Linear lab power supply

[JuanGg]

I am doing the "Learning the Art of Electronics" course and I am in need for a lab power supply. All I have is a single-channel one (discussed here: https://www.eevblog.com/forum/beginners/power-supply-ripple/) and I get by with that and a couple batteries.

I have read in multiple places that a lot can be learnt from building a power supply, so I thought I'll have a go at it.

I have a couple 15V 1A transformers and a 2x 10V 2 A one that I salvaged. All were enclosed into sealed plastic cases, so I guess current ratings will be a little higher with better dissipation. Not planing on exceding them, but good to know. I also have two 6800uF 50V caps also salvaged, which I could use for the filter caps. I will buy some new ones anyway just in case.

The specs I am aiming at are:
  -Two chanels (just two identical circuits on the same enclosure) at 0-15..20 V , 0-0.5 A
  -Hopefully arduino controlled (PWM DAC...), although I don't mind using pots for current and voltage setting, I will just use an arduino and
   some lcd / 7-segments to display voltage and current, as it's cheaper than panel meters. That would be one arduino per channel to keep both
   isolated, maybe some opto- isolated communication between them so I can do tracking or whatever.
  - 0.01 V and 1mA resolution.

What I thought:

I could maybe connect each 15 V transformer with each 10 V tap on the other one, so I will get two 25V ac outputs. Maybe taps could be switched depending on the set voltage (say 10 V tap for under 10V and the 25 V one for over that). However maybe it's not worth it given the max power disipation of +- 20 W.

My plan is to test some schematics that I found online and see how it works. I will be posting the results I get here.

I would like to start with the one attached, which I found here: https://www.electro-tech-online.com/threads/lm723-based-psu-with-min-voltage-of-0-00v.150216/page-3 (AGND just connected do GND).
I had read through the thread and also the voltage regulation chapter on The Art of Electronics and I think I got a light grasp on how the LM723 ic works. I will order the components that I haven't got around and test it out. I will only use one series pass transistor given the smaller current. Any ideas or suggestions are welcome.

Equipment that I have: The said power supply, an atx one for higher current stuff, a rigol 1054z scope, an UT61E multimeter, a 5200a function generator and recently an electronic load which is described here: https://www.eevblog.com/forum/projects/arduino-based-electronic-load/

EDIT: ALL FILES AND A PROJECT SUMMARY ARE AVAILABLE HERE:
Schematics, PCBs, firmware, 3d printed parts, front panel overlay... :
 https://github.com/Juan-Gg/Linear-lab-bench-power-supply/tree/master
A short description:
 https://juangg-projects.blogspot.com/2019/05/linear-lab-power-supply.html

Many thanks to everyone that helped me out. I hope I can give this back.
    Juan
/EDIT

[Mr. Scram]

A middle ground option could be digital pots.

[W2DML]

I recently designed a 0-12V, 0-500mA high accuracy power supply that has a computer interface for my job. I would share the design but I can't, . I recommend you look at this article for digital control of any type of regulator (linear, switching, etc.) which has resistive feedback with a DAC, might help with your design. https://www.microchip.com/forums/m688260.aspx

[lordvader88]

the LM723 is a great chip to learn from, read up on it, like I should do more of.

That design looks better than the one I've been trying to get together, I should try it too

[spec]

... I have read in multiple places that a lot can be learnt from building a power supply, so I thought I'll have a go at it...

My plan is to test some schematics that I found online and see how it works. I will be posting the results I get here.

I would like to start with the one [PSU] attached, which I found here: https://www.electro-tech-online.com/threads/lm723-based-psu-with-min-voltage-of-0-00v.150216/page-3  Any ideas or suggestions are welcome.

+ JuanGg

The wheel goes round and round

I did a lot of the design work on that PSU for the OP, so if you need to know how the various bits work, just post.

Much of the extra circuitry is there to cater for the high input voltage from the reservoir capacitors which exceeds the max VCC of the LM723.

The other complication is that the OP wanted a precision current limit rather than the standard LM723 approach and have a LED illuminate at the instant the PSU went into constant current. He also wanted the LED to have constant brightness.

My advice to you is to build a standard LM723 power supply, with standard current limiting and with no more than 35V raw voltage. Then you would only need an LM723 and three transistors: driver transistor, two power transistors. 

The physical layout is critical and you must use star point techniques and thick wires for the high current paths. Whatever you do, keep the wires to the power transistors short to avoid parasitic oscillations so you need to think about the physical layout of your PSU from day 1. I have a general physical layout that achieves all this, so if you feel you would like to take that approach just shout.

Also make sure that all components are readily accessible and can be easily changed. You can gaurentee that any component that is inaccesable will give you the most problems, even a simple resistor.

Apart from good wire routing, always remember that you can never have too big a heatsink and pay particular attention to the thermal washers between the power transistors and the heatsink. Aluminum oxide have the lowest thermal resistance, but are expensive and prone to short, ceramic is good, but the best all round choice is still thin mica. Stay clear of foam, plastic etc. Use power transistors with T03 or larger cases, and look for a low thermal resistance, junction to case. The good old 2N3055s are dirt cheap and, while not the best, are not bad for this application.

The OP had a fan on his PSU to cool the heatsink. This has a big advantage.

By the way, the OP ended up with a nice PSU with 10 turn pots for the voltage and current controls.

Have fun- the OP did building that PSU.

[spec]

Just a bit of general advice:

When you are just starting out in electronics do not try to aim too high as many of us have done. Go with something relatively simple and achievable and then build on that experience.

Newbees often think that PSUs are dead simple to build and get working: they are not. Think of a PSU as a precision, high power amplifier and remember that the LM723 has a very high voltage gain and will oscillate if you give it the chance.

The other thing is to forget about the electronics until you have built the chassis, and mounted the heatsink with fan, transformer, bridge rectifier, reservoir capacitors, controls, etc, in fact, all the big items.

The next stage is to wire up the raw supply and make sure that it is working correctly under full current load (check ripple voltage). In my experience, about 60% of cases where there have been problems with stabilized PSUs, it has been due to inadequate wiring and shortcomings of the raw supply.

When you design your power supply, one fundamental thing is to do a thermal budget to ensure that you are well below the transistor's maximum junction temperature and also check the transistors safe operating area graphs on the datasheet.

The other fundamental calculation is to do a voltage overhead budget, not forgetting ripple voltage.

The worst thing you could do for a prototype is to have bits of circuit lying around on the bench and connected by long wires.

In conclusion, do not let my lecturing put you off- with the correct approach you can build a very nice lab power supply that will last you for years. And non of the procedures that I have described are difficult. Also, there are many experienced EEV members who I am sure would be glad to advise you.

[JuanGg]

First off all, thank you everyone for answering. This is really valuable to me and I am and will be learning a lot.

A middle ground option could be digital pots.

Could be. I will try getting the thing working with regular pots, then add the microcontroller.

I recently designed a 0-12V, 0-500mA high accuracy power supply that has a computer interface for my job. I would share the design but I can't, . I recommend you look at this article for digital control of any type of regulator (linear, switching, etc.) which has resistive feedback with a DAC, might help with your design. https://www.microchip.com/forums/m688260.aspx

That article was interesting indeed. Will take note of that.

The other complication is that the OP wanted a precision current limit rather than the standard LM723 approach and have a LED illuminate at the instant the PSU went into constant current. He also wanted the LED to have constant brightness.

My advice to you is to build a standard LM723 power supply, with standard current limiting and with no more than 35V raw voltage. Then you would only need an LM723 and three transistors: driver transistor, two power transistors. 

I am more that happy with standard current limiting and no led indicator.
With 25 Vac from the transformer that would be 25*1.41 = 35 V minus a couple volts drop on the rectifier bridge, that would be 33V raw voltage.

The physical layout is critical and you must use star point techniques and thick wires for the high current paths. Whatever you do, keep the wires to the power transistors short to avoid parasitic oscillations so you need to think about the physical layout of your PSU from day 1. I have a general physical layout that achieves all this, so if you feel you would like to take that approach just shout.

Also make sure that all components are readily accessible and can be easily changed. You can gaurentee that any component that is inaccesable will give you the most problems, even a simple resistor.

I have attached a photo of the layout I am thinking off, just the size of an A4 page. Heatsinks (probably pc ones, which I have access to) and a transformer are missing, and I only "populated" one side with random components just to make provision for the board. Transformers are put in a way that evenly distributes the weight.
I am thinking of 3d printing front and back panels, joined together with aluminum profiles, to which I can attach components to in a way I can chage things easily inside later on. A sheet metal cover would complete the enclosure. I am glad to hear any suggestion.

Apart from good wire routing, always remember that you can never have too big a heatsink and pay particular attention to the thermal washers between the power transistors and the heatsink. Aluminum oxide have the lowest thermal resistance, but are expensive and prone to short, ceramic is good, but the best all round choice is still thin mica. Stay clear of foam, plastic etc. Use power transistors with T03 or larger cases, and look for a low thermal resistance, junction to case. The good old 2N3055s are dirt cheap and, while not the best, are not bad for this application.

The OP had a fan on his PSU to cool the heatsink. This has a big advantage.

I was thinking of using two separate heatsinks, one per chanel, so I could bolt directly the power transistors to them, no thermal washer needed as they would be inside the case, just some thermal paste. I have 5 3055 in TO-220 packages, will those be good enough or do I buy TO-3 ones? (not expensive at all, so I will probably get some anyway). I intend to use one temperature controlled (via the micro) fan per chanel.

By the way, the OP ended up with a nice PSU with 10 turn pots for the voltage and current controls.

Have fun- the OP did building that PSU.

I hope I will!

When you are just starting out in electronics do not try to aim too high as many of us have done. Go with something relatively simple and achievable and then build on that experience.

Newbees often think that PSUs are dead simple to build and get working: they are not. Think of a PSU as a precision, high power amplifier and remember that the LM723 has a very high voltage gain and will oscillate if you give it the chance.

The other thing is to forget about the electronics until you have built the chassis, and mounted the heatsink with fan, transformer, bridge rectifier, reservoir capacitors, controls, etc, in fact, all the big items.

The next stage is to wire up the raw supply and make sure that it is working correctly under full current load (check ripple voltage). In my experience, about 60% of cases where there have been problems with stabilized PSUs, it has been due to inadequate wiring and shortcomings of the raw supply.

I will try to make something simple enough, that can be later upgraded when I know better what I'm doing.
I will build up the enclosure and raw supply, and test it, leaving provision to add the remaining circuitry later, so I can test designs and make changes easily.

When you design your power supply, one fundamental thing is to do a thermal budget to ensure that you are well below the transistor's maximum junction temperature and also check the transistors safe operating area graphs on the datasheet.

Will take that into account.

The other fundamental calculation is to do a voltage overhead budget, not forgetting ripple voltage.

The worst thing you could do for a prototype is to have bits of circuit lying around on the bench and connected by long wires.

With 33 V raw voltage input, a 0-25 V -ish will leave plenty of overhead, even with substantial ripple. As I said, I will first build the enclosure so I can test stuff on it.

In conclusion, do not let my lecturing put you off- with the correct approach you can build a very nice lab power supply that will last you for years. And non of the procedures that I have described are difficult. Also, there are many experienced EEV members who I am sure would be glad to advise you.

Thank you very much for taking the time to answer and for all the valuable advice and sorry for not responding properly, it's late over here.

Juan

[spec]

Excellent replies: you have a good solid engineering approach. 

[Mr. Scram]

Just a bit of general advice:

When you are just starting out in electronics do not try to aim too high as many of us have done. Go with something relatively simple and achievable and then build on that experience.

Newbees often think that PSUs are dead simple to build and get working: they are not. Think of a PSU as a precision, high power amplifier and remember that the LM723 has a very high voltage gain and will oscillate if you give it the chance.

The other thing is to forget about the electronics until you have built the chassis, and mounted the heatsink with fan, transformer, bridge rectifier, reservoir capacitors, controls, etc, in fact, all the big items.

The next stage is to wire up the raw supply and make sure that it is working correctly under full current load (check ripple voltage). In my experience, about 60% of cases where there have been problems with stabilized PSUs, it has been due to inadequate wiring and shortcomings of the raw supply.

When you design your power supply, one fundamental thing is to do a thermal budget to ensure that you are well below the transistor's maximum junction temperature and also check the transistors safe operating area graphs on the datasheet.

The other fundamental calculation is to do a voltage overhead budget, not forgetting ripple voltage.

The worst thing you could do for a prototype is to have bits of circuit lying around on the bench and connected by long wires.

In conclusion, do not let my lecturing put you off- with the correct approach you can build a very nice lab power supply that will last you for years. And non of the procedures that I have described are difficult. Also, there are many experienced EEV members who I am sure would be glad to advise you.

Aren't your warnings exactly why it's a good first project? It looks simple, yet it isn't simple and you'll learn lots. Failing is part of the process.

[spec]

Just a bit of general advice:

When you are just starting out in electronics do not try to aim too high as many of us have done. Go with something relatively simple and achievable and then build on that experience.

Newbees often think that PSUs are dead simple to build and get working: they are not. Think of a PSU as a precision, high power amplifier and remember that the LM723 has a very high voltage gain and will oscillate if you give it the chance.

The other thing is to forget about the electronics until you have built the chassis, and mounted the heatsink with fan, transformer, bridge rectifier, reservoir capacitors, controls, etc, in fact, all the big items.

The next stage is to wire up the raw supply and make sure that it is working correctly under full current load (check ripple voltage). In my experience, about 60% of cases where there have been problems with stabilized PSUs, it has been due to inadequate wiring and shortcomings of the raw supply.

When you design your power supply, one fundamental thing is to do a thermal budget to ensure that you are well below the transistor's maximum junction temperature and also check the transistors safe operating area graphs on the datasheet.

The other fundamental calculation is to do a voltage overhead budget, not forgetting ripple voltage.

The worst thing you could do for a prototype is to have bits of circuit lying around on the bench and connected by long wires.

In conclusion, do not let my lecturing put you off- with the correct approach you can build a very nice lab power supply that will last you for years. And non of the procedures that I have described are difficult. Also, there are many experienced EEV members who I am sure would be glad to advise you.

Aren't your warnings exactly why it's a good first project? It looks simple, yet it isn't simple and you'll learn lots. Failing is part of the process.

It is all a matter of degree

Yes, you can learn by failure, but on the other hand, you can also be discouraged and learn nothing. It depends on your character. It also depends on your objectives: just learn, or learn and end up with a low-cost lab power supply to use for experiments. And don't forget, if the basic construction is good you can always add advanced features, like precision current control and digital meters, for example.

That is a complex power supply and the chap who built it, who was an engineer, expended quite a lot of effort to get it working.

The engineering department where I worked used to mentor the under grads and apprentices. Over the years probably 1000 went through the system, and maybe 250 started their own projects, mainly audio amplifiers but some power supplies. I think out of all those starters there was only about five audio amps completed, and they were Texan kits and no power supplies to speak off.

There are five gotchas with any power supply: star point wiring, frequency stability, max junction temperature, SOA, voltage overhead. If you learn about all of those you will have achieved an awful lot. But that is just the electronics. There is also the mechanical side: heatsink, chassis, controls, meters, fuse holder, etc etc.

So, I would advise anyone just starting in electronics to aim at the achievable, and as I said, that would be a straight forward LM723 type PSU. And, believe me, that is plenty enough of a challenge and a valuable learning experience.

[Wolfgang]

I am doing the "Learning the Art of Electronics" course and I am in need for a lab power supply. All I have is a single-channel one (discussed here: https://www.eevblog.com/forum/beginners/power-supply-ripple/) and I get by with that and a couple batteries.

I have read in multiple places that a lot can be learnt from building a power supply, so I thought I'll have a go at it.

I have a couple 15V 1A transformers and a 2x 10V 2 A one that I salvaged. All were enclosed into sealed plastic cases, so I guess current ratings will be a little higher with better dissipation. Not planing on exceding them, but good to know. I also have two 6800uF 50V caps also salvaged, which I could use for the filter caps. I will buy some new ones anyway just in case.

The specs I am aiming at are:
  -Two chanels (just two identical circuits on the same enclosure) at 0-15..20 V , 0-0.5 A
  -Hopefully arduino controlled (PWM DAC...), although I don't mind using pots for current and voltage setting, I will just use an arduino and
   some lcd / 7-segments to display voltage and current, as it's cheaper than panel meters. That would be one arduino per channel to keep both
   isolated, maybe some opto- isolated communication between them so I can do tracking or whatever.
  - 0.01 V and 1mA resolution.

What I thought:

I could maybe connect each 15 V transformer with each 10 V tap on the other one, so I will get two 25V ac outputs. Maybe taps could be switched depending on the set voltage (say 10 V tap for under 10V and the 25 V one for over that). However maybe it's not worth it given the max power disipation of +- 20 W.

My plan is to test some schematics that I found online and see how it works. I will be posting the results I get here.

I would like to start with the one attached, which I found here: https://www.electro-tech-online.com/threads/lm723-based-psu-with-min-voltage-of-0-00v.150216/page-3 (AGND just connected do GND).
I had read through the thread and also the voltage regulation chapter on The Art of Electronics and I think I got a light grasp on how the LM723 ic works. I will order the components that I haven't got around and test it out. I will only use one series pass transistor given the smaller current. Any ideas or suggestions are welcome.

Equipment that I have: The said power supply, an atx one for higher current stuff, a rigol 1054z scope, an UT61E multimeter, a 5200a function generator and recently an electronic load which is described here: https://www.eevblog.com/forum/projects/arduino-based-electronic-load/

Hi, welcome to the world of PSUs !

I have some comments to make about the design:

- much of your circuitry is busy to create a "mini"-PSU for the LM723, taking care to not exceed the 40V permanent 723 supply voltage.
  This part could be handled by a single 317HV regulator, with less components and safe input up to 60V.
- Another alternative would be a floating design (see 723 datasheets or appnotes). This would not need special precautions,
  you could implement better current sensing and the design is not voltage limited. To make this work well, however, the
  LM723 needs a little stabilized Vc supply, best derived from a small PCB transformer and an 7812 or 7815 chip.
- Current limiting will work, but is very inexact due to the temperature and manufacturing tolerances of the 723s current
  sense transistor. You could sense this by a real op-amp and then use the FC pin to shutdown the 723 in case of overcurrent.
  The current control loop also needs frequency compensation.

If you want, some 723 circuits can be found here:

https://electronicprojectsforfun.wordpress.com/power-supplies/a-collection-of-proper-design-practices-using-the-lm723-ic-regulator/

and some silly applications here:

https://electronicprojectsforfun.wordpress.com/silly-circuits/silly-circuits-a-heated-lm723-reference/

Have fun. Dont be scared off by experimenting or by committing errors. If you dont make errors, you will learn very slowwwwly.

[spec]

I have some comments to make about the design:

- much of your circuitry is busy to create a "mini"-PSU for the LM723, taking care to not exceed the 40V permanent 723 supply voltage.
  This part could be handled by a single 317HV regulator, with less components and safe input up to 60V.
- Another alternative would be a floating design (see 723 datasheets or appnotes). This would not need special precautions,
  you could implement better current sensing and the design is not voltage limited. To make this work well, however, the
  LM723 needs a little stabilized Vc supply, best derived from a small PCB transformer and an 7812 or 7815 chip.
- Current limiting will work, but is very inexact due to the temperature and manufacturing tolerances of the 723s current
  sense transistor. You could sense this by a real op-amp and then use the FC pin to shutdown the 723 in case of overcurrent.
  The current control loop also needs frequency compensation.

If you want, some 723 circuits can be found here:

https://electronicprojectsforfun.wordpress.com/power-supplies/a-collection-of-proper-design-practices-using-the-lm723-ic-regulator/

and some silly applications here:

https://electronicprojectsforfun.wordpress.com/silly-circuits/silly-circuits-a-heated-lm723-reference/

Have fun. Dont be scared off by experimenting or by committing errors. If you don't make errors, you will learn very slowwwwly.

Hmm, some interesting comments. I was not able to find any PSU schematics in the two links that you posted- just LM723 general comments and application notes. There is a world of difference between an outline schematic, like you see in application notes, and a real working PSU.

Good idea to use the LM317HV but there are still some design issues and, from what you say, to get the performance you would end up with more complexity with extra supplies mains transformers etc. But you obviously have some experience of PSU design, so would you please post a schematic with all the features that you mention. I am talking about a practical design with decoupling etc. I would be genuinely interested to see your design. 

Just to put the design in context: there were a few givens- perhaps take a look at the thread on ETO. The first is that the PSU was already built but not operating to the OP's satisfaction. The second thing is that you have over simplified the purpose of the 'busy' components which are not just there to cater for the LM723 40V Vcc limit. And finally the schematic is difficult to follow and thus difficult to understand so looks more complicated than it is. I have a rationalised schematic but the OP wanted to use his own version.

There is one problem with the posted PSU: it will not go to precisely 0V, as there is no mechanism for getting rid of the leakage current from the power transistors. I did a mod to fix that, but the OP was quite happy without it.

You over emphasize the the shortcomings of the LM723 current limit- if you look at the data sheet you will see that the tempco is -1.8mV/degC in 650mV = -0.3% degC, and that is adequate for a general purpose bench power supply, as thousands of LM723 user will testify.

As for bench power supplies in general, and given a free hand, I certainly would not use an LM723, there are much better approaches these days.

Finally, you overstate the benefit for making errors. Good design is not based on stumbling from one error to another. I would say that you learn best by taking a logical and thorough approach and learning by your successes rather than failures. And many people just make the same mistake over and over again without learning a thing.

But perhaps you are thinking about the saying, 'Anyone who has never made a mistake, never made anything'

[m3vuv]

Hi has anyone a schematic for a SG3532  based psu?,i cant find any online,cheers m3vuv.

[spec]

+ m3vuv

The SG3532 is essentially a greatly improved version of the LM723.

Because the SG3532 has the same architecture as the LM723, with a few changes here and there, you can use the SG3532 in most LM723 circuits. The main differences are a Vref of 2.5V compared to 7.15V, current limit sense voltage of 80mv rather than 650mV. Also the SG3532 has a shutdown input, which the LM723 does not have: simply disable it by connecting it to V-

The datasheet for the SG3532, which includes some PSU circuits, is attached below:

By the way, you do know that the SG3532 is obsolete (unfortunately).

[m3vuv]

yes i know its obsolete,are still available on ebay tho,the lower current sense voltage is the reason for needing one as can use a lower value current sense resistor for low current outputs.

[JuanGg]

I can't possibly respond to every single point stated, so here is a summary:

-I am a 17 year-old with next to none electronics design experience, have been playing around with breadboards, soldering iron and multimeter since I can remember, but no actual knowledge of what I am doing, hence I want to learn as much as possible. What I have some experience on is CAD (LibreCAD, FreeCAD and Fusion 360), 3d printing and coding (here is some stuff I've done: https://www.thingiverse.com/JuanGg/designs) . I am not too worried about the mechanical aspect of the power supply, I should be able to work it out.

-I want to build my own power supply because I need one and looks like there is lots to learn from it. I do not mind making mistakes and letting the magic smoke out if that teaches me something. However, I would rather do things properly, knowing what I'm doing (and the magic smoke will come out anyways...).

-I don't have any preference for the LM723 ic (although I have some on the way), just what I found online and in The Art of Electronics. I'll go for whatever you guys recommend.

-Again, the specs I am aiming at are:
    Two chanels (identical separate circuits) at 0 - 20...25 ish V and 0.5 A
    I would like to go down to 0 V, if possible.
    Some form of adjustable current limiting, doesn't need to be precise.
    Be able to controll it from a DAC, although I am ok with potentiometers if it makes it simpler.

-I have two 15 Vac 1 A and one 2x 10 V 2A transformers. I was thinking of connecting them in series so I get two isolated taps at 25 Vac, one per chanel. That would be 25 * sqrt(2) = around 35 V, minus a couple diode drops, input to the circuit.

-As spec suggested, I will start building the enclosure and the raw power supply. That is going to take me weeks for components to arrive and to design and manufacture the thing. In the meantime, some of the schematic design can be done. I will go over the linked application notes and try to make sense out of it.

Again, thank you all for the replies.

[JuanGg]

I read spec's last reply and it suddenly vanished so I can't quote it.

Don't be sorry, it's me the one that should be for not being clear. I will gladly reply to any questions asked.

What I want is two isolated supplies on the same case, no common ground, so I can connect them as I need to.

Thanks again

[spec]

I deleted the last post because I realized that you had already given the answer to my questions

But thanks for confirming any way.

Another post underway.

[spec]

-I don't have any preference for the LM723 ic (although I have some on the way), just what I found online and in The Art of Electronics. I'll go for whatever you guys recommend.

-Again, the specs I am aiming at are:
    Two chanels (identical separate circuits) at 0 - 20...25 ish V and 0.5 A
    I would like to go down to 0 V, if possible.
    Some form of adjustable current limiting, doesn't need to be precise.
    Be able to controll it from a DAC, although I am ok with potentiometers if it makes it simpler.

-I have two 15 Vac 1 A and one 2x 10 V 2A transformers. I was thinking of connecting them in series so I get two isolated taps at 25 Vac, one per channel. That would be 25 * sqrt(2) = around 35 V, minus a couple diode drops, input to the circuit.

That configuration of the three transformers to give two isolated rectified raw DC supplies is optimum, I would say, given the transformers that you have.

I would suggest that, once the metalwork is complete, wire up just one of the transformer/rectifier/ reservoir  circuits and test  just one raw supply. That will be a major step forward in your PSU build: hopefully without any blue smoke

Best, to complete just one of the stabilized supplies, then once that is working to your satisfaction start on the second PSU circuit. That way you will have just one ball in the air, rather than two.

I would aim for 1 amp output current and a ripple voltage of no more than 2V peak to peak, so a 4700uF, or higher, reservoir capacitors will be required with a working voltage of 50V or higher. The reservoir capacitors must be able to handle the ripple current, so they will be fairly large.  You would need a minimum ripple current rating of 1.8A or greater.

You can work out the value of reservoir capacitor (in Farads) for yourself if you like, by using the following formula: C=IT/Vripple, where I is the current drain (in amps), T is the half period time of your mains supply (in seconds) (10ms for Spain on 50Hz), and Vripple is the peak to peak ripple voltage (in volts).  This formula is approximate, but it is close enough for this type of PSU.

When the raw supplies are built and tested, they will be suitable for almost any stabilizing circuit, vanilla or strawberry, so they will be a valuable asset in your electronic adventures. 

Hopefully, that covers the raw supplies. I will post some thoughts about the stabilizing circuits, maybe in a couple of hours time.

[Kleinstein]

With these 3 transformers it might get a little tricky to get 2 separate identical supplies. I would more look at 2 slightly different versions: one using the 2x10 V transformer - this could end up at some 0-18 V at 1 A max. The other would than used the two 15 V transformers and could deliver up to 25,maybe 30 V, but only 0.5 A or 12 V up to 1A. Due to the filtering and not so good power factor one can expect to get a maximum DC current of about 50-65% of the AC rating.

There are some lab supply kits available like these:
https://ru.aliexpress.com/item/2016-NEW-Free-Shipping-Red-0-30V-2mA-3A-Continuously-Adjustable-DC-Regulated-Power-Supply-DIY/32660068947.html
The original plan circuit has some weak points / bugs, but it is possible to fix the worst ones (especially not using a 24 V transformer and lower current limit at some 1-1.5 A). With a few changes the circuit can give a reasonable lab supply. There are already some threads here on this type of circuit.

A LM723 circuit is also possible, but usually has a current limit that is not really accurate. Current adjustment over a large span can be tricky and might need switching the shunt. If one can live with this slight limitation it's also a possible path.

[JuanGg]

That configuration of the three transformers to give two isolated rectified raw DC supplies is optimum, I would say, given the transformers that you have.

I would suggest that, once the metalwork is complete, wire up just one of the transformer/rectifier/ reservoir  circuits and test  just one raw supply. That will be a major step forward in your PSU build: hopefully without any blue smoke

Best, to complete just one of the stabilized supplies, then once that is working to your satisfaction start on the second PSU circuit. That way you will have just one ball in the air, rather than two. ...

That is what I was thinking of, I will first build one raw supply, test it, then build one regulated supply on matrix board, test it, throubleshoot it...
If that works out, I may get some PCB's done or etch them myself, two identical ones.

According to Learning the art of electronics, I calculated the capacitor value just as you suggested, but aiming for 1 V ripple (as was in the example) This resulted in 10000 uF cap value. I have two 6800 uF 50 V caps that I salvaged I could use (It says CE Japan on them, no clue on the datasheet). Anyway, I went and bought a couple 10000 uF 50V Rubicon MXR caps of ebay (not sure if they are legit...). According to the datasheet, they have a 2.6 A ripple current rating. I also got some 8 A bridge rectifiers, LM723s and some misc stuff.

With these 3 transformers it might get a little tricky to get 2 separate identical supplies. I would more look at 2 slightly different versions: one using the 2x10 V transformer - this could end up at some 0-18 V at 1 A max. The other would than used the two 15 V transformers and could deliver up to 25,maybe 30 V, but only 0.5 A or 12 V up to 1A. Due to the filtering and not so good power factor one can expect to get a maximum DC current of about 50-65% of the AC rating. ...

I thought about doing what you propose, but I prefer having two identical ones for now, I can always change that later. Why would it be tricky to get 2 identical supplies?
I'll have a look at that kit's threads, as I said, anything that fits the purpose will be fine. However, I would like to build the thing myself, not to buy a pre-made kit. Again, I don't mind crude current limiting.

Juan

[Kleinstein]

Having a series connection of  a 10 V 2 A turn and a 15 V 1A transformer limits the current to 1 A AC and thus some 0.5-0.6 A DC.
It also needs the 3 transformers to be use together, as opposed to 2 totally separate units.

With a very large cap it would be closer to the lower value. Usually one has to accept a little more ripple, more like 5-10%.
I would consider on of the 6800 µF caps plenty for 1 A.

The circuits could be still essentially the same (e.g. same board), just a higher possible current limit for the lower voltage one, or a lower maximum voltage if the 15 V transformers in parallel.

I can understand of wanting to build the supply without a kit. However the general type of circuit would be used as a starting point. It is a little odd in some aspects, but also has some good points.

[rstofer]

There certainly seem to be some naysayers!  Add me to the list...  Longer explanation deleted except to note that the probability of success likely decreases exponentially with parts count.

And give up the idea that it's easy to cut a fat hog and produce a $500 supply for a buck ninety-eight.

[ArthurDent]

JuanGg - I followed your first power supply project with interest and although I haven't commented on this one yet, I'm watching it as well. There are a lot of suggestions from posters with many different points of view and you've done a good job of evaluating the pros and cons of the suggestions to see how they would work for what parts you have to work with.

You're making good progress.

[spec]

+ JuanGg

Attached is an outline schematic of your proposed three transformer, two power supply circuit. Could you check it to ensure that it reflects what you intend? Don't worry about the component values etc, just the architecture.

I do not know how much electronics knowledge you have, but you seem to know a lot more than I first thought. So please bear with me while I go through some points which are essential to get an accurate and stable PSU.

Just like audio power amplifiers, it is vital to connect components in the correct order and to use star points. A star point is a point of reference for certain functions in a circuit. These are indicated by SP # 1-1 etc. Also indicated by the thickness of the lines is an indication of the gauge of wire. A thick line indicated heavy current which requires low resistance wire. This is not for current handling though, it is to limit the voltage drops due to the current flowing.

I would suggest then that when you wire up the raw DC supply circuit that you follow the the same routing indicated on the schematic. The importance of the component order and the star points will be more evident when the stabilizing circuits are in place. Star point 1 is the most important as it is the voltage reference point for the whole PSU circuit.

I have also attached another schematic which illustrates the way that ordering and star points are used in a complete power supply. It also indicates the importance of decoupling (schematic is currently work in progress for another application).

[JuanGg]

Having a series connection of  a 10 V 2 A turn and a 15 V 1A transformer limits the current to 1 A AC and thus some 0.5-0.6 A DC.
It also needs the 3 transformers to be use together, as opposed to 2 totally separate units. ...

I know, I've been thinking about it. I don't really need much curent, 0.5 A would be plenty. Plus, I don't have much space in my 'lab', so I would  prefer a compact solution. And I am running out of power outlets! It seems that you can never have too many.

With a very large cap it would be closer to the lower value. Usually one has to accept a little more ripple, more like 5-10%.
I would consider on of the 6800 µF caps plenty for 1 A.

What do you mean "closer to the lower value", closer to 0.5 A than to 0.6 A ? I will use the 6800 uF ones I have.

The circuits could be still essentially the same (e.g. same board), just a higher possible current limit for the lower voltage one, or a lower maximum voltage if the 15 V transformers in parallel.

I can understand of wanting to build the supply without a kit. However the general type of circuit would be used as a starting point. It is a little odd in some aspects, but also has some good points.

Sure, just maybe a couple different component values between boards.  That one came from here : http://www.electronics-lab.com/project/0-30-vdc-stabilized-power-supply-with-current-control-0-002-3-a/ didn't it?

There certainly seem to be some naysayers!  Add me to the list...  Longer explanation deleted except to note that the probability of success likely decreases exponentially with parts count.

And give up the idea that it's easy to cut a fat hog and produce a $500 supply for a buck ninety-eight.

Constructive criticism is always welcome! I do understand that. I am not aiming for a high-end supply nor for building it for free.

JuanGg - I followed your first power supply project with interest and although I haven't commented on this one yet, I'm watching it as well. There are a lot of suggestions from posters with many different points of view and you've done a good job of evaluating the pros and cons of the suggestions to see how they would work for what parts you have to work with.

You're making good progress.

Thanks! Good to see you here. I have the other supply on hold, it works for what I need for now in its current state...

Attached is an outline schematic of your proposed three transformer, two power supply circuit. Could you check it to ensure that it reflects what you intend? Don't worry about the component values etc, just the architecture...

Yes, that is precisely what I thought. As I said, I have not much electronics knowledge, just what I have adquired as I had the need to implement into my rather simple projects. Any explanations are welcome.

I really appreciate you taking the time to do this. I am surprised at the help I am receiveing. Thank you.

[spec]

Attached is an outline schematic of your proposed three transformer, two power supply circuit. Could you check it to ensure that it reflects what you intend? Don't worry about the component values etc, just the architecture...

Yes, that is precisely what I thought. As I said, I have not much electronics knowledge, just what I have acquired as I had the need to implement into my projects. Any explanations are welcome.

I really appreciate you taking the time to do this. I am surprised at the help I am receiving. Thank you.

No sweat 

It is good that the circuit for the raw supply is now pretty much defined and that the overall architecture of the twin PSU is also defined. That is the first major milestone in the design.

It is also good that you respond quickly and intelligently to the various posts- it is not always the case on all threads 

There is one point that I forget to mention: heatsinks. I would suggest the lowest thermal resistance heatsink you can get. Also low thermal resistance insulating washers, with clamps on the power transistor/regulator case. I would also advise fan cooling.

The insulating washer has a surprisingly negative impact, so one approach is to mount directly to the heatsink. This does mean that the heatsink will be at the same voltage as the power component so the heatsink would need to be insulated from the other components and chassis/PSU case.

Another problem with a live heatsink is that it acts a huge capacitor and can cause parasitic oscillations by coupling to other components. With a bit of simple screening this is easy to fix though.

About the stabilizing circuit, here, as I see it, are the four options.
[1] LM723 based

[2] Three terminal regulator based

[3] Design available from books, internet etc

[4] Custom

LM723 BASED
The LM723 is a great PSU building block, but it has some limitations which mean that a LAB PSU using the LM723 gets complicated. Going down to 0V requires extra components. The voltage overhead is high, the absolute voltage is limited and the current limiting is basic, but probably adequate.

THREE TERMINAL REGULATOR BASED
This is the simplest and quickest approach, if the standard non-adjustable current/temperature limiting is acceptable. As you can only dissipate about 12W in a LM317 or LM337 (with a good thermal washer, case clamp, and high performance heatsink + fan cooling), before the current limiting kicks in. The PSU would produce around 1A with an input/output voltage difference (IOD) of about 12V, but as the IOD increased the current output would limit proportionally reducing to 400mA at 0V output. I can probably post a schematic of this approach if you like.

Of course you could select a 10A regulator which would give a touch more current but not that much more because of the same thermal factors.

DESIGN AVAILABLE FROM INTERNET ETC
There are 100s of designs available, but I have not been able to find one that is well proven or simple. Maybe the other EEC members could advise here.

CUSTOM
This approach has been suggested in other EEV threads about PSU construction. I think I have read them all. The general feeling is that this approach has the potential for the best performance with the simplest circuit. It promises precise voltage control, precise current control, and a simple technique to allow the output voltage range to include 0V. Overall it is probably the cheapest approach too.

Out of interest, I outlined a circuit which should do the job (famous last words). If you are at all interested in this approach I could post the schematic for you to consider. One big factor to take into account with new designs, that have not been prototyped and tested is that they may not work at first and may need some development work.

So there it is. Could you give the approach some thought and post what you think?

And finally a lecture about protection. Many of the circuits that you see around have no protection and are not worst case designs. There is often no reverse voltage protection, no protection for the emitter/base junction of the output transistor, and no over-current or over-voltage protection on the input. How do you feel about protection? Some protection can be bolted on later, but some needs to be incorporated at the architecture stage.

[ArthurDent]

Quite often the simplest circuits work well enough to cover almost all your needs. Here is a 0-30VDC/1A uA723 circuit I found that looks pretty good. You could scale it for other voltages or just use it as a starting point for your own design. They use a fairly simple circuit to generate the necessary negative voltage the uA723 needs to allow the output to go down to zero volts.

As has been pointed out trying to use common grounding points is a good idea but if you are using adequate size wire at 1A any voltage errors caused by drops along a lead probably won't be significant compared to temperature drift, etc.. If you are using an adjustable current limit where you have sub-ohm sensing resistors with higher currents that's where you really have to be careful about adding stray lead resistance that could have a large effect on your readings. Some of the better power supplies use 4-terminal resistors to get higher precision current adjustments. 

I might use a larger filter capacitor than the 1000/63 shown in the attached schematic but one of the previous circuits showed 4x4700 used which I believe to be overkill. All regulators also acts as capacitance multipliers by a factor of the combined hfe of the series transistors and that allow you to use smaller capacitors.

[Wolfgang]

Quite often the simplest circuits work well enough to cover almost all your needs. Here is a 0-30VDC/1A uA723 circuit I found that looks pretty good. You could scale it for other voltages or just use it as a starting point for your own design. They use a fairly simple circuit to generate the necessary negative voltage the uA723 needs to allow the output to go down to zero volts.

As has been pointed out trying to use common grounding points is a good idea but if you are using adequate size wire at 1A any voltage errors caused by drops along a lead probably won't be significant compared to temperature drift, etc.. If you are using an adjustable current limit where you have sub-ohm sensing resistors with higher currents that's where you really have to be careful about adding stray lead resistance that will could have a large effect on your readings. Some of the better power supplies use 4-terminal resistors to get higher precision current adjustments.   

.. you could further simplify this by using an 7905 regulator for the negative rail. It would also improve stability.

[JuanGg]

No sweat 

It is good that the circuit for the raw supply is now pretty much defined and that the overall architecture of the twin PSU is also defined. That is the first major milestone in the design.

It is also good that you respond quickly and intelligently to the various posts- it is not always the case on all threads 

There is one point that I forget to mention: heatsinks. I would suggest the lowest thermal resistance heatsink you can get. Also low thermal resistance insulating washers, with clamps on the power transistor/regulator case. I would also advise fan cooling. ...

It would be unwise not to reply quickly, as I am the first one interested in getting this thing working

I was thinking of using CPU cooling heatsinks, which usually have a fan attached to. I think the easisest approach is no thermal washers and mount them inside the case. I can esily make some plastic brackets for that purpose. (abs, so they won't melt, even if the heatsink is at 100 ºC, which it shouldn't be...). Insulating washers can be added later on if needed. I will hopefully have news about the heatsinks in a couple days time.

On the stabilizing circuit, as you present it, the "Custom" option seems the best, if it delivers good performance on a simple circuit, and thus more likely for me to better understand what is going on. I don't mind prototyping and testing. Going through the whole design process is something I would like to do.

Regarding protection, it being a lab psu it should be a litle bit foolproof (to use). If adding some is a matter of including a couple diodes here and there, a fuse or an SCR, then for sure.

Quite often the simplest circuits work well enough to cover almost all your needs. Here is a 0-30VDC/1A uA723 circuit I found that looks pretty good. You could scale it for other voltages or just use it as a starting point for your own design. They use a fairly simple circuit to generate the necessary negative voltage the uA723 needs to allow the output to go down to zero volts.  ...

That looks simple enough and it will fit my purpose just by changin a couple resitors for the 0.5 A current. About capacitors I have two salvaged 6800uF /50V and another couple 10000 uf / 50V on the way.

Regarding components, could you guys propose some common ones that will likely be used in any circuit I will end up building? I have 5x 3055 transistors, some op-amps, assorted resistors, diodes and so on, but nothing about zeners or other things. This is so I can order stuff now so it arrives when I need it. Anything I don't use can go the the parts bin, and else I can resort to the local elcetronics shop.

Juan

[ArthurDent]

"Regarding protection, it being a lab psu it should be a litle bit foolproof (to use). If adding some is a matter of including a couple diodes here and there, a fuse or an SCR, then for sure."

One problem with power supplies with crowbar protection circuits is that some people use their power supplies for battery charging. I've seen spikes on the output lines or limit adjustment errors cause the crowbar to trigger causing the output leads to short. The battery will then deliver a large amount of current into the supply blowing the SCR circuitry and quite often burning runs on circuit boards.

I have mixed feelings about having crowbar circuits in adjustable bench type power supplies, I have some with and most without. If I were using a fixed supply built into a piece of equipment where I was not changing the load then I'd be comfortable using a crowbar but probably not in an adjustable bench supply I was building.

A lot of the newer supplies have MOSFETs for the pass elements and sense if the output is over a set point and turn the output off rather than shorting the output and this is much nicer. Here is how two of my supplies allow you to set OVP/OCP points.

[JuanGg]

One problem with power supplies with crowbar protection circuits is that some people use their power supplies for battery charging. I've seen spikes on the output lines or limit adjustment errors cause the crowbar to trigger causing the output leads to short. The battery will then deliver a large amount of current into the supply blowing the SCR circuitry and quite often burning runs on circuit boards.
...

Given that, we can rule crowbar protection out for now. It is something that can be added later, and it will only prevent the output from going ,say, + 25 V. If the series pass transistor fails short, and I am powering something at 5 or 12 V, it will kill it anyway...

[spec]

Quite often the simplest circuits work well enough to cover almost all your needs. Here is a 0-30VDC/1A uA723 circuit I found that looks pretty good. You could scale it for other voltages or just use it as a starting point for your own design. They use a fairly simple circuit to generate the necessary negative voltage the uA723 needs to allow the output to go down to zero volts.

Brilliant 

As has been pointed out trying to use common grounding points is a good idea but if you are using adequate size wire at 1A any voltage errors caused by drops along a lead probably won't be significant compared to temperature drift, etc.. If you are using an adjustable current limit where you have sub-ohm sensing resistors with higher currents that's where you really have to be careful about adding stray lead resistance that could have a large effect on your readings. Some of the better power supplies use 4-terminal resistors to get higher precision current adjustments.

That is true about the voltage drops to an extent, but there is also frequency stability to consider. Using star points costs nothing so why not do it? (don't forget the huge gulps of current flowing into the reservoir capacitor)

I might use a larger filter capacitor than the 1000/63 shown in the attached schematic.

Yeah, 1mF is too small and will give a large ripple voltage. 4.7mF upwards would be better and give a peak to peak ripple of 2v rather than 10V which is not workable. Reservoir capacitors normally have a tolerance of -40% +100% so you would need to factor that in. So 6.8mF upwards would be good. 10mF would be ideal, giving a ripple voltage of 1V peak to peak.

one of the previous circuits showed 4x4700 used which I believe to be overkill. All regulators also acts as capacitance multipliers by a factor of the combined hfe of the series transistors and that allow you to use smaller capacitors.

I would say that 10mF would be ideal: low ripple voltage, giving more voltage overhead margin and less ripple voltage on the output.

Just one point about the PSU circuit- there is no decoupling.

[spec]

Quite often the simplest circuits work well enough to cover almost all your needs. Here is a 0-30VDC/1A uA723 circuit I found that looks pretty good. You could scale it for other voltages or just use it as a starting point for your own design. They use a fairly simple circuit to generate the necessary negative voltage the uA723 needs to allow the output to go down to zero volts.

As has been pointed out trying to use common grounding points is a good idea but if you are using adequate size wire at 1A any voltage errors caused by drops along a lead probably won't be significant compared to temperature drift, etc.. If you are using an adjustable current limit where you have sub-ohm sensing resistors with higher currents that's where you really have to be careful about adding stray lead resistance that will could have a large effect on your readings. Some of the better power supplies use 4-terminal resistors to get higher precision current adjustments.   

.. you could further simplify this by using an 7905 regulator for the negative rail. It would also improve stability.

There is a bit more to starpoints than just voltage drops. Bad wiring can also lead to frequency instability, just like in audio amps. And there is no reason not to do a proper job - it costs nothing

Good suggestion about using a three terminal regulator for the -5V reference supply.

[Kleinstein]

One can protect the crowbar circuit with an extra fuse. So there would be a fuse between the crowbar and the output terminal. Someone adding a strong external supply would only blow the fuse. This type of fuse might be a good idea anyway to avoid a "short" through the diode across the output - in case the battery is the other way around.
Well set the crowbar can be closer to the actually set voltage and actually protect the circuit, if no too sensitive.

The LM723 circuit with the negative supply shown below has 2 slight problems:
There is a supply voltage limit for the 723 - so with 30 V output voltage this would likely cause too much voltage at least under no load conditions. With an output level above about 25 V it might be a good idea to have some voltage limitation / regulation for the LM723. Often an extra filter cap is also a good idea, as this allows some extra 2 V more ripply on the main supply current and thus maybe half the capacitor there.

The 723 only needs a rather low negative voltage, if at all. So something like -2 V should be enough.

In that circuit the negative voltage is part of the reference. One of the good points of the 723 is it's usually low noise reference.

The simpler way to use the 723 in a supply that goes all the way down to 0 V (or at least very close) is to have the feedback divider no towards 0 V but towards a small fraction (e.g. around 1.5 V) of the reference level and no extra negative supply.  The little current flowing back from the divider might limit the lowest voltage to get open circuit to some 50 mV, depending on the minimum internal load.

It is also better to adjust the reference level and keep the feedback divider constant. I know many old supplies do it the other way round, but it's not that clever, because it changes the loop gain. Adjusting the reference side is also more convenient if a digital control is used.

[spec]

I was thinking of using CPU cooling heatsinks, which usually have a fan attached to. I think the easiest approach is no thermal washers and mount them inside the case. I can esily make some plastic brackets for that purpose. (abs, so they won't melt, even if the heatsink is at 100 ºC, which it shouldn't be...). Insulating washers can be added later on if needed. I will hopefully have news about the heatsinks in a couple days time.

Using a CPU heatsink and fan is a good idea but it will limit your PSU. I am thinking that you should be able to make a 0V to 20V supply (x2) with the transformers and reservoir capacitors that you have already.

On the stabilizing circuit, as you present it, the "Custom" option seems the best, if it delivers good performance on a simple circuit, and thus more likely for me to better understand what is going on. I don't mind prototyping and testing. Going through the whole design process is something I would like to do.

Arthur Dent's circuit is really neat and with a few mods: 6.8mF reservoir capacitor, three terminal -5V regulator, decoupling, a low thermal resistance heatsink, and to say it yet again, star point wiring, you will have a pretty good power supply that will be a useful piece of lab equipment  for the future.

But, as you are still interested in a custom approach, I will post an outline schematic to give you a feel for the custom PSU design mentioned in reply #26. As well as an adjustable precision voltage output, it also has adjustable precision current limiting. The OV output point is pretty accurate too.

[iMo]

I like 723 but I think you can do the same with discrete parts easily (and maybe better). Also from my experience the important things your PSU should manage well are:
. it has to sustain a time unlimited short at its output, and,
. produce 0V (or close to zero) when your output voltage potentiomer's wiper looses its contact.

For example a cheapo LM358 (or similar) with 78L05/8/9 as the ref source will do the same. And the over-current protection is done by a single NPN in 723, you may do it better with the second opamp in the package. And you may "filter" the ref source with RC in the same manner as with 723, in case you require uV noise

[spec]

NOTE: In the attached PSU schematic, D8 is the wrong way around. Thanks to Arthur Dent for spotting (2018_11_18).

The circuit is not ready for prototyping and may have some errors- but it is the principle that matters at this stage. The known shortcomings are:

[1] Too much loop gain

[2} No optimization

[3] Inadequate decoupling

[4] Frequency compensation components not shown

[5] Resistor values not normalised

[spec]

Wow, this thread is attracting a lot of interest, as is normally the case with PSUs. 39 posts so far, and a lot more to come, no doubt

[JuanGg]

One can protect the crowbar circuit with an extra fuse. So there would be a fuse between the crowbar and the output terminal. Someone adding a strong external supply would only blow the fuse. This type of fuse might be a good idea anyway to avoid a "short" through the diode across the output - in case the battery is the other way around.
Well set the crowbar can be closer to the actually set voltage and actually protect the circuit, if no too sensitive.

That would do it then. I had no idea the crowbar voltage could be adjusted to the set voltage. I am not planing on using the supply to charge batteries or anything like it, but better safe than sorry.

The simpler way to use the 723 in a supply that goes all the way down to 0 V (or at least very close) is to have the feedback divider no towards 0 V but towards a small fraction (e.g. around 1.5 V) of the reference level and no extra negative supply.  The little current flowing back from the divider might limit the lowest voltage to get open circuit to some 50 mV, depending on the minimum internal load.

It is also better to adjust the reference level and keep the feedback divider constant. I know many old supplies do it the other way round, but it's not that clever, because it changes the loop gain. Adjusting the reference side is also more convenient if a digital control is used.

Going close enough to 0 V should be fine. My original intent was to also be able to controll it from a micro, as I am going to use one for the displays. That would do.

Using a CPU heatsink and fan is a good idea but it will limit your PSU. I am thinking that you should be able to make a 0V to 20V supply (x2) with the transformers and reservoir capacitors that you have already.

I just got hold of the heatsinks! (photo attached). Two identical ones with a 12 V 0.28 A fan each. They look pretty substantial, don't know if it would be enough. No markings or anything, so no idea about thermal resistance.

Arthur Dent's circuit is really neat and with a few mods: 6.8mF reservoir capacitor, three terminal -5V regulator, decoupling, a low thermal resistance heatsink, and to say it yet again, star point wiring, you will have a pretty good power supply that will be a useful piece of lab equipment  for the future.

But, as you are still interested in a custom approach, I will post an outline schematic to give you a feel for the custom PSU design mentioned in reply #26. As well as an adjustable precision voltage output, it also has adjustable precision current limiting. The OV output point is pretty accurate too.

It surely will do, plus the modifications proposed by Kleinstein for it to be digitally controlled and removing the negative rail. I am doubt which way to go. I can maybe have a go at both approaches...

Attached, below, is an outline schematic for the custom PSU mentioned in reply #26. ...

Thank you. I kind of understand what everything does in the circuit. Not sure about micro control with the negative 12V rail though. Again, don't know wich way to go, the LM723 one or something more like this one. I may as well try both. For now, I will model the transformers and the heatsinks on the computer so I can start designing the case. I hope the AC mains connector and the bridge rectifier arrive soon (a week or two...) so I can assemble the raw supply.

Wow, this thread is attracting a lot of interest, as is normally the case with PSUs. 39 posts so far, and a lot more to come, no doubt

It certainly is. Thank you to anyone that is taking part.

[Kleinstein]

The circuit in #38 has still quite a few weak points. So it is a starting point at best. It's still in between the 2 main circuit types.

With a lab supply designed essentially from scratch it is a good idea to use a simulation (e.g. LTspice) to do the first tests and adjustment if the loop. It is a really powerful tool for this.

Starting from 25 V AC is just at the border, where one can still use the LM723 or an OP as voltage mode controller and emitter follower. One might still need to limit the supply for the LM723 or the OP, but the voltage swing to some 25 V is sufficient. The LM723 is not that bad: it has a reasonable good reference and the difference amplifier with access to the transconductance part can also be handy. The main drawback it the limited voltage and crude current limit - so not really a lab supply in this respect.

The other circuit type uses an output stage that is more current controlling. However this usually needs an extra auxiliary supply for the control part and thus another transformer. This type of circuit is commonly used in many commercial supplies, as it is very flexible. However it usually needs a larger capacitor at the output. 

[spec]

The circuit in #38 has still quite a few weak points. So it is a starting point at best. It's still in between the 2 main circuit types.

There is no circuit in answer #38. I take it that you mean answer #37. I really cannot see that your general  and sweeping statements from on high are of any help. And your negative statements are unwarranted. In fact, they are disruptive.

What are these weak points- please be specific. If you are thinking about the points I have already listed, what is the objective of mentioning them again. Besides that circuit, as clearly stated, is not complete. It is an outline circuit posted at this stage to show the architecture and scope. Just for the record, that architecture, which has been recommended a number of times on EEV and elsewhere has a lot going for it in terms of flexibility, cost and, most of all, performance. The other thing about it is that it can easily be built in a negative version, which is usefull when you need plus and minus supplies in one PSU, using a single transformer.

What does the statement "It's still between the 2 main circuit types mean". What 2 main circuit types? Anyway, this sounds like another instance of opinion being stated as fact. Besides which, it is up to the OP to decide which circuit to go for.

"So it is a starting point at best"  I cannot see the point of this statement.

Finally, I take it from your tone that you are an expert in PSU design and construction. So if you can post a schematic for a PSU, that is complete, practical, and meets your own undefined criteria? That would be a big help. Perhaps include a simulation too.

[xavier60]

What does the statement "It's still between the 2 main circuit types mean". What 2 main circuit types?

Too me the two types are whether the voltage control op-amp has to tolerate the full unregulated supply voltage or not.
With the type often referred to as the "floating type" , everything is referenced to the positive output terminal including the +/- control power supply rails. The op-amps see only the control rail voltages regardless of what the unregulated and output voltages are. 

[xavier60]

I see that the circuit in #37 is the exception. The op-amps have there own supply rail that is independent of the unregulated voltage.

[spec]

… Not sure about micro control with the negative 12V rail though. Again, don't know which way to go, the LM723 one or something more like this one. I may as well try both.

I would not worry about controlling a PSU with a negative supply- it is dead easy and not an issue.

All linear stabilized PSUs are essentially voltage feedback power amplifiers in their normal mode and you can alter the voltage feedback arrangements to suit your requirements. But when the PSU goes into the constant current mode you need to consider the frequency stability in that mode too. There is a quite difficult situation when both voltage feedback and current feedback are operating at the same time (there are techniques to sort this), especially with the LM723 single transistor approach.

I would not be put off by the general statements about circuit design, which sound very learned but often do not take in the whole picture. A circuit needs to be analyzed in it's entirety to reach a proper conclusion.  One example is K's treatise where he warns of the danger of varying the feed back voltage rather than the reference voltage. Well, perhaps the millions of three terminal regulators in use ought to be withdrawn and redesigned. You simply cannot design by platitudes,

As to the lecture about referencing the feedback on an LM723 back to the reference voltage, that has already been shown in the schematic in the original post. Besides which it is a well known technique in general.

K has also voiced some undefined warnings about your three transformer approach, which is difficult to fathom.

There seems to be a lot of fear and loathing about having a negative supply, but a negative supply is a low technical risk and costs little in money and complexity terms. A negative supply has a lot of advantages, especially as it allows the PSU to go to exactly to 0V which is a big issues for some lab testing. It can also be used (as in PSU #37) to keep the temperature of the control electronics more constant as they will not be exposed to all the fire and brimstone going on above them with the raw supply and the power transistors heating up and cooling down.

Don’t forget that the output transistors will have quite a high leakage current, especially when the junction heats up and you need to get rid of that current.

Another advantage of a negative supply is that it allows you to have some current sinking which is useful.

Some PSUs without a method of draining off the various standing currents get into an unhappy state around zero volts output.

Having said all this, I am not advocating the answer #37 PSU. I leave that up to you? But, as mentioned before, the Arthur Dent circuit #27 (with modifications) has a lot going for it in my book.

[spec]

What does the statement "It's still between the 2 main circuit types mean". What 2 main circuit types?

Too me the two types are whether the voltage control op-amp has to tolerate the full unregulated supply voltage or not.
With the type often referred to as the "floating type" , everything is referenced to the positive output terminal including the +/- control power supply rails. The op-amps see only the control rail voltages regardless of what the unregulated and output voltages are.

Yes, that is a good general categorization.

[spec]

I see that the circuit in #37 is the exception. The op-amps have there own supply rail that is independent of the unregulated voltage.

Yes again That is one of the major advantages of that architecture and it brings huge performance benefits, as I mention in answer #44

Interestingly, with a few mods here and there, that architecture can be turned into a high precision calibration type PSU/voltage source. 

[spec]

JuanGg

The PSU in answer #36 may be suitable for your needs: https://www.eevblog.com/forum/beginners/help-me-design-a-psu/msg1971227/#msg1971227

It is certainly the cheapest and simplest approach.

Once you complete the chassis and raw power supply you could quickly build this circuit. Then, as you go along, you could enhance the performance in a number of ways which we could discuss on EEV.

One enhancement would be to fit a bypass power transistor to improve the current output at PSU low output voltages.

Another enhancement would be to replace the LM317 with a better three terminal regulator.

[spec]

There is a topic which I have been meaning to mention.

With the transformers arrangement you have with 10V and 15V windings it may be an advantage to have a range switch which selects just one of the windings for PSU low output voltages and both windings for high output voltages. This is common practice on many linear supplies, either manually or automatically. On my linear lab PSUs the switching is automatic by a relay.

The reason for this switching is to reduce the power dissipation in the output power transistors.

[not1xor1]

I would not be put off by the general statements about circuit design, which sound very learned but often do not take in the whole picture. A circuit needs to be analyzed in it's entirety to reach a proper conclusion.  One example is K's treatise where he warns of the danger of varying the feed back voltage rather than the reference voltage. Well, perhaps the millions of three terminal regulators in use ought to be withdrawn and redesigned. You simply cannot design by platitudes,

Hi
Are you referring to Kleinstein's message? 
Sorry... but he is right and you are wrong.
Platitudes???
It doesn't matter if a theory is fancy of fashionable, but if it is confirmed by the facts.
And if you change the loop gain the phase margin does change while it generally doesn't if you just change the input signal level (in this case the voltage reference).

Regarding 3 terminal regulators I guess you are referring to the ubiquitous LM317 ?!?!
Well its feedback is fixed, not variable.
You just program a fixed current via the resistor between output and adjust which sets the output voltage via the resistor between adjust and the ground reference (just the plain, old, boring and unfashionable Ohm Law  ).

For that reason you can add a 10µF or even greater capacitor in parallel to that resistor (adj-ground) without affecting stability, while the output capacitance may cause ringing or oscillations.

[spec]

I would not be put off by the general statements about circuit design, which sound very learned but often do not take in the whole picture. A circuit needs to be analyzed in it's entirety to reach a proper conclusion.  One example is K's treatise where he warns of the danger of varying the feed back voltage rather than the reference voltage. Well, perhaps the millions of three terminal regulators in use ought to be withdrawn and redesigned. You simply cannot design by platitudes,

Hi
Are you referring to Kleinstein's message? 
Sorry... but he is right and you are wrong.
Platitudes???
It doesn't matter if a theory is fancy of fashionable, but if it is confirmed by the facts.
And if you change the loop gain the phase margin does change while it generally doesn't if you just change the input signal level (in this case the voltage reference).

Regarding 3 terminal regulators I guess you are referring to the ubiquitous LM317 ?!?!
Well its feedback is fixed, not variable.
You just program a fixed current via the resistor between output and adjust which sets the output voltage via the resistor between adjust and the ground reference (just the plain, old, boring and unfashionable Ohm Law  ).

For that reason you can add a 10µF or even greater capacitor in parallel to that resistor (adj-ground) without affecting stability, while the output capacitance may cause ringing or oscillations.

It is not a case of right or wrong it is a matter of usefulness. You are doing the same thing- going on about theory in isolation without looking at the overall picture. There are many power supplies that use both kinds of feedback.

By the way, not that it matters, but are you sure about the LM317 feedback being fixed?

Can you explain your mention of capacitors- I don't understand what you are saying?

[Kleinstein]

The Forum software seems to have a small bug: depending on the settings (Show most recent posts at the top. under Forum Look and layout), the numbering of the answers is off by 1.  This causes the confusion with the numbering.

The circuit under Reply 37 (with recent post at bottom)  has the output stage in between a low impedance (voltage setting) and high impedance (current setting). These are the 2 main classes I spoke off. The low impedance output stages are like the conventional voltage regulators, while the current setting stages are the low drop ones.  Generally these two types use a different type of compensation, which might make the still missing compensation a real challenge.  Getting the compensation right is about the main difficulty in a lab supply.

The voltage adjustment in the feedback divider is another problem. This changes the loop gain depending on the set voltage, so the regulation would be fast at low set voltage and slower at a higher voltage. There are other supplies working this way, but it's still a weak point. It is relatively easy to solve by having a fixed divider and adjusting the voltage on the other side of the OP.

For a precision current regulation the OPs should get a separate negative supply from the reference: the current through the "GND / ADJ" Pin of the -12 V regulator also flows through the current shunt and depending on the regulator this current depends on the load current and thus the controlling current from the OPs. Besides precision this might also effect stability of the current regulation.

As shown the ripple rejection could be a problem, because the collector side of Q2 would be heavily influenced by the raw voltage. It depends on the details of the compensation however.

The high loop gain due to the transistor Q2 could be a real problem: the OPs kind of need some extra gain, as there supply is smaller than the output range. This also makes the compensation really tricky.

The current regulator works with the OP at its upper supply edge. This can be tricky with some RR OPs. So one has to be careful with the choice of OP here. Especially at some 1-2 V below the upper limit RR OPs may behave a little odd.  The voltage regulator also works at it upper supply - though this might change anyway.

For the relatively low power level in question here (e.g. 0.5 A and up to about 30-35 V raw voltage) I don't think one would need switching of the transformer tap. It could be a good idea with higher power (e.g. more than 1-2 A) though.
At some 15 W of worst case power loss, cooling it not that difficult either and one could getaway without a fan.

[nemail2]

Take a look at my Lab PSU - so far everything working as intended. Tried to make it oscillate but wasn't successful yet.
https://github.com/mamama1/LabPSU_Darlington/blob/master/Hardware/schematics.pdf

It uses a DAC to drive a voltage control opamp at a rather hilarious gain (17.14something) which controls the bases of two BD139/TIP3055 Darlingtons.
The voltage control loop is coupled to the current control loop through a diode and does high-side current measuring with a 100mOhm shunt and a MAX4080 current measuring IC. the current control loop is controlled via a DAC as well. Because the DAC can only output 1,195V at max, I had to up that a bit using another opamp and a gain of 1.713something.

In the end, this PSU goes from 50mV to 20.48V in 5mV steps and from ~10mA (that's pretty much what the fixed minimum load which is at the output consumes) to 4.096A in 1mA steps. Precision is about 0.1% over the whole range, however I didn't verify that thoroughly.

I really like the idea of the high side shunt + the MAX4080, took that from Daves µSupply so credits to him.

[JuanGg]

I can't keep up with the pace this is going at!

With a lab supply designed essentially from scratch it is a good idea to use a simulation (e.g. LTspice) to do the first tests and adjustment if the loop. It is a really powerful tool for this.

Learning LTSpice has been in my to do list for a while, I guess it's time to do it now.

I would not worry about controlling a PSU with a negative supply- it is dead easy and not an issue. ...

Ok. Good to know. Between the lm723 and the custom approach, if the custom one provides more flexibility and better performance, I' ll go for it. I can give the lm723 a try if the other doesn't work out.

JuanGg

The PSU in answer #36 may be suitable for your needs: https://www.eevblog.com/forum/beginners/help-me-design-a-psu/msg1971227/#msg1971227

It is certainly the cheapest and simplest approach.

I could very well start by doing something like that, as I dont have any experience on power supplies, but it doesn't have adjustable current limit and I don't know how would it perform. As I am planing on spending quite some time making a decent case, adding micro-controlled voltage and current displays, I would like to do something more sophisticated. However, I can always upgrade it later.

There is a topic which I have been meaning to mention.

With the transformers arrangement you have with 10V and 15V windings it may be an advantage to have a range switch which selects just one of the windings for PSU low output voltages and both windings for high output voltages. This is common practice on many linear supplies, either manually or automatically. On my linear lab PSUs the switching is automatic by a relay.

The reason for this switching is to reduce the power dissipation in the output power transistors.

I mentioned that in my first post, maybe a relay driven by a comparator or even the microcontroller would do. At a maximum power disipation of around 15W/channel I don't know if it's worth it, but it wouldn't be hard to implement.

Other thing I was thinking of is that I am going to need a +12V rail to power the fans, a possible relay and the Arduino. One way I found to do this is as in the schematic attached (https://electronics.stackexchange.com/questions/153399/using-all-the-taps-of-the-transformer-at-the-same-time), using the two transformers instead of a center tapped one, and a 7812 regulator. This way the fan and anything else would draw current from the 10 V 2 A transformer. A 10V rail would do, getting regulated 12V from 10 V ac does't seem easy.

Take a look at my Lab PSU - so far everything working as intended. Tried to make it oscillate but wasn't successful yet.
https://github.com/mamama1/LabPSU_Darlington/blob/master/Hardware/schematics.pdf

That is a bit complicated-looking. I will have a read of the schematics and documentation when I am able to.

I fould a bridge rectifier I took out of a broken power supply a while ago. I will start making the case and try the raw power supply soon.
Juan

[nemail2]

Take a look at my Lab PSU - so far everything working as intended. Tried to make it oscillate but wasn't successful yet.
https://github.com/mamama1/LabPSU_Darlington/blob/master/Hardware/schematics.pdf

That is a bit complicated-looking. I will have a read of the schematics and documentation when I am able to.

In fact, it is as simple as it gets. I have to admin, however, that the readability of the schematics isn't too good because I didn't draw it very well.
But the concept is really easy and basic. It is a microcontroller controllable PSU with constant voltage and constant current mode. Whether it is very precise or not, simply depends on the parts chosen (high precision resistors for shunt, voltage dividers, feedback loop and high precision opamps).

[spec]

UPDATED: 2018_11_18

The circuit under Reply 37 (with recent post at bottom)  has the output stage in between a low impedance (voltage setting) and high impedance (current setting). These are the 2 main classes I spoke off. The low impedance output stages are like the conventional voltage regulators, while the current setting stages are the low drop ones.  Generally these two types use a different type of compensation, which might make the still missing compensation a real challenge.  Getting the compensation right is about the main difficulty in a lab supply.

I have no idea what you mean here. The voltage and current opamps drive Q2 emitter, and hence the output darlington transistor in  the same way for both voltage and current sense modes. Of course you have to compensate a feedback circuit. That is a standard procedure for all PSUs, in fact all feedback systems.

Incidentally, the handover from the constant voltage mode to the constant current mode is done by a diode AND gate and uses the reducing gm technique to help prevent the two control loops fighting with each other. The way that this is done is one of the features of this architecture.

The voltage adjustment in the feedback divider is another problem. This changes the loop gain depending on the set voltage, so the regulation would be fast at low set voltage and slower at a higher voltage. There are other supplies working this way, but it's still a weak point. It is relatively easy to solve by having a fixed divider and adjusting the voltage on the other side of the OP.

That is not true. The open loop voltage gain will be around 110dB (when the voltage gain of Q2 is reduced to around 1 in the future) whereas the change in feedback ratio amounts to a mere 2.5dB. You would get that much gain change between individual opamps of the same type, and with temperature changes. You also seem to be missing the point that the current through the 12K resistor is constant at 1mA.

I would like to see your calculations that prove that the voltage stabilization would be inadequate. I suggest that you would find the voltage stabilization would be highly accurate, way better than the average designs you see on the net etc.

Incidentally, PSU #27 has the same feedback arrangement, yet you said not a word about that.

For a precision current regulation the OPs should get a separate negative supply from the reference: the current through the "GND / ADJ" Pin of the -12 V regulator also flows through the current shunt and depending on the regulator this current depends on the load current and thus the controlling current from the OPs. Besides precision this might also effect stability of the current regulation.

The three terminal regulator is not specified so how can you make any comments about its performance. Besides which, you greatly exaggerate the situation in a negative way. There are three terminal regulators with a ground current of 10uA. Anyway you are talking in terms of a precision voltage calibration supply, not a lab supply.

As shown the ripple rejection could be a problem, because the collector side of Q2 would be heavily influenced by the raw voltage. It depends on the details of the compensation however.

Of course, just the same as millions of other PSUs that use that architecture. Remember this is not intended to be a precision reference PSU. It is a general purpose lab PSU. Once again can I stress that this is not a complete optimized design. You must know that it is a relatively simple matter to reduce the effects of ripple, if necessary.

But I do agree that this area could be improved.

The high loop gain due to the transistor Q2 could be a real problem: the OPs kind of need some extra gain, as there supply is smaller than the output range. This also makes the compensation really tricky.

Everything is a problem with you. But what you are talking about are normal design procedures to solve so called problems. Besides which, the gain controlling elements are not included in the circuit- it is an outline circuit.

The current regulator works with the OP at its upper supply edge. This can be tricky with some RR OPs. So one has to be careful with the choice of OP here. Especially at some 1-2 V below the upper limit RR OPs may behave a little odd.  The voltage regulator also works at it upper supply - though this might change anyway.

This is just general lecturing. The opamp that is specified, OPA191, is quite happy working at VCC. Incidentally both the current and voltage  opamps are working at VCC. Besides, it would be a simple matter to change the opamp for one of the many over the rail types.

You do not know what the regulator is, so how can you make such a statement. Besides which, it is not a problem that can't be easily fixed. But you are not correct even if you are assuming an LM337, which has a maximum Vin to Vout limit of 40V.

The maximum voltage possible on the negative regulator input is -1* ((25V * 1.414) - (1V +1V)) = -33.35V DC. That in itself is withing the 40V ipV/opV rating of an LM337. But as the LM337 output voltage would be -12V, this means the voltage across the LM337 would be -1 *(33.35V -12V) = -21.35V. So well within an LM337 ipV/opV limit.

For the relatively low power level in question here (e.g. 0.5 A and up to about 30-35 V raw voltage) I don't think one would need switching of the transformer tap. It could be a good idea with higher power (e.g. more than 1-2 A) though.
At some 15 W of worst case power loss, cooling it not that difficult either and one could getaway without a fan.

I am suggesting going for 1A. Besides which, if the facility is there, why not be aware of it as a possibility.

I find your comments completely negative and general. You could make the same sort of critisisems about practically any PSU circuit. And none of your claims are backed by any calculations or figures- just sweeping statements.

You are also picking holes in a circuit which has been categorically stated as work in progress and for demonstrating the principle only.

But as I said before, as you have all this knowledge about PSU design, please lets see your design for his application. Or, if not at least recommend a PSU circuit that meets your criteria and the OPs.

Finally, on a general point, not only are you wasting my time, which could be spent on something more constructive than warding of your unwarranted criticisms of the circuit but, worse sill you will be discouraging the OP who may not have the technical knowledge to establish the degree and relevance of your list of problems as you put it.

After all that, I do appreciate you posting all the above information. It does give me a feel for how you think. And I am sure that your comments are well intended.

[spec]

I can't keep up with the pace this is going at!

Neither can I 

I would not worry about controlling a PSU with a negative supply- it is dead easy and not an issue. ...

Ok. Good to know. Between the lm723 and the custom approach, if the custom one provides more flexibility and better performance, I' ll go for it. I can give the lm723 a try if the other doesn't work out.

Very wise. As mentioned before, once you have the chassis and raw supplies built, tested and the performance established you can pretty much go anyway that you want

The PSU in answer #36 may be suitable for your needs: https://www.eevblog.com/forum/beginners/help-me-design-a-psu/msg1971227/#msg1971227

It is certainly the cheapest and simplest approach.

I could very well start by doing something like that, as I dont have any experience on power supplies, but it doesn't have adjustable current limit and I don't know how would it perform. As I am planing on spending quite some time making a decent case, adding micro-controlled voltage and current displays, I would like to do something more sophisticated. However, I can always upgrade it later.

As you say you have no experience of building PSUs, this simple approach may be a good idea. By the way, you will need a practical circuit with decoupling capacitors etc. You know that even simple power supplies can be problematic. And even three terminal regulators can play up. You would not believe the the things that can go wrong.

There is a topic which I have been meaning to mention.

With the transformers arrangement you have with 10V and 15V windings it may be an advantage to have a range switch which selects just one of the windings for PSU low output voltages and both windings for high output voltages. This is common practice on many linear supplies, either manually or automatically. On my linear lab PSUs the switching is automatic by a relay.

The reason for this switching is to reduce the power dissipation in the output power transistors.

I mentioned that in my first post, maybe a relay driven by a comparator or even the microcontroller would do. At a maximum power dissipation of around 15W/channel I don't know if it's worth it, but it wouldn't be hard to implement.

Other thing I was thinking of is that I am going to need a +12V rail to power the fans, a possible relay and the Arduino. One way I found to do this is as in the schematic attached (https://electronics.stackexchange.com/questions/153399/using-all-the-taps-of-the-transformer-at-the-same-time), using the two transformers instead of a center tapped one, and a 7812 regulator. This way the fan and anything else would draw current from the 10 V 2 A transformer. A 10V rail would do, getting regulated 12V from 10 V ac doesn't seem easy.

Ahh missed that!

Another nice touch, that is easy to do, would be to have the fans temperature controlled, or have you already mentioned that too.

Getting any DC voltage converted to any other DC voltage, within reason, is dirt cheap and easy these days.
But you could probably just use a bridge rectifier and reservoir cap off the 10V 2A winding. If necessary, you could put a power resistor in series with the fan to get the right current.

I am hoping to see the test results of your raw supply and also some pictures

[ArthurDent]

I think we've seen now that there are as many ways to build a power supply as there are posters here. The circuit I found and posted in #27 is an old design but should work fine and replacing the 4.7V zener in the negative supply with a 7905 as Wolfgang suggested should improve stability slightly. The circuit spec designed in #37 with some fine tuning should be better yet (NOTE-a slight oversight, D8 was drawn backwards). 

As much as I love to try to continually improve any circuit, if the power supply you make is to allow you to test other circuits you are more interested in, I'd say decide on one circuit you think is adequate for your needs and after you've built it, get on to constructing the other stuff you want to play with that will be powered by this supply.

[fsr]

Several years ago, i built the "5A Constant Voltage/Constant Current Regulator" circuit in the LM317 datasheet, with some modifications. Of course the 5A and 35v input are bullshit, i went with far lower input voltage and current limits. But it seems to work fairly ok (didn't test it seriously, however). There are probably better, simpler circuits posted here, anyways.

But what i was going to, is that i searched for some power transistors with low thermal resistance and found the MJ15003/MJ15004 complementary pair. They have crazy specs. Thermal resistance 0.70 degrees C Junction-Case. Nice stuff.

[spec]

(NOTE-a slight oversight, D8 was drawn backwards).  [PSU # 37]

Well spotted AD: will correct

[spec]

But what i was going to, is that i searched for some power transistors with low thermal resistance and found the MJ15003/MJ15004 complementary pair. They have crazy specs. Thermal resistance 0.70 degrees C Junction-Case. Nice stuff.

One of my all-time favorites too.

ON (Motorola) Do some pretty good power devices in general. They second source a few of the Toshiba beauties, like the MJL3281A/MJL1302A, which were no longer available at one time.

[JuanGg]

I think we've seen now that there are as many ways to build a power supply as there are posters here. ...

I think I am going with the circuit in #37. I am building this as a tool, but as a learning project as well. When I have something that works for me, I'll use it until I have the need to upgrade it.

Just noticed that the fans on the heatsinks blow air into them, instead of out. If they are inside the case, this would heat everything up (not good). I am thinking on taking the fans out  and flip them so they blow air from vents on the sides/front, across the supply and out of the case, cooling everything. It's a pity because they are nicely fitted to the heatsinks... Or I could have the heatsinks sit at the back of the case, using thermal pads (switching taps to reduce power dissipation). Any ideas on thermal design etc?
Juan

[ArthurDent]

In the 'Beginners' section there is a thread called "Help me design a PSU" and post #17 has another power supply design that has all the basic building blocks we are discussing here. It appears that the idea for the negative supply using 2 caps and 2 diodes was gleaned from the same old schematic I had found and posted here but the rest is different and might be of interest in the power supply you are thinking of building. There aren't a lot of parts values shown but the circuit uses a different approach than what has been discussing. It's always good to see as many variations as you can before making a decision.

https://www.eevblog.com/forum/beginners/help-me-design-a-psu/?action=dlattach;attach=571409;image

[spec]

It appears that the idea for the negative supply using 2 caps and 2 diodes was gleaned from the same old schematic
https://www.eevblog.com/forum/beginners/help-me-design-a-psu/?action=dlattach;attach=571409;image

That technique goes back much further than that... 1920s I would guess 

[ArthurDent]

JuanGg - "Just noticed that the fans on the heatsinks blow air into them, instead of out."

The way a processor heat sink fan is used is it removes the heat from the immediate area of the processor heatsink inside the PC case then they have a large fan or two on the back of the case to remove the heat from the case. If you can design your power supply so the top of the fan/heatsink you're using is against a hole in the case you're using or you have something like a piece of duct work pipe to go from the fan to a hole in the case, that would work to force in/remove the heat from the case. I would look for an aluminum can the correct diameter and cut that to the correct length and attach that if you need small duct work round pipe.

As far as sucking air in or blowing air out I have wrestled with this question for years and concluded it probably doesn't make much difference. One piece of equipment I have, an Array 3711A a max 360V/30A/300W electronic load, solved the problem by using 4 fans, 2 sucking in and 2 blowing out.   

[spec]

I think we've seen now that there are as many ways to build a power supply as there are posters here. ...

I think I am going with the circuit in #37. I am building this as a tool, but as a learning project as well. When I have something that works for me, I'll use it until I have the need to upgrade it.

Just noticed that the fans on the heatsinks blow air into them, instead of out. If they are inside the case, this would heat everything up (not good). I am thinking on taking the fans out  and flip them so they blow air from vents on the sides/front, across the supply and out of the case, cooling everything. It's a pity because they are nicely fitted to the heatsinks... Or I could have the heatsinks sit at the back of the case, using thermal pads (switching taps to reduce power dissipation). Any ideas on thermal design etc?
Juan

The heatsink is fundamental to the success of all power supplies, even very simple low-power types.

I did a rough thermal budget for your PSU, assuming 0V to 20V and 0A to 1A and based on that you will need a heatsink of 2 degC/W or lower if you are using a single output power transistor.

You can do a very simple test to establish the actual thermal resistance of a heatsink. Clamp one of those gold-colored  high-power, wire-wound resistors directly to your heat sink where the transistor will fit. Then dissipate, say 10W in the resistor.

After no less than 15 minutes, measure the room temperature and measure the temperature of the heatsink as close to the resistor clamp area as possible. With the three parameters:  power dissipation, heatsink temperature, room temperature (ambient), we can calculate the thermal resistance of your heatsink.

[spec]

 Online JuanGg,

Just a quick question: on PSU #37 do you want to control the voltage and current by MCU, or just voltage?

I take it that the control signal will be a positive going analog signal originating at 0V: 0V to 5V or 0V to 2V5 say.

[spec]

I think we've seen now that there are as many ways to build a power supply as there are posters here. ...

I think I am going with the circuit in #37. I am building this as a tool, but as a learning project as well. When I have something that works for me, I'll use it until I have the need to upgrade it.

Just noticed that the fans on the heatsinks blow air into them, instead of out. If they are inside the case, this would heat everything up (not good). I am thinking on taking the fans out  and flip them so they blow air from vents on the sides/front, across the supply and out of the case, cooling everything. It's a pity because they are nicely fitted to the heatsinks... Or I could have the heatsinks sit at the back of the case, using thermal pads (switching taps to reduce power dissipation). Any ideas on thermal design etc?
Juan

The heatsink is fundamental to the success of all power supplies, even very simple low-power types.

I did a fag packet thermal budget for your PSU, assuming 0V to 20V and 0A to 1A and based on that you will need a heatsink of 2 degC/W or lower if you are using a single output power transistor.

You can do a very simple test to establish the actual thermal resistance of a heatsink. Clamp one of those gold-colored  high-power, wire-wound resistors directly to your heat sink where the transistor will fit. Then dissipate, say 10W in the resistor.

After no less than 15 minutes, measure the room temperature and measure the temperature of the heatsink as close to the resistor clamp area as possible. With the three parameters:  power dissipation, heatsink temperature, room temperature (ambient), we can calculate the thermal resistance of your heatsink.

As you say, suck or blow, makes little difference, provided the hot air is expelled from the equipment case, the cooling air impinges on the heatsink, and there are no stagnant areas around the heatsink

[JuanGg]

The way a processor heat sink fan is used is it removes the heat from the immediate area of the processor heatsink inside the PC case then they have a large fan or two on the back of the case to remove the heat from the case. If you can design your power supply so the top of the fan/heatsink you're using...

As you say, suck or blow, makes little difference, provided the hot air is expelled from the equipment case, the cooling air impinges on the heatsink, and there are no stagnant areas around the heatsink

Will post a screenshot of the CAD when I have it in a decent state. I am thinking both heatsinks at the back, two holes for the fans directly on the back panel and some vents.

The heatsink is fundamental to the success of all power supplies, even very simple low-power types.

I did a rough thermal budget for your PSU, assuming 0V to 20V and 0A to 1A and based on that you will need a heatsink of 2 degC/W or lower if you are using a single output power transistor.

You can do a very simple test to establish the actual thermal resistance of a heatsink. Clamp one of those gold-colored  high-power, wire-wound resistors directly to your heat sink where the transistor will fit. Then dissipate, say 10W in the resistor.

After no less than 15 minutes, measure the room temperature and measure the temperature of the heatsink as close to the resistor clamp area as possible. With the three parameters:  power dissipation, heatsink temperature, room temperature (ambient), we can calculate the thermal resistance of your heatsink.

Thanks again for the effort!. Not sure about going to 1 A with the 1 A transformer (although it was on a hermetically sealed plastic case, so current rating would be a bit more I suspect...). I don't have any of those resistors nor a way to acurately measure temperature (I have some k thermocouples around, but my DMM does not support it  ). What I do have is a 3d printer with a heated bed that i can bring up to 95 ºC. That may work.

However, I found an Intel thermal desing guide recomending this heatsink for aplications requiring about 0.6 ºC /W case to ambient. (here it is: http://download.intel.com/design/intarch/designgd/27370403.pdf , there is a graph in page 28 and a table with the heatsink on the following page, assuming ambient temperature of 42 ºC). (With the fan running I would assume). It should be more than adecuate.

Online JuanGg,

Just a quick question: on PSU #37 do you want to control the voltage and current by MCU, or just voltage?

I take it that the control signal will be a positive going analog signal originating at 0V: 0V to 5V or 0V to 2V5 say.

I would like to control both if posible, so I can set both say using an encoder/keypad. I intended to use a filtered 12 bit PWM, from 0 to 5 V. I did this on my electronic load and it worked quite well (more details here: https://www.eevblog.com/forum/projects/arduino-based-electronic-load). If I need to, maybe an i2c DAC with a voltage reference.

Juan

[spec]

All sounds good. You have a handle on the heatsink.

You may be correct about the transformer, but on paper it should do 20V at 1A and that is what I suggest we should aim for. If the transformer won't cut it, that is just bad luck, but nothing lost.

Thanks for info about the V and I control inputs.

Just to restate then, the signal will be analog and ground referenced. 0V input will be zero volts and zero amps output.

I don't care too much about the maximum input voltage (that can be changed at any time in the future) but I will work with 5V in both cases.

We have not discussed input resistance. Would around 100k Ohms be suitable?

[JuanGg]

All sounds good. You have a handle on the heatsink.

You may be correct about the transformer, ...

Perfect.

We have not discussed input resistance. Would around 100k Ohms be suitable?

No idea on input resistance on power supplies. Looked for it on the internet but no luck. How can it be calculated?
Juan

[spec]

We have not discussed input resistance. Would around 100k Ohms be suitable?

No idea on input resistance on power supplies. Looked for it on the internet but no luck. How can it be calculated?
Juan

Think of the input resistance of a scope (1M Ohm), Audio power amplifier (10k Ohm), Muiltimeter  set to 10V DC range (1G Ohm). The only criteria is, will your analog signal drive the input resistance.

Don't worry about it, 100K will be fine. And if not it can be changed anyway. 

[JuanGg]

Think of the input resistance of a scope (1M Ohm), Audio power amplifier (10k Ohm), Muiltimeter  set to 10V DC range (1G Ohm). The only criteria is, will your analog signal drive the input resistance.

Don't worry about it, 100K will be fine. And if not it can be changed anyway.

Ok. Sorry, just realized now. Input resistance of the control signals . Alright. As they will probably be driven from an op amp buffer, 100k should be more than enough.

[spec]

Think of the input resistance of a scope (1M Ohm), Audio power amplifier (10k Ohm), Muiltimeter  set to 10V DC range (1G Ohm). The only criteria is, will your analog signal drive the input resistance.

Don't worry about it, 100K will be fine. And if not it can be changed anyway.

Ok. Sorry, just realized now. Input resistance of the control signals . Alright. As they will probably be driven from an op amp buffer, 100k should be more than enough.

All good

[JuanGg]

Haven't been able to do much since last time I posted (exam week!, and more to come), here is what I have on the case so far. Hopefully I will finish designing the back panel and print it this weekend. I will leave the front panel to last, when meters and binding posts arrive. Attached are a couple of screenshots, one showing the airflow. Excuse the crudity of the pictures. There is yet a lot of work to be done.

[spec]

Looking good. Your case has a sort of military look to it.

Just one thing: I would suggest having the fans sucking or blowing out. You don't want to have the hot air from the heatsinks blowing into the case.

Apart from the power transistor/s, the components including transformers, rectifiers and reservoir capacitors will be OK with convection cooling and do not need fan cooling, providing you have sufficient vents on the case.  The bridge rectifier may need a heatsink, but the chassis will do for that.

PS: Don't let anything get in the way of your studies. Good exam results are essential- you can do PSUs anytime.

[JuanGg]

Looking good. Your case has a sort of military look to it.

Just one thing: I would suggest having the fans sucking or blowing out. You don't want to have the hot air from the heatsinks blowing into the case.

Apart from the power transistor/s, the components including transformers, rectifiers and reservoir capacitors will be OK with convection cooling and do not need fan cooling, providing you have sufficient vents on the case.  The bridge rectifier may need a heatsink, but the chassis will do for that.

It must be the green I guess. It just happens to be the only filament I have around.
Having the fans blow the hot air out is what I wanted on the first place. I will try to cut the fans out of their support and mount them reversed.
Can't the bridge rectifiers be mounted on the same heatsink as the power transistor? I won't have much metal chassis, but just a cover. (Not shown).

PS: Don't let anything get in the way of your studies. Good exam results are essential- you can do PSUs anytime.

No worries. I have my priorities right. I don't work on this on weekdays.

[spec]

Can't the bridge rectifiers be mounted on the same heatsink as the power transistor? I won't have much metal chassis, but just a cover. (Not shown).

Not advisable- look upon the power transistor heat sinks as heaters as far as other components are concerned.

Normally an aluminum bracket or sheet is sufficient. The bridge rectifiers will only be dissipating around 500mW each anyway. Rectifier cooling is not an issue.   

PS: Don't let anything get in the way of your studies. Good exam results are essential- you can do PSUs anytime.

No worries. I have my priorities right. I don't work on this on weekdays.

That's the way

[perieanuo]

The ventilation for even 1 amp is overkill,I did more with 1 40x40 quiet one.with those and 2-3 power transistors you can go to 10 amps.I end up putting an termistor on the heatsink with pwm control at 3 different temps(I have atmega328 board monitoring U-I with ina sensor showed on oled white,looks fine).And the atmega can easily monitor between U and I the temp and put some pwn on the ventilation, cost bucks and works fine, coding is simple.
Regards,pierre


Envoyé de mon iPad en utilisant Tapatalk

[fsr]

When i buit mine, i went fanless. It had 34W max dissipation. I put the transistor with mica and carefully applied thermal grease on the heatsink. The MJ15003 transistor did most of the magic. Rth J-C of 0.7 °C/W and 200 °C max junction temperature. I don't really know how much i used it at full load + low voltage, but the magic smoke remains contained up to this day
Look at the attached datasheet for details of how to calculate this stuff. Page 10, "Heatsink Requirements". Just substitute the PD and other characteristics with the ones for your transistor.
TO-3's are hard to mount, but with that thermal figures, it's worth the effort for me! Check other power packages, anyways, but it's probably hard to beat a metal can mounted with two screws.

Remember that CPUs sometimes reach power dissipations up to 135 watts (but obviously not all CPUs). So that's probably too much cooling in this case.

Also remember that hot air moves up, so you need to put ventilation holes or fans near the top of the case blowing air out. That's why good vertical computer cases have the PSU on the top, and input ventilation holes on front, near the bottom.

In my case, i just put the heatsinks on the outside, so i didn't have to deal with that. Just natural convection, and it goes up unimpeded.

[spec]

The ventilation for even 1 amp is overkill,I did more with 1 40x40 quiet one.with those and 2-3 power transistors you can go to 10 amps.I end up putting an termistor on the heatsink with pwm control at 3 different temps(I have atmega328 board monitoring U-I with ina sensor showed on oled white,looks fine).And the atmega can easily monitor between U and I the temp and put some pwn on the ventilation, cost bucks and works fine, coding is simple.
Regards,pierre


Envoyé de mon iPad en utilisant Tapatalk

I take it that by ventilation, you mean fan cooling. I don't like the term overkill- it is a sweeping statement that can be used to criticize pretty much anything. After all, your house, car, PC are all overkill.

Although I tend to agree with you that fan cooling is probably is not necessary, you cannot say for sure without doing a thermal budget and to do that you would need to know the thermal resistance of the heatsink being used. Besides which, the OP already has the fans and heatsinks which were pulled from another unit.

The other point is that, not only will this PSU design and construction result in a usable PSU, but it will be a very good experience for the OP, who is just starting in his electronics career. It is probable that this PSU will be further developed, maybe to have more output voltage and current, at a later date, so a low thermal resistance heat sink would then be an advantage.

You mention about your own power supply- feel like posting a schematic? There is always great intrest in PSUs on EEV.  

[spec]

When i buit mine, i went fanless. It had 34W max dissipation. I put the transistor with mica and carefully applied thermal grease on the heatsink. The MJ15003 transistor did most of the magic. Rth J-C of 0.7 °C/W and 200 °C max junction temperature. I don't really know how much i used it at full load + low voltage, but the magic smoke remains contained up to this day

That is real class. The MJ15003 is a beauty. Unfortunately the more mundane, but cheaper, 2N3055 has a thermal resistance junction to case of 1.52 DegC/W, more than twice that of the MJ1503. But the 2N3055 safe operating area (SOA) is better.

https://www.onsemi.com/pub/Collateral/MJ15003-D.PDF
https://www.onsemi.com/pub/Collateral/2N3055-D.PDF

Would be great to see a schematic of your PSU.

Update: just checked prices from DigiKey UK, the MJ15003 costs £6.36 and the 2N3055 costs £4.60, so not a vast price difference.

[spec]

There has been quite a bit of discussion about heat sinking, both on this thread and elsewhere on EEV.

Just to give give an overview, here is a description of a thermal budget using an actual heatsink and power transistor.

Assume, due to space and cost considerations, that the following heatsink must be used in an equipment: https://www.apexanalog.com/resources/mechdrawings/hs14.pdf. This heatsink is quite expensive at £51.11UK and fairly large at 5 by 3 by 1.3 inches. Its thermal resistance in free air is 2 degC/W.

You have one 2N3055 transistor and you want to know how much power you can theoretically dissipate.

So the list of parameters is:

• 2N3055 maximum junction temperature (Tjmax): 200 degC
• 2N3055 thermal resistance junction to case (ThRjc): 1.52 degC/W
• Insulating washer between 2N3055 case and heat sink thermal resistance (ThRwasher): 1 degC/W
• Heat sink thermal resistance to air (ThRhs): 2 degC/W
• Air temperature in the vicinity of heatsink (Tamb): 70 degC

Armed with this information, a thermal budget can be calculated from the following formula:

Tjmax = P (ThRjc + ThRwasher + ThRhs) + Tamb

P is required, so transposing gives,  P = (Tjmax-Tamb) / (ThRjc + ThRwasher + ThRhs)

Inserting actual values gets, P = (200 - 70) / (1.52 + 1 + 2) = 28.76W

That is the maximum theoretical dissipation and does not include, a safety margin, SOA considerations, and temperature effects on SOA, for example.

At the end of day, you would probably end up with a practical maximum power dissipation of 20W to 25W.

[fsr]

Would be great to see a schematic of your PSU.

I have some screen captures back from the days i built it. It's just a circuit it used to be on the LM317 datasheet i attached before, with some modifications, and also a negative version. I didn't really ever tested it, but seems to do the job. I had to build it in 3 boards, because of limitations on Eagle PCB.

If someone is interested on the PCBs, i can open Eagle and PDF them, or upload the files.

I posted it on another forum back in tha day, but it didn't catch much attention, so i never released much about it.

Again, the power transistors are MJ15003 and MJ15004.

[JuanGg]

I have printed and glued the back panel together. I won't bore you with the details, but sufices to say that big abs pieces don't get 3d printed without putting up a fight. I just printed it in pieces and acetone welded it together. Attached are some photos.
I think I am keeping the fan and heatsinks as they are, just blocking the inner sides, so air enters the back, goes through the heatsink and exits through vent holes on the sides, without going into the case.

Juan.

[spec]

so air enters the back, goes through the heatsink and exits through vent holes on the sides, without going into the case.

 

[JuanGg]

Thinking about the user interface, I saw the power supply attached, which is was very similar to what I had thought (withoght the middle chanel). I just think it is very simple and functional. The downside is that it would require isolated comunication between chanels (serial via optocouplers between arduinos?). Each chanel would have its display, and only one of them would read the encoder and buttons, sending information to the other via isolated serial. This way I could do tracking or whatever. Other way I thought of doing is using a central arduino that reads the encoder and keys, and sends the information to the other two, also providing an isolated USB, so the PSU can be interfaced with a computer. (See diagram attached). Is it feasible? Or maybe too much for a first power supply? I don't think it is very hard to acomplish, and given the arduinos are very cheap, I don't mind including three of them in the unit.

Juan

[wasyoungonce]

.............
If someone is interested on the PCBs, i can open Eagle and PDF them, or upload the files.I posted it on another forum back in tha day, but it didn't catch much attention, so i never released much about it...Again, the power transistors are MJ15003 and MJ15004.

Actually yes please...I need a basic fixed rail supply and some limited variable outputs.  I already have qty 2 PS-3005Ds 30V/5A but I really need something simple to drive small motors (want to monitor their start voltage currents) and logic ccts.  The others will do this but far easier to have a small supply with fixed and some variable outputs.  Eagle .sch .brd  etc would be great...also whats the I/P transformer ~ 15VAC?   In VA?  I do have some decent 30VAC ~@120VA but they are probably a little too high in output.

I'm supposed to be trying to use KICad but its so difficult to leave something you already know so well! 
Many thanks

edit.....Oh I love your PSU you have done.....:

https://easyeda.com/asokolsky/Analog-Lab-Power-Supply

I have to ask...why a 56V out as there is a 0-30V and other fixed O/P?.   Maybe a bit of self gratification and promotion is due your way.  I see you designed it in with your own Amp/Voltmeter and fan controller. 

https://easyeda.com/asokolsky/AmpVolt-Meter-with-Fan-Control

Well done.   Looking intently at them for an unashamed and undeserved diy for myself. 

[Mr. Scram]

Thinking about the user interface, I saw the power supply attached, which is was very similar to what I had thought (withoght the middle chanel). I just think it is very simple and functional. The downside is that it would require isolated comunication between chanels (serial via optocouplers between arduinos?). Each chanel would have its display, and only one of them would read the encoder and buttons, sending information to the other via isolated serial. This way I could do tracking or whatever. Other way I thought of doing is using a central arduino that reads the encoder and keys, and sends the information to the other two, also providing an isolated USB, so the PSU can be interfaced with a computer. (See diagram attached). Is it feasible? Or maybe too much for a first power supply? I don't think it is very hard to acomplish, and given the arduinos are very cheap, I don't mind including three of them in the unit.

Juan

Ask yourself whether you want to multiplex the controls.

[JuanGg]

Ask yourself whether you want to multiplex the controls.

I don't really know if it's worth it. I can try different ways and see how it works out. Anyway, the important thing is to get the analog PSU working. I just have a tendency to overcomplicate stuff. As I said, I'll leave the front panel to last.

[perieanuo]

The ventilation for even 1 amp is overkill,I did more with 1 40x40 quiet one.with those and 2-3 power transistors you can go to 10 amps.I end up putting an termistor on the heatsink with pwm control at 3 different temps(I have atmega328 board monitoring U-I with ina sensor showed on oled white,looks fine).And the atmega can easily monitor between U and I the temp and put some pwn on the ventilation, cost bucks and works fine, coding is simple.

I take it that by ventilation, you mean fan cooling. I don't like the term overkill- it is a sweeping statement that can be used to criticize pretty much anything. After all, your house, car, PC are all overkill.

Although I tend to agree with you that fan cooling is probably is not necessary, you cannot say for sure without doing a thermal budget and to do that you would need to know the thermal resistance of the heatsink being used. Besides which, the OP already has the fans and heatsinks which were pulled from another unit.

The other point is that, not only will this PSU design and construction result in a usable PSU, but it will be a very good experience for the OP, who is just starting in his electronics career. It is probable that this PSU will be further developed, maybe to have more output voltage and current, at a later date, so a low thermal resistance heat sink would then be an advantage.

You mention about your own power supply- feel like posting a schematic? There is always great intrest in PSUs on EEV.

hi,
for me overkill means too much.
I attached my schematic, it's not the final version, I upgraded current sense.But the idea is, I put also temperature measurement and tested the thing at 3.000 Amps with variable load.
The temp stabilised to 44-48 deg C if I remember right.I sense with 10K NTC thermister and I was calibrated with Volcraft PL-125-T2USB.
I pwm to fan starting at 30 deg C (33%, then 66% at 40deg then 100% at 50degC).
If my 40*40 fan succeed to cool the 2xTIPL790A transistors delivering 3amps at 5V from 24 Vac transformer that means those 2x80*80 fans will be toooo much for what OP wants( he has 2 amps rated transformers).
Imho, he can think of more current with those fan/cooler combination and he MUST reduce fans speed, a variable lab supply MUST be quiet, it's not just fancy stuff, when you're focused doing electronics and coding some dsp last thing you need is a stupid vent crying.
Same cooling technique I did for other people in scopes for example (one of my bosses had enough of his oscillo and power supply doing noise when he was working on research stuff).
I take seriously the cooling, I'm not perfect at this but even implement a fan cooling controled by a thermistor and a fet who starts at a defined temperature for me is a must.
Attached I'll put schematics of one of my bench ps with power stage.just for example.I had a case and stuffed inside toroidal, power stage, simple opamp regulation (V/I control) and measurement with custom atmega 328 board, current module down to 0.1mA resolution is not in schematics, but it's in the picture ).
regards,Pierre

[perieanuo]

Thinking about the user interface, I saw the power supply attached, which is was very similar to what I had thought (withoght the middle chanel). I just think it is very simple and functional. The downside is that it would require isolated comunication between chanels (serial via optocouplers between arduinos?). Each chanel would have its display, and only one of them would read the encoder and buttons, sending information to the other via isolated serial. This way I could do tracking or whatever. Other way I thought of doing is using a central arduino that reads the encoder and keys, and sends the information to the other two, also providing an isolated USB, so the PSU can be interfaced with a computer. (See diagram attached). Is it feasible? Or maybe too much for a first power supply? I don't think it is very hard to acomplish, and given the arduinos are very cheap, I don't mind including three of them in the unit.

Juan

yes it's feasible, but you can do it with only 2 arduinos not 3, or better even 1 if you put galvanic separations, but costs a little more, you can calculate.what you economise on arduino you put in isolation, but switch to clean and pro solution.I did arduino-like board and ina219 voltage/current sense (mA resolution) but only 1 channel, I had only a small box for the PS (see my earlier post).
When I'll find a bigger one, I'll do like you said, central supervision and screen on 2 oleds or a big LCD.take a look on my pdf's if you want, posted earlier.control as I sugested the vents speed, most of the time you'll need zero active cooling so zero noise...you can do it with microcontroller/dsp or a simple FET transistor+thermister, costs nothing...
regards,pierre

[JuanGg]

I pwm to fan starting at 30 deg C (33%, then 66% at 40deg then 100% at 50degC).
If my 40*40 fan succeed to cool the 2xTIPL790A transistors delivering 3amps at 5V from 24 Vac transformer that means those 2x80*80 fans will be toooo much for what OP wants( he has 2 amps rated transformers).
Imho, he can think of more current with those fan/cooler combination and he MUST reduce fans speed, a variable lab supply MUST be quiet, it's not just fancy stuff, when you're focused doing electronics and coding some dsp last thing you need is a stupid vent crying. ...

Better safe than sorry  . I will do temperature-controlled fans anyway. I did implement a pwm temperature controlled fan on my electronic load and it worked quite well. Plus, this fans are quiet enough (compared to the rigol 1054z's anyway...). In fact, one of the transformers I plan to use is 1A rated.

yes it's feasible, but you can do it with only 2 arduinos not 3, or better even 1 if you put galvanic separations, but costs a little more, you can calculate.what you economise on arduino you put in isolation, but switch to clean and pro solution.I did arduino-like board and ina219 voltage/current sense (mA resolution) but only 1 channel, I had only a small box for the PS (see my earlier post).
When I'll find a bigger one, I'll do like you said, central supervision and screen on 2 oleds or a big LCD.take a look on my pdf's if you want, posted earlier.control as I sugested the vents speed, most of the time you'll need zero active cooling so zero noise...you can do it with microcontroller/dsp or a simple FET transistor+thermister, costs nothing...
regards,pierre

I just thought it would be easier to have isolated digital comunications, so I don't have to deal with nonlinearities on analog optocouplers and such (any other way of getting galvanic isolation?). Plus, I would have to control two four-digit sevent segment displays for each chanel. Anyway, I will deal with that later.

I am waiting for the AC mains conector (it's taking ages to arrive!) so I can wire the raw psu and test it. Regarding that, do I go with tap-switching?

Juan

[fsr]

.............
If someone is interested on the PCBs, i can open Eagle and PDF them, or upload the files.I posted it on another forum back in tha day, but it didn't catch much attention, so i never released much about it...Again, the power transistors are MJ15003 and MJ15004.

Actually yes please...I need a basic fixed rail supply and some limited variable outputs.

I found the files, and posted them here: https://www.eevblog.com/forum/oshw/basic-lm317-based-cvcc-15v15v-1-5a-split-lab-psu/msg2003093/#msg2003093

[wasyoungonce]

.............
If someone is interested on the PCBs, i can open Eagle and PDF them, or upload the files.I posted it on another forum back in tha day, but it didn't catch much attention, so i never released much about it...Again, the power transistors are MJ15003 and MJ15004.

Actually yes please...I need a basic fixed rail supply and some limited variable outputs.

I found the files, and posted them here: https://www.eevblog.com/forum/oshw/basic-lm317-based-cvcc-15v15v-1-5a-split-lab-psu/msg2003093/#msg2003093

Many thanks watching that thread now

[coppercone2]

Exams more important then anything in engineering?? Set sail for the sales department!

Whats important is doing the job right at any cost.

[JuanGg]

Exams more important then anything in engineering?? Set sail for the sales department!

Whats important is doing the job right at any cost.

I try to do both things right. 

Here is some progress with the case. I have attached the transformers, changed locations a little bit. My idea is to have the raw PSU section between transformers and the rest on the sides. I'll try to wire the bridge rectifier and filter cap for one side this weekend. It will temporarily take the full 25 Vac, I'll eventually add a relay and make a PCB.

[fsr]

.............
If someone is interested on the PCBs, i can open Eagle and PDF them, or upload the files.I posted it on another forum back in tha day, but it didn't catch much attention, so i never released much about it...Again, the power transistors are MJ15003 and MJ15004.

Actually yes please...I need a basic fixed rail supply and some limited variable outputs.

I found the files, and posted them here: https://www.eevblog.com/forum/oshw/basic-lm317-based-cvcc-15v15v-1-5a-split-lab-psu/msg2003093/#msg2003093

Many thanks watching that thread now

[JuanGg]

Here are some photos of the wiring I have done so far. I just took the ac mains conector from my other power supply as it's running off a wall wart, I'll replace it when the new one arrives. I also changed the 10 A fuse that came with it with a 1 A one, which is the smallest I have around. I'll put a smaller one later.
I have wired all the transformers to the mains input, all conections heatshrunk and all exposed areas covered in electrical tape. (I know it's not the best, but it lessens the chance of me getting shocked). Transformer cores are grounded and I used reasonably thick wire. Wires coming out of the transformers are color-coded so I know hot to connect them so that the phases are in sync (I checked that with the scope). I will continue next weekend.

Juan

[JuanGg]

I have "bodged" the bridge rectifier and the filter cap just to try it out. Attached is a picture and some scope screenshots. Last one is the capacitor being charged and discharged by the 500 bleeder resistor after turn off. My electronic load can't handle more than 30 V, so I was unable to test it with it. I also don't have power resistors around, so I paralleled a dozen 1/4 W ones, but drawing 250 mA they still got too hot. Voltage dropped from 38 V to 36 V. I'll look for means of testing it properly.

Juan

[perieanuo]

hi,
for me the 2 transformers are a little on the edge for superior current limit.
Translation, I doubt they can take 1A for long time.
But for first PSU it's fine, if you want to power arduinos it will do fine.
Anyway changing them is 15' job, continue with pcb and all.
regards, Pierre

[JuanGg]

hi,
for me the 2 transformers are a little on the edge for superior current limit.
Translation, I doubt they can take 1A for long time.
But for first PSU it's fine, if you want to power arduinos it will do fine.
Anyway changing them is 15' job, continue with pcb and all.
regards, Pierre

On the wall - warts they came from said 15 V 1 A, but who knows. I am aiming at 500-600 mA anyways. I can always upgrade the transformers later. I'll make a PCB with the rectifier, cap, tap-switching relay and maybe an lm 317 for a 10 V rail (to power the Arduino, fans relays, etc.)Any thoughts on this?

Juan

[perieanuo]

hi,
for me the 2 transformers are a little on the edge for superior current limit.
Translation, I doubt they can take 1A for long time.
But for first PSU it's fine, if you want to power arduinos it will do fine.
Anyway changing them is 15' job, continue with pcb and all.
regards, Pierre

On the wall - warts they came from said 15 V 1 A, but who knows. I am aiming at 500-600 mA anyways. I can always upgrade the transformers later. I'll make a PCB with the rectifier, cap, tap-switching relay and maybe an lm 317 for a 10 V rail (to power the Arduino, fans relays, etc.)Any thoughts on this?

Juan

Me,I always put a minimum 10 amps rectifier screwed to the bottom case, then main capacitors on separate pcb or glued to the case with 2-face adhesive like 3M stuff, the control board itself separate pcb.but you can do your thing
Of course fuses for every transformer output and 220V input.those days on outputs I tend to put rearmable fuses (ptc) for currents like 0-2 Amps, tired of fuse stocking...
Regards,pierre


Envoyé de mon iPad en utilisant Tapatalk

[cdev]

1 amp is a lot of power for low voltage stuff, most of the times I use my triple bench supply I use MUCH less than its rated 7v@6a and 20v @ 3 amps x2 . However, lots of the time I need two voltages at the same time with current limiting. If I were you I would make two identical supplies and make them well regulated with adjustable current limiting and some kind of notification (LED and/or beeper would be best) when the current draw reaches that limit. A nice multiturn pot for voltage and a digital 4 digit display of the voltage and current (so you can set the voltage + current limiting easily) with a floating decimal point (around $3 on ebay) is nice too.

You could probably get away with multiple common fixed voltages but make sure you have a nice pot to adjust the current limit.

[fsr]

hi,
for me the 2 transformers are a little on the edge for superior current limit.
Translation, I doubt they can take 1A for long time.
But for first PSU it's fine, if you want to power arduinos it will do fine.
Anyway changing them is 15' job, continue with pcb and all.
regards, Pierre

On the wall - warts they came from said 15 V 1 A, but who knows. I am aiming at 500-600 mA anyways. I can always upgrade the transformers later. I'll make a PCB with the rectifier, cap, tap-switching relay and maybe an lm 317 for a 10 V rail (to power the Arduino, fans relays, etc.)Any thoughts on this?

Juan

And did the wall-warts had a bridge rectifier with capacitive filter? Because the maximum current output, ripple, and other characteristics varies with different rectifier circuits. Take a look:

[JuanGg]

And did the wall-warts had a bridge rectifier with capacitive filter? Because the maximum current output, ripple, and other characteristics varies with different rectifier circuits. Take a look:

No, they were ac ones, just the tranformer. Thank you for the very informative sheet.

[JuanGg]

Me,I always put a minimum 10 amps rectifier screwed to the bottom case, then main capacitors on separate pcb or glued to the case with 2-face adhesive like 3M stuff, the control board itself separate pcb.but you can do your thing
Of course fuses for every transformer output and 220V input.those days on outputs I tend to put rearmable fuses (ptc) for currents like 0-2 Amps, tired of fuse stocking...
Regards,pierre

I was thinking of making a separate pcb with rectifier, cap, relay, even fan controller. I won't do it just for now, as I may be needing more voltage rails for the regulating pcb. The big transformer already has fuses on every winding, but I may add some more.

1 amp is a lot of power for low voltage stuff, most of the times I use my triple bench supply I use MUCH less than its rated 7v@6a and 20v @ 3 amps x2 . However, lots of the time I need two voltages at the same time with current limiting. If I were you I would make two identical supplies and make them well regulated with adjustable current limiting and some kind of notification (LED and/or beeper would be best) when the current draw reaches that limit. A nice multiturn pot for voltage and a digital 4 digit display of the voltage and current (so you can set the voltage + current limiting easily) with a floating decimal point (around $3 on ebay) is nice too.

You could probably get away with multiple common fixed voltages but make sure you have a nice pot to adjust the current limit.

That's precisely what I want to do. Dual 0-25 V @ 0.5 A. Arduino controlled, with filtered 12-bit pwm as DAC. I will be using the arduino's ADCs (10 bit) to measure voltage and current, showing the readings on 4-digit 7 segment displays or on an LCD. I can also display the set values as well.

I don't really know what to do now. I may start breadboard the circuit in #37.
      Juan

[perieanuo]

37 may not work, missing reverse output diode, me I don't like it.
my favs are sokolsky one and mine course
regards,pierre

[JuanGg]

37 may not work, missing reverse output diode, me I don't like it.
my favs are sokolsky one and mine course
regards,pierre

Well that diode can be easily added.
I can't do the sokolsky one as-is as I don't have another transformer to derive the +-5V referenced to the output.Any ideas?

Juan

[JuanGg]

Connecting only the 15 Vac transformer, I had the raw PSU running at 1 A for 20 minutes, using my electronic load. The rectifier got warm, so did the transformer. Not too hot to keep one's hand on it for a while, so I guess it will be fine. DC voltage out dropped from 17 to 15 V and I was getting about 1 Vpp ripple.

Juan

[perieanuo]

37 may not work, missing reverse output diode, me I don't like it.
my favs are sokolsky one and mine course
regards,pierre

Well that diode can be easily added.
I can't do the sokolsky one as-is as I don't have another transformer to derive the +-5V referenced to the output.Any ideas?

Juan

Dcdc converter (1.5 euros ebay) with capacitors aside or rcd to generate -5V from AC branch.see my schematic as guide,I use 24V and 12V outputs from transformer.
I doubt #37 works...


Envoyé de mon iPad en utilisant Tapatalk

[fsr]

And did the wall-warts had a bridge rectifier with capacitive filter? Because the maximum current output, ripple, and other characteristics varies with different rectifier circuits. Take a look:

No, they were ac ones, just the tranformer. Thank you for the very informative sheet.

You're welcome.
If the wall-wart was AC, then it probably listed the transformer's AC output current, so if you use a diode bridge and capacitive filter, you can get about 620 mA of current from the 1A rated transformer. Any more current than that would be going over the maximum.

[JuanGg]

Dcdc converter (1.5 euros ebay) with capacitors aside or rcd to generate -5V from AC branch.see my schematic as guide,I use 24V and 12V outputs from transformer.
I doubt #37 works...

Will look into that. As I previously said, I'll go with whatever you guys recommend.

     
If the wall-wart was AC, then it probably listed the transformer's AC output current, so if you use a diode bridge and capacitive filter, you can get about 620 mA of current from the 1A rated transformer. Any more current than that would be going over the maximum

As they were sealed up in a plastic enclosure, I suppose they would be able to handle a bit more. I am aiming for 500-600 mA output current anyway, so no problem there.

Juan

[cdev]

Try an old ATX power supply as your power source. Do you have one?

[JuanGg]

Try an old ATX power supply as your power source. Do you have one?

I have two in fact. But they are mains earth referenced, so not suitable for my purpose. One of them is fitted with some binding posts, and that's what I have been using for a while.
Plus, I already have the transformers mounted and the enclosure half made. I am not changing that unless I have to.

Juan

[JuanGg]

Some of the stuff I ordered just arrived, including 7-segment displays, 0.1 power resistors, some perfboard etc. I have been playing around with some of them, towards making the "easy" digital part. Who writes a 7-segment display library and doesn't support decimal points? .
So far I have managed to write a function that displays a decimal number. Also I've made a simple test program that lets you set the value on the display via a rotary encoder. My idea is to have two of this displays per chanel, showing voltage and current.
(I know I should focus on the analog side first, just wanted to get something working)
    Juan

[JuanGg]

Any thoughts on this schematic? (https://dangerfromdeer.com/2016/04/06/bench-power-supply-build-part-ii/). It seems simple enough and has 0-5 V inputs to set voltage and current, which can be easily changed from pots to DACs. No auxiliary transformers are needed as well. Current sensing can be changed.

    Juan

[xavier60]

It looks mostly fine to me.
I would do some experimenting with the compensation to reduce the size of C4 which would be causing slow reaction to sudden current overloads. The AC gain of the op-amps would have to be reduced with local feedback.

[JuanGg]

It looks mostly fine to me.
I would do some experimenting with the compensation to reduce the size of C4 which would be causing slow reaction to sudden current overloads. The AC gain of the op-amps would have to be reduced with local feedback.

Ok. I can breadboard the thing and try several capacitor values, loading it with my electronic load in pulse mode.
I suppose local feedback to reduce AC gain can be added by putting capacitors in the feedback loop, just between the output and the non inverting input of the op-amps.
I will order the current sense ic and the Darlington transistor, and maybe some DACs. I suppose I can use 2n3904's and 2n3906's instead of the BC337 and  BC327.
    Juan

[xavier60]

Any thoughts on this schematic? (https://dangerfromdeer.com/2016/04/06/bench-power-supply-build-part-ii/). It seems simple enough and has 0-5 V inputs to set voltage and current, which can be easily changed from pots to DACs. No auxiliary transformers are needed as well. Current sensing can be changed.

    Juan

I mocked up just the voltage regulation loop with different parts. I used a TIP35C and BC337 with 100Ω and 1KΩ B-E resistors for the Darlington. And a TLC072 op-amp.
The presence of C4 encourages the Darlington to oscillate at certain conditions. A 100Ω series Base resistor fixed that.
The loop was stable with C4 removed and a 1000pF compensating capacitor added to the op-amp, with or without an output capacitor.
The loop became unstable when I reduced the compensating capacitor to try to improve the load transient response.
This was mainly caused by high frequency phase reversal at Q2(BC548) caused by more signal being coupled through the Base to Collector capacitance than the expected  amplified inverted signal.  Bypassing R3 with 22pF made a big improvement.
I got the best load transient response of 5us by configuring Q2 as  Common Base and swapping the op-amp inputs.
I can't do any testing of the current loop as I don't have the INA196  handy.

[JuanGg]

I mocked up just the voltage regulation loop with different parts. I used a TIP35C and BC337 with 100Ω and 1KΩ B-E resistors for the Darlington. And ...

Thanks a lot. I will draw a schematic with the changes you made and try it myself. Not sure about Q2 as common base, is it base to 3, emitter to 2 and collector to 1, as drawn on the schematic?
    Juan

[xavier60]

This shows Q2 configured as a Common Base amplifier. It is very good at transferring fast signals from ground referenced circuitry to circuitry referenced to a higher voltage.
Because the Base is tied to the 8V rail which is regarded as a signal ground, the Base acts like a shield between input and output. The Emitter is the signal input in a Common Base amplifier.
https://en.wikipedia.org/wiki/Common_base

[xavier60]

Q2 as a Common Base amplifier causes a possible complication.
Because it's non-inverting, the op-amp inputs had to be swapped.
When reference voltage  is connected to the inverting input and the feedback divider to the non-inverting input, the op-amp can't ever function as a true Miller Integrator. The minimum gain can never be less than unity.
For the voltage regulation loop, a small amount of proportional gain can and does improved stability.
For the current loop it is an uncertainty.
With Q2 as a Common Emitter amplifier again and a bit of tweaking, the response and stability are near as good as the Common Base version.
 

[JuanGg]

Q2 as a Common Base amplifier causes a possible complication.
Because it's non-inverting, the op-amp inputs had to be swapped.

Thank you very much. I'll try to prototype it myself as soon as I can.

[xavier60]

Because I don't have the IC for high-side current sensing. I have added low-side sensing.
I know it's not ideal because control circuit current flows through the shunt. It's only a few milliamps.
As expected, there is a massive current spike when the output is short circuited.
I do have a solution for this that is currently working in a bench supply that I had previously completed.
I'm trying to adapt it to the present design.
I have made small changes to the CV loop which is working very well.

[not1xor1]

Because I don't have the IC for high-side current sensing. I have added low-side sensing.
I know it's not ideal because control circuit current flows through the shunt. It's only a few milliamps.
As expected, there is a massive current spike when the output is short circuited.
I do have a solution for this that is currently working in a bench supply that I had previously completed.
I'm trying to adapt it to the present design.
I have made small changes to the CV loop which is working very well.

I think you are looking for troubles with that design.
If you do not want to use an aux. supply voltage for the regulator circuit (Harrison design) the circuit below is much simpler, much easier to compensate (may even work with just a small cap across the feedback resistor) and might work up to 30-35 Vout.



So far I've just simulated lot of variations of this circuit in all possible conditions, but failed to find the time to test it (my lab is a mess because the walls are covered by mould and I had to move most of stuff elsewhere).

[xavier60]

Q2 as a Common Base amplifier causes a possible complication.
Because it's non-inverting, the op-amp inputs had to be swapped.

Thank you very much. I'll try to prototype it myself as soon as I can.

You might already know this. For the best results, the components shown below need to be wired directly together at the heat sink.
What part number Darlington have you decided to use?

[perieanuo]

After you do your first highside current sensing, you never come back to low side


Envoyé de mon iPad en utilisant Tapatalk

[JuanGg]

. Because I don't have the IC for high-side current sensing. I have added low-side sensing.
I know it's not ideal because control circuit current flows through the shunt. It's only a few milliamps.
As expected, there is a massive current spike when the output is short circuited.
I do have a solution for this that is currently working in a bench supply that I had previously completed.
I'm trying to adapt it to the present design.
I have made small changes to the CV loop which is working very well.
 

Looking good. Thanks again. I will see if I can order those current sense ICs. Not aiming for a precision PSU here and I can calibrate current error in software, so we may get away with low side current sense for starters.

Q2 as a Common Base amplifier causes a possible complication.
Because it's non-inverting, the op-amp inputs had to be swapped.

Thank you very much. I'll try to prototype it myself as soon as I can.

You might already know this. For the best results, the components shown below need to be wired directly together at the heat sink.
What part number Darlington have you decided to use?

I'll try to make it in a perfboard as compact as possible, right next to the heatsink. I don't have any darlingtongs at hand, only some 2n3055s and tip42s. I will have to order some, so whatever you guys recommend.

I think you are looking for troubles with that design.
If you do not want to use an aux. supply voltage for the regulator circuit (Harrison design) the circuit below is much simpler, much easier to compensate (may even work with just a small cap across the feedback resistor) and might work up to 30-35 Vout.

Thanks for the input. I'm going to need an 8-12V rail anyway to power the micro, displays, relays and fans.

    Juan

[xavier60]

I'll try to make it in a perfboard as compact as possible, right next to the heatsink. I don't have any darlingtongs at hand, only some 2n3055s and tip42s. I will have to order some, so whatever you guys recommend.

It's not necessarily for the whole control circuit to be made compact and close to the heat sink. Mainly the high current rails, the input and output capacitors and those other components shown in the previous diagram that should be kept close together. I have about 15cm  wires between the heat sink mounted components and the control circuit on the breadboard.
 I'm using a TIP35C and BD137 for the Darlington at the moment. Using different transistors might alter the compensation.
I'm not familiar with common Darling types. What output voltage and  current are you aiming for?
I'm worried that although that my mockup is working well, it might not work the same with different parts, mainly the transistors and the op-amp. I'm using the TLC072 which is very well suited.

What equipment do you have to test with?

I have included the current state of the circuit.
If you go ahead, don't bother with the CC loop until the CV loop is working properly.
I really like the topology of this power supply. I hope that it is successful. There doesn't seem to be much other viable alternatives with single  control rail referenced to the - output terminal.

[JuanGg]

It's not necessarily for the whole control circuit to be made compact and close to the heat sink. Mainly the high current rails, the input and output capacitors and those other components shown in the previous diagram that should be kept close together. I have about 15cm  wires between the heat sink mounted components and the control circuit on the breadboard.

All right. I was thinking of doing one board with the rectifier, filter cap, relay, and 8V reg, then another one with the control circuit, close to the heatsink. I may do a separate small one with the components shown in #126.  The microcontroller would be closer to the front panel, as it has deal with displays, encoders and such.

I'm using a TIP35C and BD137 for the Darlington at the moment. Using different transistors might alter the compensation.
I'm not familiar with common Darling types. What output voltage and  current are you aiming for?
I'm worried that although that my mockup is working well, it might not work the same with different parts, mainly the transistors and the op-amp. I'm using the TLC072 which is very well suited.

I may order those same ones, they are not expensive and if I don't use them they will go to the parts bin, same for the op amps. I'll see if I have some with similar characteristics as the TL072. I'll try to prototype the thing with what I have anyway.

I am aiming at 20-25 V at 0.6 A. The thing is my transformer arrangement 10Vac in series with 15 Vac gives me about 38 V wich drops to 36 V whend drawing 1 A. Would that be too much input voltage? I just want to make sure.

What equipment do you have to test with?

-Rigol 1054z scope
-Cheap arbitrary function generator
-24V 4A electronic load (https://www.eevblog.com/forum/projects/arduino-based-electronic-load/)
-Two multimeters (a $5 one and a UT-61E), DPS3005 power supply module running of a wall wart.

I have included the current state of the circuit.
If you go ahead, don't bother with the CC loop until the CV loop is working properly.
I really like the topology of this power supply. I hope that it is successful. There doesn't seem to be much other viable alternatives with single  control rail referenced to the - output terminal.

Agreed. Thank you for taking the time.

Juan

[xavier60]

It would be good to find a suitable op-amp that's cheaper than the TLC072. It needs to have at least 15v/us slew rate and 10Mhz, and most important, the input common mode range has to include ground. Odd thing is the data sheet shows that the common mode includes ground only when the IC is powered by 12V. When I tested mine, I found no problem even down to 5V.

[JuanGg]

It would be good to find a suitable op-amp that's cheaper than the TLC072. It needs to have at least 15v/us slew rate and 10Mhz, and most important, the input common mode range has to include ground. Odd thing is the data sheet shows that the common mode includes ground only when the IC is powered by 12V. When I tested mine, I found no problem even down to 5V.

I have "breadboarded" the circuit shown in #129 (only the CV loop), but using a 2n3055 as the series pass transistor, 2n3904s and 2n3906s for the remaining transistors and a lm358 op-amp. (which only has 1 MHz bandwidth, but its common mode range does include ground). I used whatever I had at hand, so no big expectations. Not the best conditions either with intermittent contacts and long wires. I powered the thing from a 14 V supply and used a 9V battery for the 8V rail (voltage was near 8V). A 100K pot served as the CV ref.

With no load appart from the feedback divider, it regulates without problems from 0 to about 12V.  Using a 1K resistor as load made it oscillate, see scope screenshot.
Haven't had time for further testing. Also tried out the 6N137 opto couplers that I plan to use to comunicate both chanels of the psu.They seem to work just fine.

    Juan

[wasyoungonce]

It would be good to find a suitable op-amp that's cheaper than the TLC072. It needs to have at least 15v/us slew rate and 10Mhz, and most important, the input common mode range has to include ground. Odd thing is the data sheet shows that the common mode includes ground only when the IC is powered by 12V. When I tested mine, I found no problem even down to 5V.

I have "breadboarded" the circuit shown in #129 (only the CV loop), but using a 2n3055 as the series pass transistor, 2n3904s and 2n3906s for the remaining transistors and a lm358 op-amp. (which only has 1 MHz bandwidth, but its common mode range does include ground). I used whatever I had at hand, so no big expectations. Not the best conditions either with intermittent contacts and long wires. I powered the thing from a 14 V supply and used a 9V battery for the 8V rail (voltage was near 8V). A 100K pot served as the CV ref.

With no load appart from the feedback divider, it regulates without problems from 0 to about 12V.  Using a 1K resistor as load made it oscillate, see scope screenshot.
Haven't had time for further testing. Also tried out the 6N137 opto couplers that I plan to use to comunicate both chanels of the psu.They seem to work just fine.

    Juan

The oscillation could be due to the breadboard introducing capacitance in connections.  This is known especially if trying to build buck/boost PSUs.... as well as linear


Sent from my iPhone using Tapatalk

[xavier60]

The oscillation could be caused by layout or the slow op-amp or both.
I didn't expect to see the power transistor and main current paths on the breadboard.
I have the components shown in post #126 physically soldered together at the heat sink and separate from the  breadboard. The component connections are no longer than their own leads allow. The idea is to keep the physical loop formed by C1, Q3, C2 and R9, fairly tight.
No "load related" current should be allowed to flow through any parts for the control circuit. These unwanted current paths can also be caused by connecting test  equipment. I had trouble caused by a HF ground loop between my DSO and the bench supply that I used to power the mockup.

[xavier60]

I am aiming at 20-25 V at 0.6 A. The thing is my transformer arrangement 10Vac in series with 15 Vac gives me about 38 V wich drops to 36 V whend drawing 1 A. Would that be too much input voltage? I just want to make sure.
Juan

That voltage drop indicates a light load for that arrangement. You need about that voltage for 30V max out of the regulator output.
Need to keep the SOA rating of the transistor in mind. The transistor stress is difficult to determine  because temperature and time are also factors.
http://www.hammondmfg.com/pdf/5c007.pdf

[xavier60]

Dominant pole compensation can be temporally applied by adding 1uF between the op-amp's output and inverting input.
The 47uF output capacitor should be a low ESR type.

[JuanGg]

. The oscillation could be due to the breadboard introducing capacitance in connections.  This is known especially if trying to build buck/boost PSUs.... as well as linear
   

.    The oscillation could be caused by layout or the slow op-amp or both

Most likely. I will look for a better op-amp and make it in a perfboard.

.  I didn't expect to see the power transistor and main current paths on the breadboard.
I have the components shown in post #126 physically soldered together at the heat sink and separate from the  breadboard. The component connections are no longer than their own leads allow. The idea is to keep the physical loop formed by C1, Q3, C2 and R9, fairly tight.
No "load related" current should be allowed to flow through any parts for the control circuit. These unwanted current paths can also be caused by connecting test  equipment. I had trouble caused by a HF ground loop between my DSO and the bench supply that I used to power the mockup. 

There is no way I am going to draw more than some milliamps from that setup. Just wanted to test if the thing even worked with the components I used. If all looks ok, I'll solder the thing as proposed and socket the op- amp so it can be changed easily.I'll put the power transistor on the heatsink I have already mounted on the case.

[JuanGg]

.     That voltage drop indicates a light load for that arrangement. You need about that voltage for 30V max out of the regulator output.
Need to keep the SOA rating of the transistor in mind. The transistor stress is difficult to determine  because temperature and time are also factors.
http://www.hammondmfg.com/pdf/5c007.pdf 

The 10Vac transformer is 2A rated, while the 15 Vac is 1 A.
As per the datasheet, ≈40V c-e and 0.6 A is well within 2n3055's SOA.

Another thing to keep in mind is the current return for the fan, relay, Arduino and displays. That current has to go somewhere.

[xavier60]

.     That voltage drop indicates a light load for that arrangement. You need about that voltage for 30V max out of the regulator output.
Need to keep the SOA rating of the transistor in mind. The transistor stress is difficult to determine  because temperature and time are also factors.
http://www.hammondmfg.com/pdf/5c007.pdf 

The 10Vac transformer is 2A rated, while the 15 Vac is 1 A.
As per the datasheet, ≈40V c-e and 0.6 A is well within 2n3055's SOA.

Another thing to keep in mind is the current return for the fan, relay, Arduino and displays. That current has to go somewhere.

How are you deriving the control rail?
For simple analog control, there will be only a few milliamps of unwanted current through the shunt.
I'm also adding a 10ma load to the output which will return back to the left side of the shunt, so no problem.
Fan and relay current can be returned to the left side of the shunt also. But a micro-controller needs to be referenced to the right side.
A separate control supply will be needed.

Extra: High side sensing will solve some problems if it can be made to work properly.

[xavier60]

I have made a PCB for the regulator circuit from post #129.
Only the CV loop is working for now. It's performing the same as it did on the breadboard, very well.
I found some LM358 op-amps here, so I fitted one in place of the TLC072. It worked just fine.
I have the compensation set so that load transient recovery is 5us.

[JuanGg]

.     That voltage drop indicates a light load for that arrangement. You need about that voltage for 30V max out of the regulator output.
Need to keep the SOA rating of the transistor in mind. The transistor stress is difficult to determine  because temperature and time are also factors.
http://www.hammondmfg.com/pdf/5c007.pdf 

The 10Vac transformer is 2A rated, while the 15 Vac is 1 A.
As per the datasheet, ≈40V c-e and 0.6 A is well within 2n3055's SOA.

Another thing to keep in mind is the current return for the fan, relay, Arduino and displays. That current has to go somewhere.

How are you deriving the control rail?
For simple analog control, there will be only a few milliamps of unwanted current through the shunt.
I'm also adding a 10ma load to the output which will return back to the left side of the shunt, so no problem.
Fan and relay current can be returned to the left side of the shunt also. But a micro-controller needs to be referenced to the right side.
A separate control supply will be needed.

Extra: High side sensing will solve some problems if it can be made to work properly.

As we need an 8 V rail, it makes sense to derive it from the 10 Vac transformer, doing so from the full 25 Vac will be wasting power. The 10 Vac transformer is 2A rated as well, so there is more current available from that rail without limiting the power supply output current. I thought of doing it as attached. There is probably a simpler way, but that's what I came up with. In addition, taps can be switched depending on the output voltage to reduce power disipation. The relay can be controlled from the micro as that does not have to be particularly quick. I was thinking of using just an lm317 for the control rail. All current will be returned to the left of the shunt but from the arduino, which can take tens of mA, not drawing current from its pins. High side sensing would be definitely better, do I go ahead and buy some INA196 ICs? Any other solution?

I have made a PCB for the regulator circuit from post #129.
Only the CV loop is working for now. It's performing the same as it did on the breadboard, very well.
I found some LM358 op-amps here, so I fitted one in place of the TLC072. It worked just fine.
I have the compensation set so that load transient recovery is 5us.

I can't thank you enough for taking the time.  Nice to know that the LM358 does the job.

Juan

[perieanuo]

Put tap switching before bridge


Envoyé de mon iPad en utilisant Tapatalk

[JuanGg]

Put tap switching before bridge

How would I get the 12V rail then?

[perieanuo]

Simply.
You series the 2 secondaries, 1 bridge, and in bridge insert one end and the middle or the other end of your series.
You have 10 or 25V on your bridge then
You want 2 loads?
If you want 12V and 35 V from same gnd you need 1 bridge only.and the from your 12 or 35 get whatever voltages you want, fixed or variable.
I don't get why 2 bridges help and the current return with 35v is a bit strange


Envoyé de mon iPad en utilisant Tapatalk

[JuanGg]

Simply.
You series the 2 secondaries, 1 bridge, and in bridge insert one end and the middle or the other end of your series.
You have 10 or 25V on your bridge then
You want 2 loads?
If you want 12V and 35 V from same gnd you need 1 bridge only.and the from your 12 or 35 get whatever voltages you want, fixed or variable.
I don't get why 2 bridges help and the current return with 35v is a bit strange


Envoyé de mon iPad en utilisant Tapatalk

I don't know if I understood you correctly. I don't want to have fans, relay and displays drawing current from the 15V transformer at any time. Plus, 35V (theoretically, in reality it goes up to 38-9V) is too close or even over max input voltage of 780x series regulators or similar.

[perieanuo]

Simply.
You series the 2 secondaries, 1 bridge, and in bridge insert one end and the middle or the other end of your series.
You have 10 or 25V on your bridge then
You want 2 loads?
If you want 12V and 35 V from same gnd you need 1 bridge only.and the from your 12 or 35 get whatever voltages you want, fixed or variable.
I don't get why 2 bridges help and the current return with 35v is a bit strange


Envoyé de mon iPad en utilisant Tapatalk

I don't know if I understood you correctly. I don't want to have fans, relay and displays drawing current from the 15V transformer at any time. Plus, 35V (theoretically, in reality it goes up to 38-9V) is too close or even over max input voltage of 780x series regulators or similar.

You're free to choose what you like.
You find regulators for 40 V, I never saw one burned at this voltage.the 4 diodes for 2-nd unnecessary bridge, use them to lower voltage or any other mean to do this.
Put only one diode on 35v and you got lower V for your 7812 for ventilator...


Envoyé de mon iPad en utilisant Tapatalk

[JuanGg]

Simply.
You series the 2 secondaries, 1 bridge, and in bridge insert one end and the middle or the other end of your series.
You have 10 or 25V on your bridge then
You want 2 loads?
If you want 12V and 35 V from same gnd you need 1 bridge only.and the from your 12 or 35 get whatever voltages you want, fixed or variable.
I don't get why 2 bridges help and the current return with 35v is a bit strange


Envoyé de mon iPad en utilisant Tapatalk

I don't know if I understood you correctly. I don't want to have fans, relay and displays drawing current from the 15V transformer at any time. Plus, 35V (theoretically, in reality it goes up to 38-9V) is too close or even over max input voltage of 780x series regulators or similar.

You're free to choose what you like.
You find regulators for 40 V, I never saw one burned at this voltage.the 4 diodes for 2-nd unnecessary bridge, use them to lower voltage or any other mean to do this.
Put only one diode on 35v and you got lower V for your 7812 for ventilator...


Envoyé de mon iPad en utilisant Tapatalk

I wouldn't mind doing it that way, the thing is:
80 mA (fan) + 2*40 mA (led displays) + 40 mA (Arduino) + 30 mA (relay) makes for a 230 mA load, plus regulating circuitry. (It'll be lower on average, specially the fan, just doing a worst case calculation here).
The 15 Vac transformer is 1A rated, resulting in practical ≈620 mA max current as per the chart in #135. I would be limited to ≈ 390 mA output current, and disipating ≈5W on the regulator. Just a thought. Thanks for the input.
    Juan

[perieanuo]

I wouldn't mind doing it that way, the thing is:
80 mA (fan) + 2*40 mA (led displays) + 40 mA (Arduino) + 30 mA (relay) makes for a 230 mA load, plus regulating circuitry. (It'll be lower on average, specially the fan, just doing a worst case calculation here).
The 15 Vac transformer is 1A rated, resulting in practical ≈620 mA max current as per the chart in #135. I would be limited to ≈ 390 mA output current, and disipating ≈5W on the regulator. Just a thought. Thanks for the input.
    Juan

hi,
in real life you don't need tap switching.I don't do it even for 5 amps/0-27Vdc (limits of my 2'nd power supply).So I review my own theory, just cut off your budget and pain and series the secondaries, do the regulation and you're done.don't forget measuring current on low side will put problems for a beginner more than adding 4 $ with high side sensing, the right way to do it in your case.
If you do tap switch for future developement, that's another thing.
Anyway with your current/voltage specs you are well within SOA, you can do tap switch only for echology reasons, but buying relay+diodes+... isn't echological no more, it's just like the big lie buying electrical car will do good to the ozone
I just tried to simplify your design and get you focused on important stuff like keep it simple and get the voltage and current measurement simply and the PS stable.tap switch implies further decisional voltage comparator, watching what happens when commuting if you forgot to remove load you'll have to think what you do with the spike when you switch.imho too complicate for almost zero result.
tap switching method is for amps not for mAmps.But what the hell, I'm talking too much, as I said it before, do as you like
pierre

[xavier60]

The power transistor is mounted in direct contact with the heat sink bracket to reduce thermal resistance.
Although this power supply circuit responds to output short circuits much faster than most others, the LM358 takes 20us to take control of output current which peaks to 20 amps.  The TLC072 responds in 2us.
The 3 BD137 transistors from left to right are 8V series pass, Darlington 1st transistor and constant current  load.
Ill post an updated schematic after I do more testing. At this stage I have made only minor changes.

[coromonadalix]

@Xavier60    Your montage is bad 

You use an aluminum plate for your montage and put it back on an other "big" heatsink with an huge thermal pad, and you say you dont put an thermal pad or any micas on the power transistor ??

You want to avoid thermal loss and you're creating it the other way  loll

[not1xor1]

@Xavier60    Your montage is bad 

You use an aluminum plate for your montage and put it back on an other "big" heatsink with an huge thermal pad, and you say you dont put an thermal pad or any micas on the power transistor ??

You want to avoid thermal loss and you're creating it the other way  loll

it is anyway better than a transistor directly mounted to the heatsink via thermal pad as in this case the thermal pad is much larger.
I would rather object regarding the single screw for the power transistor since that usually do not ensure a proper contact with the heatsink.

[JuanGg]

hi,
in real life you don't need tap switching.I don't do it even for 5 amps/0-27Vdc (limits of my 2'nd power supply).So I review my own theory, just cut off your budget and pain and series the secondaries, do the regulation and you're done.don't forget measuring current on low side will put problems for a beginner more than adding 4 $ with high side sensing, the right way to do it in your case.
If you do tap switch for future developement, that's another thing.
Anyway with your current/voltage specs you are well within SOA, you can do tap switch only for echology reasons, but buying relay+diodes+... isn't echological no more, it's just like the big lie buying electrical car will do good to the ozone
I just tried to simplify your design and get you focused on important stuff like keep it simple and get the voltage and current measurement simply and the PS stable.tap switch implies further decisional voltage comparator, watching what happens when commuting if you forgot to remove load you'll have to think what you do with the spike when you switch.imho too complicate for almost zero result.
tap switching method is for amps not for mAmps.But what the hell, I'm talking too much, as I said it before, do as you like
pierre

You are right, I'll do away with tap switching, just wanted to try to as I had two transformers and some relays laying around, but as you said it's not worth it. Better keep it simple. But I'll have to look for a way of not getting control and fan current going through the 15V transistor, so I can have at least 500 mA output current.

The power transistor is mounted in direct contact with the heat sink bracket to reduce thermal resistance.
Although this power supply circuit responds to output short circuits much faster than most others, the LM358 takes 20us to take control of output current which peaks to 20 amps.  The TLC072 responds in 2us.
The 3 BD137 transistors from left to right are 8V series pass, Darlington 1st transistor and constant current  load.
Ill post an updated schematic after I do more testing. At this stage I have made only minor changes.

Not many people are willing to do what you are doing. Thank you. I am looking foward to seeing the revised schematic and how you are getting the 8V rail.

    Juan

[xavier60]

The VCC1 8V regulator has a lot of components, it needs to have good performance,  mainly a small dropout voltage which is 0.8V and tolerance to high input voltage.
Regulation will understandably be lost if the unregulated supply drops below 8.8V.
 The control circuit using the TLC072 also works properly with VCC1 set to 7V.
R23 is just to allow the current to range to include zero in case the CC op-amps input offset happens to be a particular polarity.
If D3 is replaced with a silicon diode, there will be an increase in controlled overshoot current when the output terminals are short circuited.
If the preload circuit is replaced with a resistor, transient response will suffer in situations where B-E of Q5 is not forward biased.
In its present state, I estimate the circuit is safe to work up to 3A and 30V with an adequate heatsink and tap changing.
The next step will be to check the performance with paralleled power transistors. I have also ordered some Darlingtons for testing.

Any power supply design can become dysfunctional if not laid out properly.

* C8 changed to 330pF

[xavier60]

The purpose of Q4 is to allow the CC op-amp to operate open loop until it takes control of the power supply's output.  At this point Q4 turns on, switching the compensating capacitor into circuit.
The waveform shows the shunt's voltage drop at the point the output is shorted and the current limit set to 2 amps.
The large initial spike is because of the 2us it takes the op-amp to respond. After that is an intentionally controlled overshoot.
If Q4 is jumpered out, the current will peak to about 20 amps and take hundreds of microseconds to settle.

[JuanGg]

The VCC1 8V regulator has a lot of components, it needs to have good performance,  mainly a small dropout voltage which is 0.8V and tolerance to high input voltage.
Regulation will understandably be lost if the unregulated supply drops below 8.8V.
 The control circuit using the TLC072 also works properly with VCC1 set to 7V. ...

Ok. Thank you. I will make a bill of materials and order the exact same parts, including the op-amp, so I can have a go myself. Any part number for the voltage reference? I'll order some ina196s just in case we want to try out high side sensing. Any other component suggestions that may come in handy are welcome.
I suppose adjusting the current range is a matter of resistor values.
I was thinking of mounting the perfboard as attached (not many options given the space I have to work with) Rectifiers and filter caps will go between the transformers, no tap switching.
Will a well-laid out perfboard do? If not I can make a pcb.
Also, will the 8 V rail be able to handle a couple hundred milliamps, or would it be better to have a separate regulator for the fan, displays and arduino?

    Thanks again, Juan

[Kleinstein]

The reference could be something like TL431, at least the circuit looks like it is made for this reference. Changing the reference part is likely one of the easier parts. Another interesting (low noise) reference would be the LM329.

It may be tricky to use the VCC1 also use for other uses, as the current would also flow through the shunt. So about the only part one might consider running from here is something like an DAC if needed.  One may even consider a different supply (negative side) for the OP.

Alone from the noise and interference, I would use a separate regulator for the µC and display.

Perf board should be OK, if one takes some care with star ground and similar. But this is also important for a board.
If the filter caps are separate, it may take an extra small capacitor on the perf board. 

The shunt should of cause higher value for the lower current at some 500 mA.

[JuanGg]

The reference could be something like TL431, at least the circuit looks like it is made for this reference. Changing the reference part is likely one of the easier parts. Another interesting (low noise) reference would be the LM329.

It may be tricky to use the VCC1 also use for other uses, as the current would also flow through the shunt. So about the only part one might consider running from here is something like an DAC if needed.  One may even consider a different supply (negative side) for the OP.

Alone from the noise and interference, I would use a separate regulator for the µC and display.

Perf board should be OK, if one takes some care with star ground and similar. But this is also important for a board.
If the filter caps are separate, it may take an extra small capacitor on the perf board. 

The shunt should of cause higher value for the lower current at some 500 mA.

Thank you. That's what I was thinking, a PWM driven fan and multiplexed 7-segments would be no good for the control rail. I'll add a separate regulator, before the shunt. The micro still has to be referenced after the shunt I'm afraid. Displays take 5V, clk and data, I don't know if they can be driven with the micro referenced to the other side of the shunt...will have to test it.
I have some 0R1 resistors around, I can parallel more or less to get different shunt values. Still looking at high side current sensing, which would certainly make things simpler.

    Juan

[nick_d]

Regarding obsolete parts available on Ebay, many are fakes. Do not trust them. If you have a WORKING circuit and can drop in the chip then fine, it will work or not (but may prove unreliable over time). On the other hand if you are developing then you can end up implementing elaborate workarounds for non-problems due to weird characteristics of the fake chips and wasting enormous development time. Don't do it.
cheers, Nick

[xavier60]

The reference could be something like TL431, at least the circuit looks like it is made for this reference. Changing the reference part is likely one of the easier parts. Another interesting (low noise) reference would be the LM329.

It may be tricky to use the VCC1 also use for other uses, as the current would also flow through the shunt. So about the only part one might consider running from here is something like an DAC if needed.  One may even consider a different supply (negative side) for the OP.

Alone from the noise and interference, I would use a separate regulator for the µC and display.

Perf board should be OK, if one takes some care with star ground and similar. But this is also important for a board.
If the filter caps are separate, it may take an extra small capacitor on the perf board. 

The shunt should of cause higher value for the lower current at some 500 mA.

Thank you. That's what I was thinking, a PWM driven fan and multiplexed 7-segments would be no good for the control rail. I'll add a separate regulator, before the shunt. The micro still has to be referenced after the shunt I'm afraid. Displays take 5V, clk and data, I don't know if they can be driven with the micro referenced to the other side of the shunt...will have to test it.
I have some 0R1 resistors around, I can parallel more or less to get different shunt values. Still looking at high side current sensing, which would certainly make things simpler.

    Juan

Having circuitry on both sides of the shunt could be made to work with a lot of thought because the shunt voltage can be kept low enough to not affect even TTL, but transients should be expected that would need to be filtered out which means using LC filtering. It gets messy.
An independent supply  from a separate winding or transformer is best.
I have tested the design with a 0R1 shunt if needed, no problem. I would prefer 0R05 or less.
Perf board will be fine if the guideline in post #126 is followed. The most important detail is the single connection between the - output and the ground of the control circuit. I have represented this in the schematic also.
There is an advantage to using the same type parts in the schematic, mainly the transistors because it has all been tested.
 The voltage divider and current set resistor,R12,  can be changed to suit.

[xavier60]

The position of the shunt resistors on my PCB is a compromise. Ideally they should be in the middle of the PCB, but they would have obscured access to the transistors mounted on the heat sink bracket.
Yes, U2 is a TL431CLP. Ill update the schematic in post #154.

[JuanGg]

Having circuitry on both sides of the shunt could be made to work with a lot of thought because the shunt voltage can be kept low enough to not affect even TTL, but transients should be expected that would need to be filtered out which means using LC filtering. It gets messy.
An independent supply  from a separate winding or transformer is best.
I have tested the design with a 0R1 shunt if needed, no problem. I would prefer 0R05 or less.
Perf board will be fine if the guideline in post #126 is followed. The most important detail is the single connection between the - output and the ground of the control circuit. I have represented this in the schematic also.
There is an advantage to using the same type parts in the schematic, mainly the transistors because it has all been tested.
 The voltage divider and current set resistor,R12,  can be changed to suit.

Ok, I'll try to look for more transformers to power the micro, fans and displays.
I have attached a bill of materials for your chematic, I hope I haven't missed anything. Couldn't find some of the transistors, added some similar ones: BD135 instead of BD137 and BD559 instead of BD558, I've gone through the datasheets and the characteristics are almost the same. I'll order this tonight, as well as some smd to dip adapters and misc stuff. It's going to take at least a week or two, I hope I can get something done this christmas.
I'll prototype the thing on a perfboard, but at the prices PCB services are running at, it may be an option to have some made later on (as I'm planing on having two channels... we'll see)
    Juan

[xavier60]

Ok, I'll try to look for more transformers to power the micro, fans and displays.
I have attached a bill of materials for your chematic, I hope I haven't missed anything. Couldn't find some of the transistors, added some similar ones: BD135 instead of BD137 and BD559 instead of BD558, I've gone through the datasheets and the characteristics are almost the same. I'll order this tonight, as well as some smd to dip adapters and misc stuff. It's going to take at least a week or two, I hope I can get something done this christmas.
I'll prototype the thing on a perfboard, but at the prices PCB services are running at, it may be an option to have some made later on (as I'm planing on having two channels... we'll see)
    Juan

Although the BD135 will be fine, try to keep all transistors above 60V where necessary, Is BD139 available instead?
There should be no BD558, only BC558, maybe a misprint. Although I haven't used the transistor types consistently, use BC546, BC556 where the voltage is high and BC548,  Bc558 for low voltage positions or just about any BC series at hand.
The BC546, BC556, BC547, BC557, BC548, BC558 are very commonly used general purpose signal transistors. You should take the time to get familiar with the data sheets. Take note of the hFE or Current Gain grading. If you get the ungraded versions, they should be tested but usually have plenty of current gain.
I have ordered more TIP35C transistors on Aliexpress, very cheap. Ill be checking to see if they are fakes.
I tested with an actual mains transformer last night. There was a few millivolts of ripple at the output. I have changed C8 to 330pF

[xavier60]

I have attached a bill of materials for your chematic, I hope I haven't missed anything. Couldn't find some of the transistors,
[/quote]
The XLsX file won't open for me, trying 3 different viewers.
Can you upload a text file?

[JuanGg]

Ok, I'll try to look for more transformers to power the micro, fans and displays.
I have attached a bill of materials for your chematic, I hope I haven't missed anything. Couldn't find some of the transistors, added some similar ones: BD135 instead of BD137 and BD559 instead of BD558, I've gone through the datasheets and the characteristics are almost the same. I'll order this tonight, as well as some smd to dip adapters and misc stuff. It's going to take at least a week or two, I hope I can get something done this christmas.
I'll prototype the thing on a perfboard, but at the prices PCB services are running at, it may be an option to have some made later on (as I'm planing on having two channels... we'll see)
    Juan

Although the BD135 will be fine, try to keep all transistors above 60V where necessary, Is BD139 available instead?
There should be no BD558, only BC558, maybe a misprint. Although I haven't used the transistor types consistently, use BC546, BC556 where the voltage is high and BC548,  Bc558 for low voltage positions or just about any BC series at hand.
The BC546, BC556, BC547, BC557, BC548, BC558 are very commonly used general purpose signal transistors. You should take the time to get familiar with the data sheets. Take note of the hFE or Current Gain grading. If you get the ungraded versions, they should be tested but usually have plenty of current gain.
I have ordered more TIP35C transistors on Aliexpress, very cheap. Ill be checking to see if they are fakes.
I tested with an actual mains transformer last night. There was a few millivolts of ripple at the output. I have changed C8 to 330pF

I ordered already, from "LCSC", let's see how it works out. Yes, BD139 was available  . Oh well. And yes, BD558 was a typing error.
I have some BC547 and BC557 transistors around. Anyway, I got ten signal transistors of each kind (minimum order quantity) and four TIP3455C so I should have plenty to play with. I'll eventually end up using them I suppose. While they arrive (at least a week or two), I'll see if I can finish the transistor labs on "Learning the art of electronics" and go through the datasheets so I can get a better understanding.

I have attached a bill of materials for your chematic, I hope I haven't missed anything. Couldn't find some of the transistors,

The XLsX file won't open for me, trying 3 different viewers.
Can you upload a text file?

Uploaded it to the original post. I should have thought of that.
    Juan

[xavier60]

I get no search results for " TIP3455C "

[JuanGg]

I get no search results for " TIP3455C "

This thing manages to mess everything up. Here is a screenshot and a pdf. Sorry for the inconvenience, I should be the one dealing with this stuff.

    Juan

[xavier60]

The top trace represents a 0 to 6 amp ramped load on the regulator's output.
The bottom trace was at the Base of Q2. The transconductance is very linear.
The gm of 40 is a bit higher than I would have preferred.
I did notice a problem, some voltage fluctuation that is present all the way up to the Base of the TIP35C, but not at the Emitter.
It could mean that the transistor is failing. It is a spare that was left over from a repair job.
I bought them on Aliexpress and are ST branded. I hope the ones on order are ok.

[JuanGg]

The top trace represents a 0 to 6 amp ramped load on the regulator's output.
The bottom trace was at the Base of Q2. The transconductance is very linear.
The gm of 40 is a bit higher than I would have preferred.
I did notice a problem, some voltage fluctuation that is present all the way up to the Base of the TIP35C, but not at the Emitter.
It could mean that the transistor is failing. It is a spare that was left over from a repair job.
I bought them on Aliexpress and are ST branded. I hope the ones on order are ok.

Are you putting 6A through this thing?
The ones I ordered are ST as well.
    Juan

[nick_d]

No way the AliXpress ones are actual ST. They will be fakes definitely. That does not mean they are no good. Likely they will have similar characteristics to what they are marked as, but not always. See my earlier post and do not waste valuable development time tracking down possible weird characteristics of the fakes. If it works great -- if it doesn't, swap in the Digikey or other genuine part before proceeding.
cheers, Nick
PS I know this from bitter experience... Hello fake LM386...

[xavier60]

The TIP35C transistors I have ordered are supposed to be the newer TO-247 types. It would be possible to test the transistor's spec somehow, but how does one check for poor quality bonding?
The control circuit  can easily drive the power transistor to  high currents, but the transistor needs to be kept cool and in its Safe Operating Area.
I have increased the current carrying capacity  of the PCB by placing 0.8mm solid core copper wire along the ground track.
The BD137, TIP35C Darlington has a current gain of over 10,000. There is only about 0.7ma through R4.

[xavier60]

I should have been more suspicious when I had to file the tab flat on the ST marked fake.
This is what I have on order. I wonder what brand they are?

Extra: The description says "Usongshine" . I can't find anything about it.

https://www.aliexpress.com/item/10PCS-LOT-TIP35C-TIP35-35C-35-TO3P-transistor-new-original/32810647217.html?spm=2114.search0104.3.1.bd184394dD2rIC&ws_ab_test=searchweb0_0,searchweb201602_3_10065_10068_10130_10547_319_317_10548_10696_453_10084_454_10083_10618_10307_538_537_536_10131_10132_10133_10059_10884_10887_100031_321_322_10103,searchweb201603_51,ppcSwitch_0&algo_expid=32f8061b-7b70-4998-af35-5105fcc333de-0&algo_pvid=32f8061b-7b70-4998-af35-5105fcc333de

[xavier60]

The top trace represents a 0 to 6 amp ramped load on the regulator's output.
The bottom trace was at the Base of Q2. The transconductance is very linear.
The gm of 40 is a bit higher than I would have preferred.
I did notice a problem, some voltage fluctuation that is present all the way up to the Base of the TIP35C, but not at the Emitter.
It could mean that the transistor is failing. It is a spare that was left over from a repair job.
I bought them on Aliexpress and are ST branded. I hope the ones on order are ok.

Are you putting 6A through this thing?
The ones I ordered are ST as well.
    Juan

Let me know when you have received the parts, I will update the schematic.
I have made a few minor changes. The gm has been reduced to 20.
There are reasons for the current overshoot being allowed to occur when the output is shorted.
I have added a method for easily setting the overshoot.
I have not introduced any new part values.

[JuanGg]

The top trace represents a 0 to 6 amp ramped load on the regulator's output.
The bottom trace was at the Base of Q2. The transconductance is very linear.
The gm of 40 is a bit higher than I would have preferred.
I did notice a problem, some voltage fluctuation that is present all the way up to the Base of the TIP35C, but not at the Emitter.
It could mean that the transistor is failing. It is a spare that was left over from a repair job.
I bought them on Aliexpress and are ST branded. I hope the ones on order are ok.

Are you putting 6A through this thing?
The ones I ordered are ST as well.
    Juan

Let me know when you have received the parts, I will update the schematic.
I have made a few minor changes. The gm has been reduced to 20.
There are reasons for the current overshoot being allowed to occur when the output is shorted.
I have added a method for easily setting the overshoot.
I have not introduced any new part values.

I should receive them within a week from now. I'll let you know. Thanks again. I bought a few of each just in case.

I'll start winter break tomorrow, so I'll have more time. I can start planning the perfboard layout or continue working on the case. I'll also look for transformers for the micro and displays.
    Juan

[ignilux]

Gents-

Following along diligently, but finding it a little difficult to keep up. I've attached an annotated schematic highlighting my main questions, but I'll attempt to summarize below:

• Does PS2 represent a rectified mains transformer input? If so, what is the voltage? I saw a number of transformers discussed earlier on
• Does the uppermost line leading off to the right represent the positive output terminal?
• Does the lowermost line represent the negative output terminal?
• You've discussed using a microcontroller. Do the CV/CC Ref's correspond to control inputs, or are the related to the TL431 reference?
• VCC1 just represents the regulated control voltage created by the TL431 circuit, right? It is NOT an isolated supply?

I've built a few '317 and '338 style power supplies, blew 'em up, and I'm working on finding a suitable topology for my next project. I've been simulating various circuits involving throwing a current control loop around various voltage regulator ICs, but have so far been frustrated trying to compensate the darn things. Anyway, just wanted to say that this is a neat project, and I'm keeping an eye on it. Thanks to everyone who has contributed thus far!

[xavier60]

Gents-

Following along diligently, but finding it a little difficult to keep up. I've attached an annotated schematic highlighting my main questions, but I'll attempt to summarize below:

• Does PS2 represent a rectified mains transformer input? If so, what is the voltage? I saw a number of transformers discussed earlier on
• Does the uppermost line leading off to the right represent the positive output terminal?
• Does the lowermost line represent the negative output terminal?
• You've discussed using a microcontroller. Do the CV/CC Ref's correspond to control inputs, or are the related to the TL431 reference?
• VCC1 just represents the regulated control voltage created by the TL431 circuit, right? It is NOT an isolated supply?

I've built a few '317 and '338 style power supplies, blew 'em up, and I'm working on finding a suitable topology for my next project. I've been simulating various circuits involving throwing a current control loop around various voltage regulator ICs, but have so far been frustrated trying to compensate the darn things. Anyway, just wanted to say that this is a neat project, and I'm keeping an eye on it. Thanks to everyone who has contributed thus far!

Yes, the schematic follows conventional layout. Unregulated input on the left and regulated output on the right.
To reduce clutter, I have omitted the obviously needed bits, such as the usual transformer, rectifiers and large capacitors, or just about any DC supply such as a SMPS.
 The references can be from a micro-controller  via Digital Pots, DAC or filtered PWM. Or simply from mechanical Pots powered from VCC1 8V.
When Pots are used, make certain that the circuit is fail safe to the possibility of loss of wiper contact which is a real probability with cheap 10 turn Pots. Oiling(PAO) has been fixing them. Read my posts for more details.
 The recommended absolute input voltage is limited  by the transistor spec, 60V.
The real world limits are determined by difficult to measure and calculate interaction of factors such as operating current, power transistor Safe Operating Area rating and the heat sink's thermal resistance to ambient. Transformer tap switching helps a lot.
VCC1 may need to be from an isolated supply if more is to be powered than currently shown.
Q8's dissipation will be rather  high at higher loads because of its large voltage drop. And also, this current will be measured by the shunt resistor.

I have been testing the regulator circuit a lot with good results.
It's yet to be seen if the results can be reproduced.
Transistors can have a large difference in current gain from one device to another of the same part type.
Well designed circuits tolerate this variation.

[xavier60]

I have done just about enough tweaking for now.
I have deleted a transistor from the preload circuit.
I used the method in the linked PDF to do Phase Margin tests.
My measurements indicate a PM of 90° at the 0db crossing of 60Khz.
The regulator remains stable with no output capacitor.
The PM is 90° again at the much higher frequency of 300Khz.

http://www.ti.com/lit/an/snva364a/snva364a.pdf

[Kleinstein]

The phase margin test is good for the given load the output. The case without the capacitor at the output is not really relevant, as the difficult case is normally a load with low ESR capacitance combined with a kind of current sink.  This would be the hard case to judge stability. Under these difficult condition the phase margin will usually go down quite a bit - that is OK unless the phase margin turns negative.
A good phase margin with an easy load not always guaranties stability with a difficult load.

Looking at the circuit, I am not sure the regulator will be stable for all reasonable loads. Something like 100 µF with 10 mOhms ESR and some current sink could cause instability.

The output impedance can help to judge stability at different loads.

[xavier60]

The phase margin test is good for the given load the output. The case without the capacitor at the output is not really relevant, as the difficult case is normally a load with low ESR capacitance combined with a kind of current sink.  This would be the hard case to judge stability. Under these difficult condition the phase margin will usually go down quite a bit - that is OK unless the phase margin turns negative.
A good phase margin with an easy load not always guaranties stability with a difficult load.

Looking at the circuit, I am not sure the regulator will be stable for all reasonable loads. Something like 100 µF with 10 mOhms ESR and some current sink could cause instability.

The output impedance can help to judge stability at different loads.

The capacitor(NCC KME 47-100) on the output that I thought was low ESR, isn't. I measured about 500mΩ ESR.
When I replaced it with a 100µF MLCC, the worst case PM dropped to 15°. It also causes the load transient response to have a 50% overshoot.
I guess this is why I see regulator designs that have multiple series RC's across the output.
Testing with a constant current load didn't seem to cause any problem.

[Kleinstein]

The capacitors at the output have different purpose. A low ESR capacitor at the output can help to buffer fast transients, where the active regulator can no longer react. However it also is a difficult load to the regulator and make regulation more tricky with more overshoot. A capacitor with defined ESR (e.g. in the 1 Ohms range) does not help much to buffer the output and deliver much power. The purpose is mainly to improve stability.  Still it makes regulation at other frequencies slower. So ideally the capacitors should be relatively small, just enough to get good response.

50 % overshoot for a tricky load like the 100 µF capacitor is quite good.  The 100 µF are more like not a good idea to have at the output though. The 100 µF with 0.5 Ohms ESR might be suitable for the output, to provide some damping. Chances are one could use a smaller capacitor.

[xavier60]

I temporarily fitted the 100µF MLCC just for testing the Phase Margin.
I can eliminate the overshoot by peaking Q1, but the regulator becomes unstable with no output capacitor although this does  not  have to be a requirement.
If this poor tolerance to low ESR capacitors is generally expected of regulators, the design that I have posted should not be regarded as a failure.
Because it's quite happy with a standard 47µF electrolytic across the output.
I want this design to be successful in having good performance without being overly complicated.
The transient response is currently 15µs.

[Kleinstein]

Having a low ESR capacitor at the output is a difficult situation for any voltage regulator. So a low phase margin for this case is kind of normal -  not so good designs have a chance to oscillate under such conditions. I would consider only 50 % overshoot as still good for the difficult conditions. It can actually get better with the 47 µF electrolytic in parallel.

For this type of regulator with a relative high impedance output stage it is normal to require some kind of output capacitor.

A 47 µF electrolytic cap with some ESR sounds like a good solution.  It is still relatively small to what other lab supplies have.

[JuanGg]

I am getting a bit lost with all this capacitors and phase margins. I'll have a careful read.
Anyway, i have done some work on the case, mainly on the front panel, see photos and schematic below. I think it's rather neat as it allows the arduino to drive the indicator leds and read the pushbuttons from the same pin, alternating it as an input our output. 7 segment displays are just modules from e-bay.

I have gotten hold of some more transformers, so this is all I have (see last photo).
From left to right,
    - two 12 Vac 500 mA transformers (to power the arduino, fans and displays),
    - 22 Vac 2.2 A,
    - 2 taps @ 33 Vac 2A each . (Measured with no load, 2 A because wire gauge is the same as in other 2 A transformer)
    - and on the right what I had before, 2 taps@ 10 Vac 2A each, two 15 Vac 1 A.

Here is the options I see, all using the 12 Vac transformers for the arduino and displays:
    -Use what I have currently installed, 10 Vac tap + 15 Vac transformer per channel.
    -Use two 10Vac taps in series  (so 20 Vac) for one channel and the 22 Vac transformer for the other.
    -Use one 33 Vac tap of the big transformer per channel.  This will result on a rather high ≈ 45 Vdc input voltage, perhaps too close to the
     max 50 V of the filter caps.

That's it. Sorry for the nonsense. Thank you.

[xavier60]

A lot of options there. I think that he 33Vac transformer is just too high for most uses. The power supply would have to be derated to 500ma or less to keep the power transistors safe.
You could use the 22Vac transformer to make one medium voltage channel at about 1.5 amps.
And the 2x10Vac+2X15Vac transformer for a switchable low voltage high current  /  high voltage low current channel.
The safe minimum output voltage for each range at full current will be determined by the heat sinks.
Keep in mind that the thermal washers will add about 0.8° per watt of thermal resistance. You could easily add over temperature protection anyway.

[xavier60]

If I decide to go further with this project, this transformer looks nice.
 https://www.jaycar.com.au/12v-30v-100va-6a-multi-tapped-dual-type-2170-transformer/p/MM2015
Switches or relays could be used to select the 12/15 taps as well as series or parallel for higher current on the low voltage ranges.
Relays would allow auto switching to be added which would greatly lower heat dissipation  if the output is shorted on the high voltage ranges.

[JuanGg]

I think I'll leave it as-is for simplicity's shake. I can make changes later on.
    Juan

[xavier60]

I think I'll leave it as-is for simplicity's shake. I can make changes later on.
    Juan

Yes, because you are going for lower output current, the secondary switching isn't really needed.

[JuanGg]

All right. I'll make a perfboard with the main rectifier and filter cap, plus the auxiliary PSU: another small rectifier taking 12 Vac from the small transformers and regulating it to 12 V with a LM7812 regulator (for the fan) and to 5 V with an LM317 (arduino and displays).
Or maybe I should put the auxiliary PSU on the main board?(second photo, on the right, mounted vertically).
Regulators will be disipatting 0.5 W and 1.5 W respectively. As per the datasheet, LM7812 has a Juction to Ambient thermal resistance of 20ºC/W, so it should heat up to 35º assuming an ambient temperature of 25ºC (fine withought heatsink). LM317 has a Junction to Ambient thermal resistance of 38ºC/W, so it would heat up to 80ºC, I suppose it'll be fine without heatsink, but I can bend it and solder the tab to the perfboard just in case.

    Juan

[xavier60]

I also prefer having a separate board for the AUX rails along with the main rectifier. Keep ground return current paths for the 12V and 5V rails separate. They should meet only at a common point on the regulator board.
IF you power the op-amp with the 12V rail, don't run a ground wire from the AUX supply board to the regulator board. The only ground connection to the regulator board should be the 5V ground from the control board.
The fan might impose some ripple onto the 12V rail.

Extra: Im not certain how the LEDs should be powered. What voltage do the displays need?

[xavier60]

I see it's 5V for the displays. I need to think about that.
Does the  micro-controller have its own accessible internal reference?

[JuanGg]

5 V would be for the micro and displays. I don't know if the displays would make the 5 V rail noisy and if that would impact performance. The arduino itself has a 5 V reg, but its an smd package and I don't think that it can handle 1 W. It can be used for control circuitry though.

What are the op-amps powered from? I thought they would be from the 8V rail. Fans, being PWM driven, would definitely add some ripple to the 12 V rail.

The micro has an internal 1.1 V reference for the ADC's that can be enabled or disbled.

    Juan

[xavier60]

5 V would be for the micro and displays. I don't know if the displays would make the 5 V rail noisy and if that would impact performance. The arduino itself has a 5 V reg, but its an smd package and I don't think that it can handle 1 W. It can be used for control circuitry though.

What are the op-amps powered from? I thought they would be from the 8V rail. Fans, being PWM driven, would definitely add some ripple to the 12 rail.

The micro has an internal 1.1 V reference for the ADC's that can be enabled or disbled.

    Juan

You can stay with the 8V regulator as in my schematic to power the op-amps. Using the 12V instead would have eliminated many parts but might cause complications.
If the internal reference is used, there sould be no need to worry about fluctuations on the 5V

[JuanGg]

I have like ten LM7812 regulators, so I could use one for the fans and another one for the regulating circuitry. I have ordered all parts, so any option is feasible. I also have some LM317's.
    Juan

[xavier60]

I have like ten LM7812 regulators, so I could use one for the fans and another one for the regulating circuitry. I have ordered all parts, so any option is feasible. I also have some LM317's.
    Juan

Just occurred to me that increasing the supply voltage to the op-amps will affect the response to overloads as well as some other component value changes. You would need to use an LM317 set to 8V.

[JuanGg]

Just occurred to me that increasing the supply voltage to the op-amps will affect the response to overloads as well as some other component value changes. You would need to use an LM317 set to 8V.   

Ok. So I would have a 7812 regulator for the fan, an LM317 for the 8 V control rail, and another LM317 / 7805 for the displays. The Arduino can be powered from either 12 or 8 V and it'll generate it's own 5 V.
Would it be adequate to daysy-chain all regulators: Unreg. V > 12 V > 8 V > 5 V?
    Juan

[xavier60]

Just occurred to me that increasing the supply voltage to the op-amps will affect the response to overloads as well as some other component value changes. You would need to use an LM317 set to 8V.   

Ok. So I would have a 7812 regulator for the fan, an LM317 for the 8 V control rail, and another LM317 / 7805 for the displays. The Arduino can be powered from either 12 or 8 V and it'll generate it's own 5 V.
Would it be adequate to daysy-chain all regulators: Unreg. V > 12 V > 8 V > 5 V?
    Juan

I would use the 12V for the fans only and use the 8V for the Arduino as well as for the regulator control rail. At the moment, I can't see any big problem with daysy-chaining so long as it helps to share the dissipation.

[soldar]

Would it be adequate to daysy-chain all regulators: Unreg. V > 12 V > 8 V > 5 V?

I would rather have them be independent as it makes troubleshooting and repair easier. OTOH it could be a case where staggering them would balance power dissipation better. Your call.

[xavier60]

Also, the regulator mentioned earlier that is expected to reach 80° bothers me.  It's difficult to know what temperatures components really reach after the case is closed unless there is some continuous air flow. 

[JuanGg]

Also, the regulator mentioned earlier that is expected to reach 80° bothers me.  It's difficult to know what temperatures components really reach after the case is closed unless there is some continuous air flow. 

I'll build the thing up on a headboard, test it and measure temperatures just in case. I am thinking of deriving both the 8V and the 5 V rail from the 12 V one. Another thing to keep in mind is return current paths. Fan and displays wouy go before the shunt. I think the displays will be fine being driven from the other side of the shunt. I'll just put in a couple resistors in series with the clk and data lines, just in case.
    Juan

[xavier60]

Also, the regulator mentioned earlier that is expected to reach 80° bothers me.  It's difficult to know what temperatures components really reach after the case is closed unless there is some continuous air flow. 

I'll build the thing up on a headboard, test it and measure temperatures just in case. I am thinking of deriving both the 8V and the 5 V rail from the 12 V one. Another thing to keep in mind is return current paths. Fan and displays wouy go before the shunt. I think the displays will be fine being driven from the other side of the shunt. I'll just put in a couple resistors in series with the clk and data lines, just in case.
    Juan

Because AUX current will be from an independent supply rather than from the main unregulated rail, you should be able to reference everything to the top side of the shunt. Do a sketch to confirm.

[JuanGg]

That's right  sorry. I'll draw an schematic anyways.

PD: Merry Christmas!

    Juan

[JuanGg]

I have soldered the perfboard with the raw PSU and the 12 V regulator. I have mounted the two transformers for the auxiliary supplies as well. There are now a total of 5 transformers in the case.

12 V regulator seems to work well and doesn't even get warm drawing 300 mA. It has little noise with no load and under resistive loads, but when connecting one of the 12 V fans, I get what's on the screenshot below. I suppose it's due to the fact of the fan being a brushless motor and drawing current spikes or whatever. I'll try adding some more output capacitance (now i'ts just a 0.1 uF as recomended by the datasheet), but what I've tried so far (10 uF, 100 uF) does not have an effect.
    Juan

[xavier60]

Looks like a combination of things. Those small high speed fans must draw very high current pulses along with the secondary voltage being on the low side for a  regulated  12V rail. Increasing the size of the capacitor after the bridge will help but looks like it will never run both fans well at full speed.

[Kleinstein]

Extra capacitance after the regulator has a limited effect, as ideally the voltage is not changing much. Some of the fans even have quite some capacitance inside. In this case a resistor (e.g. 10 Ohms range) before the fan can help. So the fan would still see voltage drops, but the rest of the circuit would see less.

[JuanGg]

Capacitor after the bridge is 1000 uF.  This 12 V regulator is running one of the fans, as there will be an identical board for the other channel. Fans do not need to run at full speed. I was thinking of using PWM, and this may make it worse. I'll try with the resistor.
I am just worried that this noise can spread to the rest of the PSU, as I am deriving the 8 V and 5 V rail from this 12 V.
I could try to derive all rails from the unregulated supply, to make them more independent. I'd have to desolder the 12 V reg from the rectifier board, and put all regulators on the main board.
    Juan

[xavier60]

Capacitor after the bridge is 1000 uF.  This 12 V regulator is running one of the fans, as there will be an identical board for the other channel. Fans do not need to run at full speed. I was thinking of using PWM, and this may make it worse. I'll try with the resistor.
I am just worried that this noise can spread to the rest of the PSU, as I am deriving the 8 V and 5 V rail from this 12 V.
I could try to derive all rails from the unregulated supply, to make them more independent. I'd have to desolder the 12 V reg from the rectifier board, and put all regulators on the main board.
    Juan

A series resistor should make things better but no matter what you do, expect the 8V regulator to drop out. The 5V regulator might be ok. And they need to be powered directly from the unregulated low voltage rail. Because the 8V rail isn't being used for the reference, some ripple shouldn't cause a big problem until you find a more suitable transformer.
Try adding more capacitance to see if it causes a useful drop in ripple.

[JuanGg]

A series resistor should make things better but no matter what you do, expect the 8V regulator to drop out. The 5V regulator might be ok. And they need to be powered directly from the unregulated low voltage rail. Because the 8V rail isn't being used for the reference, some ripple shouldn't cause a big problem until you find a more suitable transformer.
Try adding more capacitance to see if it causes a useful drop in ripple.

I'll desolder the 12 V regulator from the rectifier board, and run the raw DC to the main board, where I'll regulate it to 12, 8 and 5 V. I'll try adding the series resistor. What do you mean by a more suitable transformer?

    Juan

[xavier60]

A series resistor should make things better but no matter what you do, expect the 8V regulator to drop out. The 5V regulator might be ok. And they need to be powered directly from the unregulated low voltage rail. Because the 8V rail isn't being used for the reference, some ripple shouldn't cause a big problem until you find a more suitable transformer.
Try adding more capacitance to see if it causes a useful drop in ripple.

I'll desolder the 12 V regulator from the rectifier board, and run the raw DC to the main board, where I'll regulate it to 12, 8 and 5 V. I'll try adding the series resistor. What do you mean by a more suitable transformer?

    Juan

Something with higher voltage and maybe higher current. My last power supply project uses a 15Vac 10VA transformer to supply a 12V regulator which powers a 60mm fan, but the fan draws only 220mA at full speed.

[JuanGg]

Something with higher voltage and maybe higher current. My last power supply project uses a 15Vac 10VA transformer to supply a 12V regulator which powers a 60mm fan, but the fan draws only 220mA at full speed.

My fan draws about 100 mA at full speed. I'll change my 500 mA transformer if I come across a suitable one.

Parts I ordered just arrived. They sent 470 uF caps instead 47 uF... but the rest seems fine. I suppose it's time to start doing the perfboard layout.

    Juan

[JuanGg]

Another thing is mounting power devices to the heatsink. Would it be ok to have the TIP35 mounted directly with no thermal washers and leave everything else without heatsink/bent and soldered to the board?
    Juan

[xavier60]

Another thing is mounting power devices to the heatsink. Would it be ok to have the TIP35 mounted directly with no thermal washers and leave everything else without heatsink/bent and soldered to the board?
    Juan

It's ok to mount the TIP35C with no washer. The dissipation of Q3 and Q9  will vary a lot with operating conditions and difficult to calculate with certainty. mainly with Q3 because the gain of the TIP35C is unknown. They would be ok for most conditions for testing but I would not permanently leave them un heat sunk. Q3 won't need a washer.
If you leave Q9 un heat sunk, you might have to decrease the preload current.

[xavier60]

Something that needs to be aware of, some CPU fans control their own speed according to temperate sensed in the hub. If the fan is always drawing in fresh air, it should stay at low speed. All of my stock Intel P4 fans have this behavior.

[JuanGg]

   It's ok to mount the TIP35C with no washer. The dissipation of Q3 and Q9  will vary a lot with operating conditions and difficult to calculate with certainty. mainly with Q3 because the gain of the TIP35C is unknown. They would be ok for most conditions for testing but I would not permanently leave them un heat sunk. Q3 won't need a washer.
If you leave Q9 un heat sunk, you might have to decrease the preload current. 

I have some to-220 insulating pads. I suppose they could be used on the To-126 this transistors come in. I'll heatsink everything.

   Something that needs to be aware of, some CPU fans control their own speed according to temperate sensed in the hub. If the fan is always drawing in fresh air, it should stay at low speed. All of my stock Intel P4 fans have this behavior. 

I'll check for that.

Instead of laying the perfboard on paper I thought I'd have a go at designing a PCB, even if it's just for component placement. I am using Autodesk Eagle, just because it can be synced to Fusion 360, so I can put PCBs on my mechanical assemblies.  For now, I have just input the schematic.

Another thing, can I use any schottky diode, (say IN5817). Instead of the Bat 85 which was only available on Sot-323, and I'll prefer soldering an axial diode on a perfboard. I can use the sot-323 if there's no other option.
    Juan

[Kleinstein]

The diodes for joining the control signals are not that critical. So 1N5817 should be Ok. Using normal 1N4148 might effect the minimal voltage - it depends one details of the rest of the plan.

[xavier60]

D3 can be left out for now until the CV loop is tested. Ill test the circuit with a 1A schottky diode here, it's likely to alter the allowed current overshoot.  R26 might have to be changed.
I use machined IC socket pins where I expect to have to later change a component value.

[JuanGg]

I have designed a PCB from the schematic. It's far from finished, I just wanted to have a go at it. It's the first PCB I lay out (except a couple rather simple hand made ones) so don't expect much. Its probably a non optimal layout, but I may have an idea on where components could be placed.

D3 can be left out for now until the CV loop is tested. Ill test the circuit with a 1A schottky diode here, it's likely to alter the allowed current overshoot.  R26 might have to be changed.
I use machined IC socket pins where I expect to have to later change a component value.

Ok, thank you. I'll prototype the CV loop on a perfboard.

    Juan

[ArthurDent]

On the fans putting garbage on the +12VDC going to the rest of the circuitry, just add a second 7812 and associated caps to only run the fans. It is cheap and will effectively isolate the fan noise from the rest the stuff.

[JuanGg]

On the fans putting garbage on the +12VDC going to the rest of the circuitry, just add a second 7812 and associated caps to only run the fans. It is cheap and will effectively isolate the fan noise from the rest the stuff.

The 12 V reg on the board is going to power just the fan, 5 V reg will do the displays (maybe arduino as well) and the 8 V the analog circuitry.

    Juan

[xavier60]

I have designed a PCB from the schematic. It's far from finished, I just wanted to have a go at it. It's the first PCB I lay out (except a couple rather simple hand made ones) so don't expect much. Its probably a non optimal layout, but I may have an idea on where components could be placed.

D3 can be left out for now until the CV loop is tested. Ill test the circuit with a 1A schottky diode here, it's likely to alter the allowed current overshoot.  R26 might have to be changed.
I use machined IC socket pins where I expect to have to later change a component value.

Ok, thank you. I'll prototype the CV loop on a perfboard.

    Juan

I had hoped that the main current paths would have been routed directly across the PCB  with input and output capacitors on board Like I illustrated in post #126. The shunt resistors are ok where they are.
The idea is to keep the main + and - paths close to each other.

[xavier60]

I don't expect the layout to cause problems.
There is one problem with my layout that I'm finding it difficult to find a solution for. There is a bit too much track length between the top side of the shunt and the star point.

[JuanGg]

I had hoped that the main current paths would have been routed directly across the PCB  with input and output capacitors on board Like I illustrated in post #126. The shunt resistors are ok where they are.
The idea is to keep the main + and - paths close to each other.
   

The main filter cap is on a separate board, perhaps it would be a good idea to add more capacitance right on the PCB. EDIT: I see this is on #126.

Output capacitor is better on the PCB or right at the front panel binding posts?

Also, voltage sense divider could be connected to the binding posts to partially compensate for drop in the wires. Is this a good idea?

Also, star grounding as I did, is it fine? Or should I use a groundplane?. I did it single sided, so adding a groundplane is simple enough.

I'll try to make a Rev B with your suggestions.

    Juan

[xavier60]

I had hoped that the main current paths would have been routed directly across the PCB  with input and output capacitors on board Like I illustrated in post #126. The shunt resistors are ok where they are.
The idea is to keep the main + and - paths close to each other.
   

The main filter cap is on a separate board, perhaps it would be a good idea to add more capacitance right on the PCB. EDIT: I see this is on #126.

Output capacitor is better on the PCB or right at the front panel binding posts?

Also, voltage sense divider could be connected to the binding posts to partially compensate for drop in the wires. Is this a good idea?

Also, star grounding as I did, is it fine? Or should I use a groundplane?. I did it single sided, so adding a groundplane is simple enough.

I'll try to make a Rev B with your suggestions.

    Juan

I always put the output capacitor on the PCB and I have never bothered with sensing at the terminals because there will be the resistance of the power leads between the power supply and load anyway.
Theoretically, having the capacitor at the output  terminals should improve the phase margin and stability but I haven't tried it, so I can't be certain.
The star point looks fine. I have no experience with ground planes either because I make my own single sided PCB's.

[JuanGg]

I always put the output capacitor on the PCB and I have never bothered with sensing at the terminals because there will be the resistance of the power leads between the power supply and load anyway.
Theoretically, having the capacitor at the output  terminals should improve the phase margin and stability but I haven't tried it, so I can't be certain.
The star point looks fine. I have no experience with ground planes either because I make my own single sided PCB's.

Alright. I was thinking of making a perfboard prototype and then have a couple PCBs made, as I have two channels. It's tempting to use the other layer, maybe I should route all grounds on the second layer and a couple more things. It'll definitely make the layout easier. No problems to replicate that on perfboard.
    Juan

[JuanGg]

Here is Rev B. There is still some tweaking and double checking to be done. I have routed all grounds on the botom layer. I have used traces as thick and as short as I could and did my best to follow guidelines in #126. I'll add the fan circuitry and some connectors to the schematic so I can see how much room I have left (size of the perfboard I have, 8 x 12 cm, but not much more left inside the case). Also I have attached some CAD screenshots.

    Juan

[xavier60]

Here is Rev B. There is still some tweaking and double checking to be done. I have routed all grounds on the botom layer. I have used traces as thick and as short as I could and did my best to follow guidelines in #126. I'll add the fan circuitry and some connectors to the schematic so I can see how much room I have left (size of the perfboard I have, 8 x 12 cm, but not much more left inside the case). Also I have attached some CAD screenshots.

    Juan

That looks good.
I have been working on a problem that I have just noticed. If the output voltage is set higher than the unregulated rail can supply, the CV op-amps goes into saturation, output pin goes to full 8 volts. This causes C4 to over charge causing the CC loop to lose its quick limiting response to the output being short circuited.
Putting a 1A schottky diode across R10 minimizes the problem.  See if you can add the diode, anode to ground.

Extra: It looks like R13, 10K on the layout.

[JuanGg]

Like that? Or to the output ground?

(My resistor numbers don't mach yours R20 = R10)

[xavier60]

It's the other 10K, between the inverting input and ground.

[JuanGg]

I see, sorry. My mistake

[xavier60]

You should also add the CC LED. It can be very useful.  A CV LED also if there is space.

[xavier60]

As with my layout, there is an undesirable length of track between the shunt and the ground star point. Can you calculate the track resistance?
Try to make the track wider , also the tracks that interconnect the shunt resistors.

[xavier60]

The compromise I made with mine was to put the star point half way along the ground track. I also have 0.8mm solid copper wire soldered along its length. I measure the total resistance at about 1.5mΩ putting about 0.75mΩ in series with the shunt.
My star point is actually a single connection between the  control circuit's ground print and the  ground track.

[JuanGg]

I'll move things around as per your suggestions. Time for next version. Regarding CC and CV LEDs, I could measure this with the Arduino and show it on the displays (maybe on the mA one, as I will only be using 3 out 4 digits). I also have to think about measuring voltage and current as well. The Arduino has a 12 bit 10bit ADC with an internal reference if 1.1 V.   For measuring voltage, a simple resistor divider, and for current, maybe tapping off something at the input of the CC op amp.
I can calibrate errors in software.

    Juan

[xavier60]

I'll move things around as per your suggestions. Time for next version. Regarding CC and CV LEDs, I could measure this with the Arduino and show it on the displays (maybe on the mA one, as I will only be using 3 out 4 digits). I also have to think about measuring voltage and current as well. The Arduino has a 12 bit ADC with an internal reference if 1.1 V.   For measuring voltage, a simple resistor divider, and for current, maybe tapping off something at the input of the CC op amp.
I can calibrate errors in software.

    Juan

With my previous bench supply project, I put CC and CV LEDS on the regulator board as well as having the PIC monitor the CC op-amp.
With its layout, the shunt got positioned off the PCB and allowed the main current paths to run parallel from left to right. It is the floating type design with everything is referenced to the + output.
Amplifying the shunt voltage could be tricky. With my bench supply, I had split control rails. So I was able to use auto-zeroing op-amps, mainly because it's important to have low input offset and drift for the shunt amplifier. It wouldn't have been easy for me to correct errors in the code because I have only 10 bits.
You will need an op-amp that is rail to rail, input and output.
I got some OPA2192's to replace the TLC072 CC/CV op-amp in my bench supply, mainly to reduce offset error also.
The specs say "Rail-to-Rail Input and Output".  Its other specs would be an over kill for your application though, its high speed and high output current.

[JuanGg]

LM358 op amp is rail to rail. I am running out of space on the board if I want to get it manufactured (prices go right up out of the 100 x 100 mm size) I see what I can fit. I'll have to mount the shunt resistors vertically, and the regulators as well. This is where I'm at:

    Juan

[Kleinstein]

The LM358 is not rail to rail, it is an single supply OP only.  For the circuit likely used, this is not a problem, as there is no need to go close to the positive supply. The OPA2192 would be rail to rail - however it is still relatively expensive. The OPA2171 would be a cheaper, though less precise alternative.

For making the  board smaller one could change the non critical resistors and capacitors (e.g. at the regulators) to SMD form factor.  This can safe quite a bit of space. Size 0805 is not that difficult to solder by hand. For manual placement and soldering, there is no big penalty for mixing THT and SMT. If available I even tend to prefer THT chips and SMD passives (expect electrolytic).

[JuanGg]

The LM358 is not rail to rail, it is an single supply OP only.  For the circuit likely used, this is not a problem, as there is no need to go close to the positive supply. The OPA2192 would be rail to rail - however it is still relatively expensive. The OPA2171 would be a cheaper, though less precise alternative.

For making the  board smaller one could change the non critical resistors and capacitors (e.g. at the regulators) to SMD form factor.  This can safe quite a bit of space. Size 0805 is not that difficult to solder by hand. For manual placement and soldering, there is no big penalty for mixing THT and SMT. If available I even tend to prefer THT chips and SMD passives (expect electrolytic).

Right, I got confused with another datasheet (LMV358). LM358 comon mode voltage range does include ground. Pinouts seem standard, so I could change which op-amp I use later on. It seems I can fit everything on the space I have (and THT resistors are handy for jumping traces). The TLC072 op amps only came on smd.
    Juan

[Kleinstein]

For jumping traces SMD resistors in 1206 size also work.  If just for a single private used circuit one can even solder 0805 parts to the slightly larger 1206 taps.  It's extra mechanical stress to the parts, but should be OK for resistors, if no too much solder is used.

The OP pinout is pretty standard with most types.  One trap for the young players here is the LT1013 - in SMD (SO8) it has a different pinout.

[xavier60]

LM358 op amp is rail to rail. I am running out of space on the board if I want to get it manufactured (prices go right up out of the 100 x 100 mm size) I see what I can fit. I'll have to mount the shunt resistors vertically, and the regulators as well. This is where I'm at:

    Juan

The layout is looking more tidy.
Some space can be saved by using narrower tracks around the driver transistors.
I had to use adapters for the OP2192 op-amps, https://i.ebayimg.com/images/g/0XkAAOSwtfhYqoh0/s-l300.jpg
I solder 0.5mm wire  into the holes so I can put them into a standard DIP socket.
I tried an OP2192 in the current power supply project. It worked no differently to the TLC072 as expected.
With both types, I have seen the inputs of the CC op-amp go down to -0.4V with no problem.

[JuanGg]

The layout is looking more tidy.
Some space can be saved by using narrower tracks around the driver transistors.
I had to use adapters for the OP2192 op-amps, https://i.ebayimg.com/images/g/0XkAAOSwtfhYqoh0/s-l300.jpg
I solder 0.5mm wire  into the holes so I can put them into a standard DIP socket.
I tried an OP2192 in the current power supply project. It worked no differently to the TLC072 as expected.
With both types, I have seen the inputs of the CC op-amp go down to -0.4V with no problem.

And the tydiness goes down the drain... I have added the micro, PWM filtering, connectors... The only thing missing is the current measuring. I have half of a LM358 left. I didn't terminate it in case it could be used for that. I have attached my current schematics as well.
I bought those adpters to, I'll maybe change the SOIC outline for a DIP one just in case.
This is taking some effort, and I hope it works out. Do I go ahead and prototype the CV part on perfboard, or will it work for sure on a manufactured PCB?

    Juan

[Kleinstein]

The PWM filter is rather low impedance with 1 K and 1 µF.  Even with the LM358 one can use larger resistors like 10 K or 22 K.  If one has an OP, one could also use a proper 2nd or  3rd order active filter to get less residual ripple.

The CV part is not that complicated. The main uncertainty is in some resistor or capacitor values.  So one could include the CV part on the board. If for some odd reason it would not work as expected one could still keep it unpopulated and add the CV part separate.

[JuanGg]

The PWM filter is rather low impedance with 1 K and 1 µF.  Even with the LM358 one can use larger resistors like 10 K or 22 K.  If one has an OP, one could also use a proper 2nd or  3rd order active filter to get less residual ripple.

The CV part is not that complicated. The main uncertainty is in some resistor or capacitor values.  So one could include the CV part on the board. If for some odd reason it would not work as expected one could still keep it unpopulated and add the CV part separate.

I still have some space left on the board, so I could use a more complex filter. I have to do some reading on that. I just did it that way on my electronic load and it worked fine.

I definitely prefer getting the PCB made and then change a couple values or cut a couple traces than making a full prototype first. I just want to make sure I don't make any stupid mistakes.

    Juan

[xavier60]

The layout is looking more tidy.
Some space can be saved by using narrower tracks around the driver transistors.
I had to use adapters for the OP2192 op-amps, https://i.ebayimg.com/images/g/0XkAAOSwtfhYqoh0/s-l300.jpg
I solder 0.5mm wire  into the holes so I can put them into a standard DIP socket.
I tried an OP2192 in the current power supply project. It worked no differently to the TLC072 as expected.
With both types, I have seen the inputs of the CC op-amp go down to -0.4V with no problem.

And the tydiness goes down the drain... I have added the micro, PWM filtering, connectors... The only thing missing is the current measuring. I have half of a LM358 left. I didn't terminate it in case it could be used for that. I have attached my current schematics as well.
I bought those adpters to, I'll maybe change the SOIC outline for a DIP one just in case.
This is taking some effort, and I hope it works out. Do I go ahead and prototype the CV part on perfboard, or will it work for sure on a manufactured PCB?

    Juan

Although I have done a lot of testing here, I would prefer that you test the CV loop at least  before committing it to a PCB.
I still need to test with a larger group of TIP35C's which I don't expect to see for a week or so.

[JuanGg]

  Although I have done a lot of testing here, I would prefer that you test the CV loop at least  before committing it to a PCB.
I still need to test with a larger group of TIP35C's which I don't expect to see for a week or so.

Thanks again for doing so. I'll do just that, although I won't be able to until next week or so.
    Juan

[JuanGg]

Regarding PCB grounding, should I add a groundplane or keep the star grounding as is? I have been doing some reading on the subject and there are different views.
   Juan

P.d: happy 2019!

[Kleinstein]

For a linear lab supply the star ground (or similar) is usually more suitable than a ground plane - especially a bad ground-plane can be a real pain.

[JuanGg]

With my current layout, digital ground return looks like below. That can't be any good. As I have clearly separated analog and digital sections, should I separate grounds in any way? Also, no current should flow between the output ground and the regulator's ground, should I move the star point?
Or maybe I should have separate digital and analog groundplanes, joined under the traces going from one section to the other?
8 V reg is only for analog, while 5V one is only digital, I should separate those too.

Thank you.
    Juan

[xavier60]

Unless someone has a better idea, I would link AUX reg ground to Digital ground to main star point.
The 12V regulator should be placed on the left so that fan current has the shortest path in and out.
Regulator input bypass capacitors can inject some ripple current into ground rails.
One input bypass capacitor that is carefully placed is sometimes best.
Are you panning to temporarily power the AUX regulators from the main unregulated rail at all?

[JuanGg]

Unless someone has a better idea, I would link AUX reg ground to Digital ground to main star point.
The 12V regulator should be placed on the left so that fan current has the shortest path in and out.
Regulator input bypass capacitors can inject some ripple current into ground rails.
One input bypass capacitor that is carefully placed is sometimes best.
Are you panning to temporarily power the AUX regulators from the main unregulated rail at all?

Ok, so all digital grounds tied together, and then to star point. Each analog ground to star point. Fan connector and transistor is as close as it can be to the 12 V reg. I am not planning on powering the aux regulators from the main unregulated supply, as I have all transformers and rectifiers set up.
I will make a couple changes and include a single input bypass cap. BTW, will the op-amp need a bypass cap?

    Juan

[xavier60]

I have to be certain that we are not misunderstanding each other. Ill see If I can make a sketch.
 I don't bother with bypass capacitors for op-amps because my PCBs are usually small. It's difficult to say at what power track length an op-amp will suddenly need a capacitor.
But when a capacitor is added, a HF loop is formed and needs to be broken with resistor or choke in series with the supply rail.  You can make provision for the parts and fit them if necessary.

[xavier60]

Ideally the the digital to analog ground path should not carry any currents but it's difficult to avoid.
The return current path current for the 8V is low and doesn't fluctuate much at all.
Even a possible HF loop isn't really a big problem because the 100nF capacitor would filter a lot of it out at the CV op-amp.

[JuanGg]

Ideally the the digital to analog ground path should not carry any currents but it's difficult to avoid.
The return current path current for the 8V is low and doesn't fluctuate much at all.
Even a possible HF loop isn't really a big problem because the 100nF capacitor would filter a lot of it out at the CV op-amp.

Ok, I'll do the grounding as suggested. What values should I use for R1 and C1?

I have prototyped the CV loop. After making a couple mistakes I got it working. It regulates from 0 to around 30 V and with no load there is barely any ripple on the output. There is no overshoot at power on. I have only done the analog side, using an LM317 as the 8 V reg and a pot for setting the voltage. As I haven't got the SOIC to DIP adapters yet, I made my own for the op-amp. I also had to parallel and series some resistors to accound for values I haven't got around (some on the way), so resistor values aren't spot on, but within tolerance. I'll attach the power transistors to a heatsink and do further testing.
    Juan

[xavier60]

Ideally the the digital to analog ground path should not carry any currents but it's difficult to avoid.
The return current path current for the 8V is low and doesn't fluctuate much at all.
Even a possible HF loop isn't really a big problem because the 100nF capacitor would filter a lot of it out at the CV op-amp.

Ok, I'll do the grounding as suggested. What values should I use for R1 and C1?

I have prototyped the CV loop. After making a couple mistakes I got it working. It regulates from 0 to around 30 V and with no load there is barely any ripple on the output. There is no overshoot at power on. I have only done the analog side, using an LM317 as the 8 V reg and a pot for setting the voltage. As I haven't got the SOIC to DIP adapters yet, I made my own for the op-amp. I also had to parallel and series some resistors to accound for values I haven't got around (some on the way), so resistor values aren't spot on, but within tolerance. I'll attach the power transistors to a heatsink and do further testing.
    Juan

47Ω and 100nF will be fine.
Good to see it working there. Some tests should be done to make certain that it isn't just working in a marginal way.

The Base of Q2 in the schematic from post #177 should be close to 5V.
A load transient recovery test should give a recovery time of about 15µs with very little overshoot.
What I'm mainly curious about is the ratio of voltage change at the Base of Q2 and output current.
This needs to be done with the load current stepped between 0.1A and 0.5A or higher, at about 50Hz.

[JuanGg]

       47Ω and 100nF will be fine.
Good to see it working there. Some tests should be done to make certain that it isn't just working in a marginal way.

The Base of Q2 in the schematic from post #177 should be close to 5V.
A load transient recovery test should give a recovery time of about 15µs with very little overshoot.
What I'm mainly curious about is the ratio of voltage change at the Base of Q2 and output current.
This needs to be done with the load current stepped between 0.1A and 0.5A or higher, at about 50Hz.   

I'll make those measurements. I have a CC load that I can hook up to my function generator. Testing is to be done at 0.1 A, 50 Hz square wave, then 0.2 A etc right?
What about output voltage, do I test at, say, at 5, 15, 25 V?
    Juan

[xavier60]

       47Ω and 100nF will be fine.
Good to see it working there. Some tests should be done to make certain that it isn't just working in a marginal way.

The Base of Q2 in the schematic from post #177 should be close to 5V.
A load transient recovery test should give a recovery time of about 15µs with very little overshoot.
What I'm mainly curious about is the ratio of voltage change at the Base of Q2 and output current.
This needs to be done with the load current stepped between 0.1A and 0.5A or higher, at about 50Hz.   

I'll make those measurements. I have a CC load that I can hook up to my function generator. Testing is to be done at 0.1 A, 50 Hz square wave, then 0.2 A etc right?
    Juan

The load transient test can be stepped from 0A to 0.5A or up to anything the supply can handle. The duty cycle can be reduced to minimize dissipation.
The other test needs to step from 0.1A because the trans-conductance is non-linear at lower currents. You can even use a ramp current to observe the linearity.  The reason for the 50Hz is so that the capacitor in your DSO doesn't alter the waveforms much if AC coupled.
The voltage doesn't really matter so long as the regulator doesn't drop out of regulation.

[xavier60]

The transfer characteristics starting at the Base of Q2 to the Emitter of the TIP35C is mainly a voltage controlled current source, or transconductance amplifier. A change in voltage at the Base of Q2 causes a change of current at the power supply's output.
The amplification factor or gm is the ratio of voltage change at the Base of Q2 to the current change at the output.
The one main remaining thing that I have not been able to test is how much the gm changes for individual TIP35C transistors.

[JuanGg]

47Ω and 100nF will be fine.
Good to see it working there. Some tests should be done to make certain that it isn't just working in a marginal way.

The Base of Q2 in the schematic from post #177 should be close to 5V.
A load transient recovery test should give a recovery time of about 15µs with very little overshoot.
What I'm mainly curious about is the ratio of voltage change at the Base of Q2 and output current.
This needs to be done with the load current stepped between 0.1A and 0.5A or higher, at about 50Hz.

Attached is a photo of my setup: Function generator is connected to the electronic load, which I can configure to give me a set CC or in relation to the waveform from the function gen. I don't trust this load too much (as I made it). I used the High-res adquisition as the load was injecting a fair ammount of noise. I am monitoring current Q2 base or and output voltage on the scope.
I attached a small heatsink to the TIP35C, and only ran it for a couple seconds each. It got a bit warm (far from too hot to touch) when running it for a minute or so.

All measurements have been done with the power supply set to 20 V.

-Drawing 0.2 A CC or 0.5 A CC, Q2's base remains at 4.54 V.
-Other measurements as attacched.
-Load transients give about 300 uS recovery time (as I see it, see screenshots)
    Juan

[xavier60]

The result of the transconductance test is rather low, looks like a gm of about 7.
The cause of that is actually here. I have just realized that I have been using a Darlington power transistor after all of my fake TIP35C's failed.
Reducing the emitter resistor on Q1 will raise the gm to a more reasonable figure.
This doesn't explain the poor load transient response. The step in the response indicates that the voltage measurement is being taken at some distance from the regulator's output and/or something wrong with the compensating components.
This is my result with a 2N3055 now fitted.


EDIT: change Q1's Emitter resistor to 47Ω

[JuanGg]

The result of the transconductance test is rather low, looks like a gm of about 7.
The cause of that is actually here. I have just realized that I have been using a Darlington power transistor after all of my fake TIP35C's failed.
Reducing the emitter resistor on Q1 will raise the gm to a more reasonable figure.
This doesn't explain the poor load transient response. The step in the response indicates that the voltage measurement is being taken at some distance from the regulator's output and/or something wrong with the compensating components.
This is my result with a 2N3055 now fitted.


EDIT: change Q1's Emitter resistor to 47Ω

I'll change that resistor. The oscilloscope probe was right on the output of the power supply, hooked to the output capacitor. I'll check in case I have forgotten something/used different value capacitors/whatnot.
    Juan

[JuanGg]

Ideally the the digital to analog ground path should not carry any currents but it's difficult to avoid.
The return current path current for the 8V is low and doesn't fluctuate much at all.
Even a possible HF loop isn't really a big problem because the 100nF capacitor would filter a lot of it out at the CV op-amp.

Would a groundplane shaped as attached fulfill the outline in post #250? It just seems cleaner and simpler to me, as I can't possibly take every single ground to the star point, and have to join them before. That would be, of course, keeping the groundplane out of the shunt resistors so main current paths are on the top only.
    Juan

[xavier60]

Ideally the the digital to analog ground path should not carry any currents but it's difficult to avoid.
The return current path current for the 8V is low and doesn't fluctuate much at all.
Even a possible HF loop isn't really a big problem because the 100nF capacitor would filter a lot of it out at the CV op-amp.

Would a groundplane shaped as attached fulfill the outline in post #250? It just seems cleaner and simpler to me, as I can't possibly take every single ground to the star point, and have to join them before. That would be, of course, keeping the groundplane out of the shunt resistors so main current paths are on the top only.
    Juan

Although ground planes allow current paths to share, the cross talk will be minimal because of the ground plane's low impedance.
Just keep a close eye on the fan current path.

[JuanGg]

Fan current path is as short as it can be. I think it'll be fine.
    Juan

[JuanGg]

Here is where I'm at with the PCB. I have included the groundplane, and rearranged things a bit so I don't cut it too much. Also consolidated hole sizes and made higher voltage traces have bigger clearances. The only thing missing is a way to measure current. Maximum drop on the shunt is 0.5 A * 0.025 Ohm = 0.0125 V Would that be too low for an acurate measurement?
EDIT: I bought some INA 196 IC which are high/low side shunt amplifiers. I could use that, and it could maybe added to the CC loop. It has a gain of 20 V/V. They are fairly expensive at $2.45 each in one-off quantity.
I also have one ADC channel left that could be used to detect CV or CC operation.
    Juan

[xavier60]

Here is where I'm at with the PCB. I have included the groundplane, and rearranged things a bit so I don't cut it too much. Also consolidated hole sizes and made higher voltage traces have bigger clearances. The only thing missing is a way to measure current. Maximum drop on the shunt is 0.5 A * 0.025 Ohm = 0.0125 V Would that be too low for an acurate measurement? I also have one ADC channel left that could be used to detect CV or CC operation.
    Juan

You could monitor the CC op-amp's output via a 2:1 resistor divider.
I could be missing something, I have not found a parametric search that helps me find op-amps with wanted combinations of specs.
The only suitable op-amp that I'm aware of is the OPA2192. Actually the same high precision op-amp should be used for the CC op-amp also.
The data PDF for the OPA2192 says "Common-mode (V–) – 0.5 (V+) + 0.5" witch means that it suits the CC job.

 

[xavier60]

Rail to rail auto zeroing op-amps do exist. ISL28134

[xavier60]

The  INA196  has  ±0.5mv typical offset error. That's a ±20ma measurement error. The worst case offset will cause a ±80ma error.
Even with using an op-amp, because we are using single rail, an offset error that wants to cause the output signal to go negative, can't. It is lost signal that can't easily be compensated for.
On the other hand, the OPA2192 has ±5µv offset which gives ±0.2ma error. I think that this might be over kill depending on what resolution is required.
Some time ago I did test the design with a 50mΩ shunt.

ADC inputs usually ignore negative voltage anyways unless small positive offset is intentionally added which still won't help in this case.

[JuanGg]

Ok. What I was aiming for is 0.05 V resolution (can't go any lower with 10 bits ADCs without switching ranges) and 1 mA resolution. I do have 12 big and even 16 bit PWM.
The OPA2192 fortunately has the usual footprint. The thing is that it is fairly expensive (4.5 € each) and it is not available at LCSC, so if I go and buy it at digikey or mouser, shipping would cost a fortune. It isn't that cheap on eBay and who knows if they are fake. I don't mind getting it if absolutely necessary though, although I am in kind of a budget.

Would it be possible to increase the shunt resistor value, so offsets are negligible, maybe switch to high side? Is there any reason to have such small values on shunt resistors for this sort of currents?
I just want to keep the thing simple. Thanks again.
    Juan

[xavier60]

Ok. What I was aiming for is 0.05 V resolution (can't go any lower with 10 bits ADCs without switching ranges) and 1 mA resolution. I do have 12 big and even 16 bit PWM.
The OPA2192 fortunately has the usual footprint. The thing is that it is fairly expensive (4.5 € each) and it is not available at LCSC, so if I go and buy it at digikey or mouser, shipping would cost a fortune. It isn't that cheap on eBay and who knows if they are fake. I don't mind getting it if absolutely necessary though, although I am in kind of a budget.

Would it be possible to increase the shunt resistor value, so offsets are negligible, maybe switch to high side? Is there any reason to have such small values on shunt resistors for this sort of currents?
I just want to keep the thing simple. Thanks again.
    Juan

I could try testing it with 100mΩ but there could be complications. The way the CC op-amp is configured,  it is not a true Miller Integrator. Its minimum gain is one. Increasing the shunt resistor is the same as increasing the  gain.
And the offset error gets improved by  factor of 4.
I bought the OPA2192 from UTsource.
For now, I'm keen on seeing the CV loop working properly.

[xavier60]

It's mostly fine with a 100mΩ shunt,just getting some current undershoot when tested into a dead short, not a practical problem really.
Removing C5, 33pF fixes it.

[JuanGg]

.I could try testing it with 100mΩ but there could be complications. The way the CC op-amp is configured,  it is not a true Miller Integrator. Its minimum gain is one. Increasing the shunt resistor is the same as increasing the  gain.
And the offset error gets improved by  factor of 4.
I bought the OPA2192 from UTsource.
For now, I'm keen on seeing the CV loop working properly. 

.  It's mostly fine with a 100mΩ shunt,just getting some current undershoot when tested into a dead short, not a practical problem really.
Removing C5, 33pF fixes it.

I'll get the op-amp from there, shipping seems cheaper. So, one would be needed for the CC loop and another for current measurement. I could get two dual ones, and use the remaining two for the CV loop and as a buffer for filtered PWM. Or even use the TLC072 for this last two things.

I'll continue testing the CV today. I wasn't able to do so yesterday.

If nothing, by increasing the shunt resistance we are decreasing parts count. However, with the new op-amp it wouldn't be needed right?
Thanks a lot for the effort.
    Juan

[xavier60]

.I could try testing it with 100mΩ but there could be complications. The way the CC op-amp is configured,  it is not a true Miller Integrator. Its minimum gain is one. Increasing the shunt resistor is the same as increasing the  gain.
And the offset error gets improved by  factor of 4.
I bought the OPA2192 from UTsource.
For now, I'm keen on seeing the CV loop working properly. 

.  It's mostly fine with a 100mΩ shunt,just getting some current undershoot when tested into a dead short, not a practical problem really.
Removing C5, 33pF fixes it.

I'll get the op-amp from there, shipping seems cheaper. So, one would be needed for the CC loop and another for current measurement. I could get two dual ones, and use the remaining two for the CV loop and as a buffer for filtered PWM. Or even use the TLC072 for this last two things.

I'll continue testing the CV today. I wasn't able to do so yesterday.

If nothing, by increasing the shunt resistance we are decreasing parts count. However, with the new op-amp it wouldn't be needed right?
Thanks a lot for the effort.
    Juan

It's happy with 250mΩ also. I'm aiming for 4 amps for mine, I keep overlooking  that yours is lower.
So, a lot of options there.  Is there really a need to buffer the references depending on how you are doing the filtering?
You could use the TLC072's and high resistance shunt for now and upgrade to the OPA2192 if it's necessary to go back to the low resistance shunt.

[xavier60]

My earlier 20 amp bench supply has a 10mΩ shunt. Before I replaced the TLC072 CV/CC op-amp with an OP2192, I resorted to trying all of my spare TLC072's in the socket to find the one with the lowest offset.

[Kleinstein]

One usually could compensate for an offset of the OP for the current reading / control in software or if needed in hardware. So I don't think one would really need the rather expensive OPA2192.  If single OPs are used, there would be the option to do offset adjustment with a pot.

Using a 250 mOhms shunt for 4 A would cause quite a lot of heat at the shunt and thus potential drift of shunt value. So it would need a rather high power one.  So the more normal way is to aim for something like 100-200 mV burden voltage, which would be more like 25-50 mOhm for a 4 A range.

[xavier60]

One usually could compensate for an offset of the OP for the current reading / control in software or if needed in hardware. So I don't think one would really need the rather expensive OPA2192.  If single OPs are used, there would be the option to do offset adjustment with a pot.

Using a 250 mOhms shunt for 4 A would cause quite a lot of heat at the shunt and thus potential drift of shunt value. So it would need a rather high power one.  So the more normal way is to aim for something like 100-200 mV burden voltage, which would be more like 25-50 mOhm for a 4 A range.

I think JuanGg is targeting 0.5 amps, so 250mΩ and TLC072's should do fine.
Ill stay with 25mΩ and 4 amps for mine. And because I'm using analog meters, Ill likely use the TLC072 also.

[JuanGg]

Yes, I am after about 0.5 A. I could go a bit higher but not much, given the transformers I have. This PSU would be used for low power stuff. I have a switching PSU that can go up to 5 A  if I need it.

I suppose changing shunt and op-amp would be a matter of resistor values, so the PCB would remain the same. I can go with the TLC072 for now and upgrade later. As I am planning on building two channels, I could even do one of each kind.

I was thinking of using a simple RC low-pass filter.
There is no need to buffer the CV ref, as the op amp has "infinite" input impedance. The CC ref has a rather high (470 k) input impedance, so if using a stiff RC filter shouldn't be a problem. The thing is, if I have one op amp left of the four (CV, CC, Cmeas, remaining one), it could buffer the CC ref instead of being left unused.
How would the current measuring op-amp be connected?
Please correct me if I'm wrong.
    Juan

[xavier60]

Yes, I am after about 0.5 A. I could go a bit higher but not much, given the transformers I have. This PSU would be used for low power stuff. I have a switching PSU that can go up to 5 A  if I need it.

I suppose changing shunt and op-amp would be a matter of resistor values, so the PCB would remain the same. I can go with the TLC072 for now and upgrade later. As I am planning on building two channels, I could even do one of each kind.

I was thinking of using a simple RC low-pass filter.
There is no need to buffer the CV ref, as the op amp has "infinite" input impedance. The CC ref has a rather high (470 k) input impedance, so if using a stiff RC filter shouldn't be a problem. The thing is, if I have one op amp left of the four (CV, CC, Cmeas, remaining one), it could buffer the CC ref instead of being left unused.
How would the current measuring op-amp be connected?
Please correct me if I'm wrong.
    Juan

The left over op-amp is inconvenient. If it's used for the CC buffer, there will be more offset uncertainty. It could be used for the voltage reference to separate 2  RC filters so that the same value R and C can be used for both stages.
The current measurement op-amp would have to be connected as a inverting amplifier because of the negative going shunt voltage.
The CC op-amp already has a 10MΩ resistor connected to the inverting input to make sure it regulates to zero. You will need to add a settable amount of offset to the CC reference in software.
You can do the same sort of thing with the current measurement op-amp so that it has a small positive offset when the shunt current is zero.
Then hope they don't drift much.

Edit: Just realized that the spare op-amp would be buffering the high level CC reference signal and the offset would have little affect. I think.

[JuanGg]

I have changed Q1's emitter resistor from 100 Ohms to 47 Ohms. Atached are the results. Transient response remains the same. I also checked all the conections and component values. They seem fine. Note that I am using BD135 in place of BD137. BD135 has a maximum beta of 250, while BD137 and BD139 have 160 as per the datasheet. That may have an influence.
    Juan

[JuanGg]

The left over op-amp is inconvenient. If it's used for the CC buffer, there will be more offset uncertainty. It could be used for the voltage reference to separate 2  RC filters so that the same value R and C can be used for both stages.
The current measurement op-amp would have to be connected as a inverting amplifier because of the negative going shunt voltage.
The CC op-amp already has a 10MΩ resistor connected to the inverting input to make sure it regulates to zero. You will need to add a settable amount of offset to the CC reference in software.
You can do the same sort of thing with the current measurement op-amp so that it has a small positive offset when the shunt current is zero.
Then hope they don't drift much.

Edit: Just realized that the spare op-amp would be buffering the high level CC reference signal and the offset would have little affect. I think.

I'll do some reading on inverting amplifiers. EDIT: Read the corresponding pages on Learning the Art of Electronics. Seems straightforward enough.
I can easily add offset in software. I could even program a calibration procedure that I can run when I need to, checking voltages and currents with my DMM and have the micro figure out the offsets.

I also have three options for the ADCs, I could use either the 5 V as the reference voltage, enable the 1.1 V internal reference or provide an external reference. 5 V will not be as stable as the 1.1 V reference, and offsets would matter more I suppose.
   Juan

[JuanGg]

Attached is what I understand that is to be used:
I have changed the four parallel shunt resistors from 0.1 to 1 Ohm to get the 250 mOhm shunt. Same footprint, can be changed later on.
I have set the gain at 8.2:  0.25 Ohm * 0.5 A = 0.125 V max drop on the shunt. If we want to amplify it to a maximimum of 1.1 V (using the internal reference of the micro), we need a gain of 1.1 V/0.125 V = 8.8, using E12 values for the resistors, 1 M and 8.2 M leaves a gain of 8.2.
So we would get a max reading with 1.1/8.2 = 0.134 V across the shunt, that is 0.134 V/0.25 Ohm = 536 mA.

Don't know if input impedance would be right.
Please correct me if I'm wrong.
    Juan

[xavier60]

I have changed Q1's emitter resistor from 100 Ohms to 47 Ohms. Atached are the results. Transient response remains the same. I also checked all the conections and component values. They seem fine. Note that I am using BD135 in place of BD137. BD135 has a maximum beta of 250, while BD137 and BD139 have 160 as per the datasheet. That may have an influence.
    Juan

That bit of nonliterary makes it difficult to calculate the gm. Try stepping from a higher initial current like 0.2A to 0.6A.
If the voltage step is still present in the load transient test, there has to be something wrong. while applying the 0A to 0.5A step load, scope the output voltage with only that one probe ground connected. Leave the other probe ground leads disconnected.

[xavier60]

Attached is what I understand that is to be used:
I have changed the four parallel shunt resistors from 0.1 to 1 Ohm to get the 250 mOhm shunt. Same footprint, can be changed later on.
I have set the gain at 8.2:  0.25 Ohm * 0.5 A = 0.125 V max drop on the shunt. If we want to amplify it to a maximimum of 1.1 V (using the internal reference of the micro), we need a gain of 1.1 V/0.125 V = 8.8, using E12 values for the resistors, 1 M and 8.2 M leaves a gain of 8.2.
So we would get a max reading with 1.1/8.2 = 0.134 V across the shunt, that is 0.134 V/0.25 Ohm = 536 mA.

Don't know if input impedance would be right.
Please correct me if I'm wrong.
    Juan

R15 can be very low without causing loading problems. A value of around 1K will be practical.

[xavier60]

Attached is what I understand that is to be used:
I have changed the four parallel shunt resistors from 0.1 to 1 Ohm to get the 250 mOhm shunt. Same footprint, can be changed later on.
I have set the gain at 8.2:  0.25 Ohm * 0.5 A = 0.125 V max drop on the shunt. If we want to amplify it to a maximimum of 1.1 V (using the internal reference of the micro), we need a gain of 1.1 V/0.125 V = 8.8, using E12 values for the resistors, 1 M and 8.2 M leaves a gain of 8.2.
So we would get a max reading with 1.1/8.2 = 0.134 V across the shunt, that is 0.134 V/0.25 Ohm = 536 mA.

Don't know if input impedance would be right.
Please correct me if I'm wrong.
    Juan

A divider from rail can be used to apply a small offset to the non-inverting input

[JuanGg]

R15 can be very low without causing loading problems. A value of around 1K will be practical.
 

I'll make it 1 K and 8.2 K then.

A divider from rail can be used to apply a small offset to the non-inverting input

Alright. I'll add a pot so it can be adjusted.
     Juan

[xavier60]

R15 can be very low without causing loading problems. A value of around 1K will be practical.
 

I'll make it 1 K and 8.2 K then.

A divider from rail can be used to apply a small offset to the non-inverting input

Alright. I'll add a pot so it can be adjusted.
     Juan

You could have a fixed resistor between the non-inverting input and ground and make provision for the other resistor and fit it only if necessary.

[JuanGg]

You could have a fixed resistor between the non-inverting input and ground and make provision for the other resistor and fit it only if necessary.
 

Wouldn't a resistor to ground just keep the non-inverting input at ground, without any offset?
    Juan

[xavier60]

You could have a fixed resistor between the non-inverting input and ground and make provision for the other resistor and fit it only if necessary.
 

Wouldn't a resistor to ground just keep the non-inverting input at ground, without any offset?
    Juan

Yes, there will be a 50% chance that the op-amp's own offset will cause a small positive voltage at the output.
If the op-amp's own offset happens to be reversed, the output will tend to stay at zero volts even when there is some voltage across the shunt. This offset can be as high as 2mV.

[xavier60]

The other optional resistor goes from the non-inverting input to rail. It needs to be a value that applies about 2mV to the non-inverting input if necessary.

[JuanGg]

Got it. Thanks.
    Juan

[JuanGg]

I was making more measurements and double-checking and I realized there was a ground loop on my setup. The electronic load's current measuring output (an RCA connector, outputs a voltage) was connected to the oscilloscope, ground included. This resulted in a drop through the wires and current going through the scope's ground. This explains the voltage step in transient measurements.
I have only connected the center pin, so current measurements are not accurate, but the rest is. That makes all previous measurements useless. Trap for young players 

Attached are the new ones.
Transconductance test, from 0 to about 0.6 A (couldn't do from 0.1A, function generator constraints). Note the 20 mV/div on the blue trace (Q2's base). By the way, how do you calculate gm?

Transient tests turned out way better than before, with 40 uS recovery time. Note the 10mV/div.

     Juan

[xavier60]

Ill have a closer look later, looking better though.
The gm is the ratio of output current to input voltage  using units of amps and volts .
Because of the rather crude design there is expected nonlinearity at low currents which makes the gm difficult to calculate there.

[xavier60]

I should have mentioned that it's the change in current to change in input voltage that's used to calculate gm.
The gm still looks low. It's not a real problem, just curious as to why. The TIP35C's arrived today. Ill  do more testing.
It looks like the transient test is only showing the current off side. The current switching slew rate looks slow.
You might not have any control over that.
I drive a large MOSFET directly with my function generator via a buffer amplifier although this is not totally necessary.

[xavier60]

The ST TIP35C's from Utsource have current gain of all close to 100 @ 1 amp.
My ST BD137's have current gain of about 500. This likely explains why I measure a higher gm.
It is 25 when tested with a 0.1A to 0.6A ramp. I used a resistor to create the 0.1A load.

[xavier60]

To see how the gm affects load transient response, I reduced it to 5 by increasing R4 to 220Ω.
For the 0.5A test, the voltage dip increased from 120mV to 200mV and the recovery time stayed within 20µs.
A gm of over 15 is adequate.
I can tweak the compensation to reduce the recovery response to 50mV and 2µs. I'd rather leave it on the sluggish side to be safe.
My Agilent U8002A dips by 75mv and has 30µs recovery time, but it has a larger output capacitor.

[JuanGg]

 I can try to do transient measurements using a MOSFET and some resistors, instead of using the electronic load, which may be too slow. I can order more transistors and try them out. I'll also try a couple things more and finish off the PCB.
P.D I am back to being a full-time student, so I can't spend much time on this, mainly only on the weekends.
    Juan

[xavier60]

I can try to do transient measurements using a MOSFET and some resistors, instead of using the electronic load, which may be too slow. I can order more transistors and try them out. I'll also try a couple things more and finish off the PCB.
P.D I am back to being a full-time student, so I can't spend much time on this, mainly only on the weekends.
    Juan

Don't bother with getting more transistors. You could just measure the current gain of the ones you have.
You might have mentioned that the CV reference range is 0V to 1.1V? the spare op-amp could be used to amplify it.
The reference range for the CC side can be compensated for by changing one resistor, the 470K,
 

[JuanGg]

Don't bother with getting more transistors. You could just measure the current gain of the ones you have.
You might have mentioned that the CV reference range is 0V to 1.1V? the spare op-amp could be used to amplify it.
The reference range for the CC side can be compensated for by changing one resistor, the 470K, 

Ok I'll measure that too.
PWM goes from 0-5V, so CC and CV references go from 0 to 5 V.
ADC reference is 1.1  V.
    Juan

[JuanGg]

I have measured some of my BD135 with my cheap DMM's transistor tester. They all have a gain of about 170 at room temperature.
I have also done the transient test using a resistor load switched by a MOSFET driven from the function generator. Results attached, current ON and current OFF responses are superimposed, so they fit on the same screenshot. Output voltage was set to 5 V to reduce power disipation on the poor little resistors. Modifying output voltage didn't noticeably change the response. Current is either 0 or 0.5 A, at 50 Hz.

I have also done some experimenting with the current measuring circuit, while it does work, it has some offset that affects low current measurements. I'll do some more testing on that.

I also have to do some work on PWM filtering. PWM frequency will be about 2 kHz and from 0 to 5 V. I have tried a two-stage low-pass filter with no noticeable ripple on the output with about a 500 K load. I'll have to do some work on this.
    Juan

[xavier60]

I have measured some of my BD135 with my cheap DMM's transistor tester. They all have a gain of about 170 at room temperature.
I have also done the transient test using a resistor load switched by a MOSFET driven from the function generator. Results attached, current ON and current OFF responses are superimposed, so they fit on the same screenshot. Output voltage was set to 5 V to reduce power disipation on the poor little resistors. Modifying output voltage didn't noticeably change the response. Current is either 0 or 0.5 A, at 50 Hz.

I have also done some experimenting with the current measuring circuit, while it does work, it has some offset that affects low current measurements. I'll do some more testing on that.

I also have to do some work on PWM filtering. PWM frequency will be about 2 kHz and from 0 to 5 V. I have tried a two-stage low-pass filter with no noticeable ripple on the output with about a 500 K load. I'll have to do some work on this.
    Juan

Those results are much the same as mine and good to see that the design has little overshoot when unloaded.
The offset is expected and can be compensated for so long as it makes the op-amp want to go positive because it can't go negative.
Some op-amps  need a pull down resistor to help the output go down to the correct voltage to keep the loop closed and in control.

[JuanGg]

Those results are much the same as mine and good to see that the design has little overshoot when unloaded.
The offset is expected and can be compensated for so long as it makes the op-amp want to go positive because it can't go negative.
Some op-amps need a pull down resistor to help the output go down to the correct voltage to keep the loop closed and in control. 

I'll try adding offset and putting in a pull-down resistor. I'll also reduce the gain of the current sense amplifier so I have a bit more room to play with.

I was wondering if PWM from 0 to 5 V would be precise enough to give a consistent output, given that the Arduino and displays are powered from the same 5 V rail. I'll try to route the PCB so drop on the traces is minimum.

Also, to check if the supply is in CV or CC mode, a simple voltage divider on the output of the CC op-amp and going to an analog in would do right? EDIT: You proposed that on an earlier post.

    Juan

[JuanGg]

This is how the PCB is doing. I have changed the SOIC footprints to DIP so it's easier to test different op-amps or whatever. I also added the current amplifier right next to the shunt and the voltage divider that monitors the CV CC modes. I have moved things around a bit and added two 5 V current paths, one for a connector to the displays and another one just for the arduino itself. PWM filtering is still pending.
    Juan

[xavier60]

My other bench supply project uses filtered 32Khz PWM from a PIC16F1938 powered from a TL431 controlled regulator.
I didn't worry about load variations because I used an LCD character display.
A simple solution for your project would be to use a TL431 controlled regulator and sense the 5V rail voltage at the VDD pin of the micro-controller.

[xavier60]

This is how the PCB is doing. I have changed the SOIC footprints to DIP so it's easier to test different op-amps or whatever. I also added the current amplifier right next to the shunt and the voltage divider that monitors the CV CC modes. I have moved things around a bit and added two 5 V current paths, one for a connector to the displays and another one just for the arduino itself. PWM filtering is still pending.
    Juan

Yes, that should help. Do you know the display current per segment?

[JuanGg]

Yes, that should help. Do you know the display current per segment? 

I have 2x 4-digit 7 segment displays. Each draws about 35-40 mA with all segments lit at full brightness. So together about 80 mA maximum. That is ≈1.5 mA /segment.

That would be a TL431 just for the micro, to keep input voltage stable, and keep the 7805 for the displays. I have that regulator, I'll do some testing.

I can lower PWM resolution (I don't need 12 bits when I can only measure with 10 bits) and thus increase frequency, which would help. I don't need fast settling time for the filters anyway.

    Juan

[xavier60]

Because the current draw of the MC should be low, a TL431 could be used to shunt regulate 5V from the 8V rail.

[xavier60]

Because I used a MOSFET in my project for the 5V regulator, it has no current limiting.
 I haven't tested the circuit below. Using an LM317 instead of a MOSFET provides current limiting.
The main advantage of this regulator is that the regulated rail can be remote sensed.

[JuanGg]

Now that I think about it, the Arduino itself has an on-board 5 V regulator, which I could feed from either the 12 V or the 8 V rail. 12 V rail is going to be noisy. I don't know if the Arduino is going to upset the 8 V rail.
    Juan

[JuanGg]

I have done some testing with the PWM, I just built up a two pole RC passive filter, with 1k and 10k resistors and 10uF capacitors (just what I had around). PWM is 4 kHz in the end.  Output is not perfectly clean, I'll have to do some measurements. I had the arduino read a potentiometer and generate PWM modifying duty cycle acordingly.

I have measured the 5V derived by the Arduino from the 8V. There is a lot of noise when the Arduino is on (off when pressing reset), see screenshots. I have connected the filtered PWM to the PSU on the perfboard, see some measurements below, that is with the 12V fan on (separate 12V rail, as I measured, better to keep that separated) , PSU and Arduino running from the 8V rail. I have to do more measurements and improve the test setup, which is not the best with wires running all over the place.

    Juan

[xavier60]

High frequency noise on the 5V shouldn't be much of a problem, it would mostly get filtered out by the PWM filter.
When measuring noise, also check how much noise there is when the probe is connected to where the probe ground clip is connected.
The problem with a low value first filter resistor is the high current that can induce spikes across the capacitor's ESR and ground paths.
I used a 5.6K and a 1uF MLCC for the first stage. The second resistor doesn't have to be 10x the value of the first resistor, 5x should be ok, it doesn't have to be a precision filter.
 The second TIP35C order arrived. Transistors in both orders are ST and TO-247. The current gain of the second lot is 60, the first is 100.
I'll be using the 60's for future testing.

[JuanGg]

High frequency noise on the 5V shouldn't be much of a problem, it would mostly get filtered out by the PWM filter.
When measuring noise, also check how much noise there is when the probe is connected to where the probe ground clip is connected.
The problem with a low value first filter resistor is the high current that can induce spikes across the capacitor's ESR and ground paths.
I used a 5.6K and a 1uF MLCC for the first stage. The second resistor doesn't have to be 10x the value of the first resistor, 5x should be ok, it doesn't have to be a precision filter.
 The second TIP35C order arrived. Transistors in both orders are ST and TO-247. The current gain of the second lot is 60, the first is 100.
I'll be using the 60's for future testing.

How do you check for noise where the ground clip is connected? Floating the probe?
I did some measurements with my old setup, and I found noise everywhere, even more when the fan was on. I got sick of the mess of wires and soldered the arduino, voltage regulators and filter on the perfboard. Needless to say, much better results. I also found that the transformer that I'm using drops a bit too much when the fan is on, so the 12V reg cannot regulate any more and the fan gets 10V, still runs fine though. I'll probably put in a 9 or 10 V reg instead.

I'll also connected the displays and the encoder, so I can set the PWM, from 0 to 4096. It seems to work fine now. I am just using a 10k and 10uF RC filter for now.

See an scope screenshot of the output attached. That is with everything connected, fan and displays, and drawing 200 mA with my electronic load. BW limit is not enabled.

I'll do further testing.

    Juan

[xavier60]

Probing the ground clip is just to see how much of the noise is because of ground loops.
Use a faster time base so we can see what the noise origin is.

I know that the regulation loop is very quiet. I see 2mV pp on the output with 1A load.
The noise level stayed the same after I turned off the input supply. So most of the noise I am
seeing is stray pickup.

I have made another PCB for low profile mounting and to test 2 power transistors. All is good so far.

[JuanGg]

I have made a couple more measurements, with a faster timebase. Noise on the ground clip is very similar to the noise on the output, that is with the PSU switched on or completely disconected. It does not change much with output voltage or current. Noise drops a tiny amount when disabling the fan, which was left on together with the displays for these measurements. See screenshots below.
I suppose that with a circuit built on a proper pcb with a groundplane, and inside an earthed enclosure, will perform better.

I also found that when switching on and off the mains switch, i see spikes on the oscilloscope, with the probe unconnected. I assume that is normal.
    Juan

[Wolfgang]

Do you have PLC (power line communications) or LED lights in your vicinity ? That could be the reasons for this type of noise ?

[xavier60]

Sometimes measuring the frequency of the interference can give a clue, sometimes not.
It took me years to realize that the hash that was messing with my DSO's triggering was originating from its own SMPS!

[JuanGg]

I do have a led light bulb less than 1 m away from the setup. I should have tried measuring with it powered off. I'll try just in case.
    Juan

[JuanGg]

I just made a wire loop with one of may probes and moved it around my bench. Switching lights on and off makes no difference. I had everything but the scope unplugged, and about the same amount of noise is present everywhere Bringing the probe closer to the scope's screen increases the noise, as expected. It maybe that. Or even radio signals pickup.
    Juan

[xavier60]

I just made a wire loop with one of may probes and moved it around my bench. Switching lights on and off makes no difference. I had everything but the scope unplugged, and about the same amount of noise is present everywhere Bringing the probe closer to the scope's screen increases the noise, as expected. It maybe that. Or even radio signals pickup.
    Juan

Does the interference signal keep the same characteristics when the loop is near the screen?
The next best thing to do is to take the DSO to another location to see what difference it makes.

[JuanGg]

Attached are some scope screenshots with a 50 Ohm terminator on the BNC no probe, probe pickup on my bench, near the scope screen and in another room downstairs from my lab. There is some difference but not much.

I'll finish off the PCB. I think I'll stick with the simple RC low-pass filters, the seem to work quite well. I'll upload some pdfs/screenshots should anyone want to take a look. Needless to say the files will be made public/open.

Should a reverse-biased diode be added across the main power transistor, for protection?, and also a reverse protection diode on the output?

    Juan

[xavier60]

Looks like there is a chance of damage if another power source is connected to the output while power supply is powered down.
A reverse diode from C to E of the power TIP35C should give some protection.
I think that the most likely accident is a battery being connected with reversed polarity to the power supply's output while powered up.
My bench supply tolerates this, it simply goes into CC mode. I had to add a diode and polyswitch to protect the preload circuit.
I later added a transistor that detects the reverse polarity and clamps the Gate of the MOSFET after a short time. The output capacitor remains unprotected.
I have made provision for this protection on the recent PCB for this project but have not tested anything yet.

[JuanGg]

Wouldn't just a reverse protection diode across the output protect both the circuit and the output capacitors?
I'll add a reverse protection diode and a diode on the power transistor to the circuit.
Any more protection measures I should look at?
    Juan

[Wolfgang]

Juan, are there other switching PSUs around your bench (wall warts, e.g.) ?

[xavier60]

Wouldn't just a reverse protection diode across the output protect both the circuit and the output capacitors?
I'll add a reverse protection diode and a diode on the power transistor to the circuit.
Any more protection measures I should look at?
    Juan

It depends on what you are protecting for. If a high current battery is connected in reverse across the output which is protected by a diode, there is going to be a high fault current. Then you will need a fuse which will add to the power supply's output resistance.
I started doing some tests and got sidetracked by another problem.
I got as far a momentarily applying reverse 9V to it and everything survived. Electrolytic capacitors can tolerate reverse voltage for some seconds before they explode.
Some power supplies have an over voltage protection crowbar circuit on the output. I'm not going to bother with that.
I don't yet fully understand what effect reverse voltage has on the preload circuit.

[xavier60]

This is the schematic of what I'm currently testing, mainly to show how I will be protecting the preload circuit. I have ordered 100mA Polyswitches.
Nothing much has changed except for having 2 power transistors.

Extra: There is supposed to be a 47uF across the 8V rail.

[JuanGg]

Juan, are there other switching PSUs around your bench (wall warts, e.g.) ?

I unplugged all of them and it made no difference. I also moved to another room and noise was a bit lower, but pretty much the same.

[JuanGg]

It depends on what you are protecting for. If a high current battery is connected in reverse across the output which is protected by a diode, there is going to be a high fault current. Then you will need a fuse which will add to the power supply's output resistance. 

I am not planning on using this to charge batteries, and if I do, I can always add a fuse later on. I'll stick with the diode.

This is the schematic of what I'm currently testing, mainly to show how I will be protecting the preload circuit. I have ordered 100mA Polyswitches.
Nothing much has changed except for having 2 power transistors.

That looks neat. Although the main PSU circuitry remains unprotected right?
    Juan

[xavier60]

The main regulator circuitry doesn't care much about momentarily  applied reverse polarity, it simply goes into CC mode. Dissipation will be high.
A transistor can be added to detect reverse polarity and set the output current to zero.

[xavier60]

Just thought of a neat idea for my project.
A diode in series with a piezo buzzer across the output terminals will give immediate indication that reverse polarity has been applied.

[JuanGg]

Just thought of a neat idea for my project.
A diode in series with a piezo buzzer across the output terminals will give immediate indication that reverse polarity has been applied.

I like that idea. I had thought of a led indicator or something, but a buzzer will definitely warn you before something blows up.

[JuanGg]

Here is what I have on the PCB. I still have to make some final touches, but I think this is what I'm going for. I'll wait a couple days before ordering should anyone want to take a look and suggest any changes.
xavier60, I hope you don't mind to be included on the silkscreen, that is the least I can do.
    Juan

[xavier60]

xavier60, I hope you don't mind to be included on the silkscreen, that is the least I can do.
    Juan

I don't mind. 

[xavier60]

Here is what I have on the PCB. I still have to make some final touches, but I think this is what I'm going for. I'll wait a couple days before ordering should anyone want to take a look and suggest any changes.
xavier60, I hope you don't mind to be included on the silkscreen, that is the least I can do.
    Juan

Can you post a schematic mainly of the PWM filters and measurement op-amps areas?

[JuanGg]

Can you post a schematic mainly of the PWM filters and measurement op-amps areas?

Attached are the full schematics. I don't know if I should stick with the RC filters or not.They seemed to do just fine. The spare op-amp was "terminated" as a voltage follower from the ouput of the current sense amplifier.

    Juan

[xavier60]

 Single stage filters aren't really good enough. Not just because of the poor theoretical performance, PWM signal can leak past the capacitor's parasitic resistance and inductance.
A double stage filter can give much better filtering without causing the voltage and current settings to be too sluggish.
 http://sim.okawa-denshi.jp/en/CRCRtool.php
4.7K, 1uF,33K and 100nF give a good result.

EXTRA: or double up the existing parts, 10K, 1uF, 10K and 1uF. The settling time is still short enough.

[JuanGg]

Single stage filters aren't really good enough. Not just because of the poor theoretical performance, PWM signal can leak past the capacitor's parasitic resistance and inductance.
A double stage filter can give much better filtering without causing the voltage and current settings to be too sluggish.
 http://sim.okawa-denshi.jp/en/CRCRtool.php
4.7K, 1uF,33K and 100nF give a good result.

EXTRA: or double up the existing parts, 10K, 1uF, 10K and 1uF. The settling time is still short enough.

I added the second stage to the filters Component values can be changed anytime. I have also consolidated some drill sizes and labeled functional blocks.
    Juan

[xavier60]

Q5 needs C-E swapped.
R32 no longer has a clear benefit.`It was originally meant to break up a loop formed between bypass capacitors on the rail between two separate areas of the board.

[JuanGg]

Alright, thanks for spotting that. I'll make those changes and hopefully order tomorrow or the following day, as most prototype services in China close for the Chinese new year and I want to get this thing made sooner than later. I'll do some more testing on the current sense amplifier if I'm able to.
    Juan

[JuanGg]

Attached is the final PCB and the revised schematics. I hope everything is ok. I'll order this afternoon.
    Juan

[JuanGg]

Just ordered 10 of them (for the same price as 5 ?) and a couple spare components. About 1 € each board including shipping. Really cheap. Let's see how it goes.
    Juan

[xavier60]

We didn't discuss on/off control. Ill be turning off the reference voltage to the Pots.
I guess that you could do the same sort of thing. The filters will make the rise and fall times a little slow,shouldn't be a problem in practice.

[JuanGg]

We didn't discuss on/off control. Ill be turning off the reference voltage to the Pots.
I guess that you could do the same sort of thing. The filers will make the rise and fall times a little slow,shouldn't be a problem in practice.

That is what I had thought, just turning off the PWM. Rise and fall times won't be significant for manual operation I think. I can always add a hard switch that completely disconnects the output if needed.
    Juan

[JuanGg]

I just realized I designed the current sense amplifier to give a maximum of 1.1 V in order to use the 1.1 V Arduino ADC internal reference, and tapped off the voltage feedback divider to measure voltage, which gives a maximum of 5 V  . I'll just increase the gain of the current sense amplifier to give 5 V out at full current, and use the 5 V rail as a reference (perhaps less precise, but I am using it for the PWM anyway) Fortunately, it can be easily fixed and PCB remains the same. I hope I haven't messed up anything else.
    Juan

[JuanGg]

PCBs have got manufactured without any problems and have already been shipped. I will get them in a couple weeks. Meanwhile, I'll work on the software if I have time.
     Juan

[JuanGg]

I have just received the PCBs and some misc components. They look fine as far as I can tell. Attached is a photo of the front, back and one with the big components fitted just for testing. I am not too happy about the PWM filter capacitors covering the Arduino's USB port, and the film ones that I ordered have too big footprints. I'll have to take the arduino out to program it anyways, so the serial comunications don't interfere.
I'll assemble it this weekend and see how it goes.
    Juan

[JuanGg]

R8 and R16 (on my schematics) are 10 M. I don't happen to have resistors that high around. I'll have to order some. I assume they can't be 1 M or so (which is the highest I've got). I'll solder just the CV loop for now.
    Juan

[xavier60]

Don't fit R16, it might not be needed for your op-amp anyway.
Changing R8 to 1MΩ shouldn't cause a noticeable problem. We can check that later.
R17 needs to be recalculated depending on what current range you are aiming for.

[JuanGg]

Ok. I'll do just that. I have soldered the CV loop and it seems to work just fine from the PWM. I have to do further testing. I had to parallel/series a couple resistors in order  to get values I didn't have around. Film capacitors clear the USB port no problem.
Hopefully tomorrow I'll try the CC out.
I am aiming for 0.5 A max output current.
    Juan

[JuanGg]

Sorry for not posting in a while. I've been busy.

I have got the code working to a usable extent, and the CV loop works fine controlled from the micro. I can set the voltage/current with the encoder and show the measured values on the displays

Don't fit R16, it might not be needed for your op-amp anyway.
Changing R8 to 1MΩ shouldn't cause a noticeable problem. We can check that later.
R17 needs to be recalculated depending on what current range you are aiming for.

So do I just put a jumper where R16 was?
How is R17 calculated? I don't really understand how the circuitry around the CC op-amp works. Thank you.
    Juan

[xavier60]

Sorry for not posting in a while. I've been busy.

I have got the code working to a usable extent, and the CV loop works fine controlled from the micro. I can set the voltage/current with the encoder and show the measured values on the displays

Don't fit R16, it might not be needed for your op-amp anyway.
Changing R8 to 1MΩ shouldn't cause a noticeable problem. We can check that later.
R17 needs to be recalculated depending on what current range you are aiming for.

So do I just put a jumper where R16 was?
How is R17 calculated? I don't really understand how the circuitry around the CC op-amp works. Thank you.
    Juan

R16 is to give the CC op-amp some positive offset if necessary. If you don't have the 10MΩ resistor, fit nothing in its position.
R17 and R15 form a divider that applies a positive voltage to the non-inverting input of the CC op-amp. R14 keeps the inverting input at 0V.
The current regulating setting is very close to the voltage applied to the non-inverting input divided by the shunt resistance.
125mV will give close to 500mA.
https://www.eevblog.com/forum/projects/linear-lab-power-supply/?action=dlattach;attach=629695

Edit: corrected "R16" to "R17" in the 2nd line.

[JuanGg]

R16 is to give the CC op-amp some positive offset if necessary. If you don't have the 10MΩ resistor, fit nothing in its position.
R17 and R15 form a divider that applies a positive voltage to the non-inverting input of the CC op-amp. R14 keeps the inverting input at 0V.
The current regulating setting is very close to the voltage applied to the non-inverting input divided by the shunt resistance.
125mV will give close to 500mA.

Allright, thank you.
    Juan

[xavier60]

R16 is needed only if CC mode can't be set all the way down to zero amps.

[JuanGg]

 I have tried the CC loop and it seems to work fine. I calculated R17 to be 400k. As a 470 k was already fitted I tried it temporarily with that. I'll have to do further testing and calibration and maybe add a calibration routine to the code, so I don't have to manually update the hard-coded calibration constants.
This project is taking shape, but lots of things to be done yet, including finishing the case, mounting the PCBs and adding more functionality to the code.
    Juan

[xavier60]

I sometimes solder in IC socket pins to where I expect to have to change component values later.
I get them from chopping up machined pin IC sockets. Or they can be bought in SIL form.

[ArthurDent]

The good SIL socket strips that are .1 inch on center have an insert that effectively grabs the component lead and makes a very good contact. They are made so you can easily snip off the number you need.

[JuanGg]

Sorry for not posting in a while. Exams...

I sometimes solder in IC socket pins to where I expect to have to change component values later.
I get them from chopping up machined pin IC sockets. Or they can be bought in SIL form.

I tried a couple values and ended up using 320 K, mainly because I didn't have an easy (no more than two resistors) way to make 400 K.

The good SIL socket strips that are .1 inch on center have an insert that effectively grabs the component lead and makes a very good contact. They are made so you can easily snip off the number you need.

I have used those before. They are really nice, although a bit more expensive than the standard ones (which I have a lot of) that you have to take one pin off to cut them to length.

I have worked a bit more on the code, and updated the calibration constants so it is reasonably accurate in setting and measuring voltage and current. Everything works as expected and I find the interface easy to use. Constant current mode is indicated by all decimal points showing up on the current display, I saw this on a comercial power supply on this forum somewhere. See first two images.

So here is the problem:
While I was working on the Constant Current mode detection, which is done by a 1/2 voltage divider on the CC op-amp's output, i noticed the decimal points flashing. Attached is a scope screenshot of the output of said divider going to the microcontroller. That is with no load at all and max current set to 100 mA. The "oscillation" goes away when I set the max current to around 300 mA and then starts again going down to 30 mA or so. It seems to work fine drawing some current trough a resistor, but it sometimes oscilates again. I would expect this right on the edge of current limiting, but not drawing 0-30 mA with constant current set to 100 mA.

I went and measured the current sense amplifier's output, same conditions as first measurement, which is the second scope screenshot. Any ideas why this happens?

    Juan.

[Kleinstein]

It looks like there could be a problem with the current control loop under some conditions.
For a first test one could short out the transistor Q5, so that the control would be more conventional for a start.
I know the extra anti wind-up can be a good idea, but it could also cause some problems. So it would be good to know if the regulation work without.  From simulations I know that the anti-windup part can be difficult.

[xavier60]

I'm able to recreate the same symptom here also by changing R8 to 1MΩ.
R8 maintains a consistent voltage on C3 so that there will be a  consistent controlled current overshoot when a sudden overload occurs.
It also causes an unwanted feedback path. This is why I chose the extreme value of 10MΩ.
The anti-windup circuit is stable for me with a 2.7MΩ for R8. Or R8 can be removed for now.

[JuanGg]

I went to the local electronics shop and got the only 10 M resistor they had. I fitted it in and the problem goes away. I'll try with 5 or 3 M for the second channel, which I am still building.
    Juan

[xavier60]

I went to the local electronics shop and got the only 10 M resistor they had. I fitted it in and the problem goes away. I'll try with 5 or 3 M for the second channel, which I am still building.
    Juan

Except for one small problem that I overlooked, I don't expect any show stoppers from here on.
There is some voltage overshoot when transitioning from CC to CV. This is caused by windup of the CV loop while in CC mode.
Because the amount of overshoot is related to the current setting and output capacitor size, the overshoot will be minimal with your implementation of the design.
A resistor in series with a capacitor placed across the top divider resistor prevents the overshoot.

[xavier60]

The addition of D8 prevents windup of the CV loop while the power supply is operating in CC mode.
It limits the charging up of  C3 while the CV op-amp's output is pegged at close to full rail voltage.

[JuanGg]

The addition of D8 prevents windup of the CV loop while the power supply is operating in CC mode.
It limits the charging up of  C3 while the CV op-amp's output is pegged at close to full rail voltage.

I'll add that diode just in case.

I have assembled the second board and got the opto-isolated serial communication working. I also tested and wired the thermistors that measure heatsink temperature and tested the fans. So far everything works as expected. I have 3d printed some PCB mounts and I'll assemble the thing shortly. Then I'll have to calibrate both channels and bend the aluminum sheet top cover.
    Juan

[xavier60]

I have since changed D8 to a silicon signal  diode.
I have been testing here with the regulator now powered from  rectified and filtered transformer secondaries.
There is a small amount output ripple while in  CC mode.
I'm blaming this on the Early Effect that causes BJTs to not act as perfect constant current sources. I other words, current increases with increasing Collector to Emitter  voltage.
Your shunt is ten times higher in value. I'm expecting this to greatly reduce the current ripple.

[xavier60]

I omitted one of the transistors in the output path to reduce the regulators dropout voltage, expecting this to cause problems.
With a few tweaks, it's working well. The dropout is now 1 volt at 5 amps.
The big surprise is the linearity improvement  of the output stage. The waveform is from the Base of Q2 with the output loaded with a 5 amp peak triangle wave. I purposely allowed the dwell at 0 amps.
R14 provides some feed forward which greatly reduces the ripple current in CC mode.

[JuanGg]

I'll test my PSU on CC mode. If it doesn't look too bad, I may leave it as-is.

I have most of the mechanicals done. Those heatsinks were a pain to drill through without a drillpress. I used some thermal paste i got of ebay.
I don't think fans on this thing will ever need to come on.
PCBs are held in place nice and solid via 3d printed brackets.

   Juan

[xavier60]

The recent changes had caused the regulator to produce uncontrolled output whenever the 8V rail was below 3 volts because the output of the TLC072 goes Hi-Z at this low supply voltage. A pull-down resistor on the output of the CV op-amp solves this.

The 2nd schematic is a simplified version. I have omitted the anti-windup circuitry for the CC op-amp.
Sudden and large overload current causes D5 to conduct via the CC op-amp's feedback path. This short circuits the op-amp's negative feedback path and also provides a current path to ground to rapidly discharge the feedback capacitor, C4.
It's not as fast as the original anti-windup circuit, but fast enough at about 5µs.

 Update:

I found a problem with the simple version. If the rise time for overload current is very fast, the non-inverting input of the CC op-amp is driven too far negative causing a messy transition from CV to CC mode.
I could have added another diode.
Instead, I have now connected the op-amp's GND pin to the other side of the shunt.
The extra supply fluctuation to the TLC072 doesn't unsettle it at all, even before adding the RC filtering.
I have updated both schematics.

[xavier60]

With the front panel done,  I can put it to practical use. I still need to build a board for fan and relay control.
When I placed it amongst my other test equipment, I noticed that it causes the image on my DSO's CRT screen to wobble.
I guess that there is too much magnetic leakage from the transformer.  Ill try replacing the transformer with a toroidal type.

Extra: My Agilent U8002A PSU is even closer to the DSO and only causes image wobble when loaded. I figure that there is very little magnetic leakage from its toroidal transformer, only from its 900µH DC choke.

I have chosen this transformer. https://au.mouser.com/datasheet/2/410/media-845752.pdf

Knobs,
https://www.aliexpress.com/item/KAISH-10x-Guitar-Amp-Effect-Pedal-Knobs-1-4-Davies-1900H-Style-Knob-Set-Screw-Various/32829147338.html?spm=a2g0s.9042311.0.0.118f4c4dt6ZdMN

[JuanGg]

With the front panel done,  I can put it to practical use. I still need to build a board for fan and relay control.
When I placed it amongst my other test equipment, I noticed that it causes the image on my DSO's CRT screen to wobble.
I guess that there is too much magnetic leakage from the transformer.  Ill try replacing the transformer with a toroidal type.

Extra: My Agilent U8002A PSU is even closer to the DSO and only causes image wobble when loaded. I figure that there is very little magnetic leakage from its toroidal transformer, only from its 900µH DC choke.

I have chosen this transformer. https://au.mouser.com/datasheet/2/410/media-845752.pdf


Knobs,
https://www.aliexpress.com/item/KAISH-10x-Guitar-Amp-Effect-Pedal-Knobs-1-4-Davies-1900H-Style-Knob-Set-Screw-Various/32829147338.html?spm=a2g0s.9042311.0.0.118f4c4dt6ZdMN

That looks really nice, I am looking forward to see it finished.

Here is how mine is doing:
I have cut and folded the aluminum metal cover, which was relatively easy as it's pretty thin. The software side still needs some work, but I am on it (doing it the analog way has its advantages).
I left it running for an hour or so on a 20 power resistor at 10 V. The heatsink gets to about 25 ºC and then the fan kicks in and cools it down in a matter of seconds. See photos attached.

[xavier60]

Yours is looking neat also. I especially like the plastic work. I have a more tedious method of making a cover. I will be screwing together 4 lengths of extruded angle aluminum.
I have made a PCB for the simple version that mainly omits the anti-windup for the CC op-amp. The combination of the diode from the op-amp's inverting input to ground and moving of the op-amp's GND pin to the negative side of the shunt seems to work well enough for fast current limiting.
I have included the option for a TL431 shunt regulated 5V reference. And also the constant current preload.

The anti-windup idea is still useful for larger designs like my 30V 20A bench supply. Its micro-controller controls the charge on the CC feedback capacitor from a lookup table so that there can be the proper amount of allowed current over shoot for all CC settings.

[JuanGg]

The anti-windup idea is still useful for larger designs like my 30V 20A bench supply. Its micro-controller controls the charge on the CC feedback capacitor from a lookup table so that there can be the proper amount of allowed current over shoot for all CC settings.     

That's interesting.

I have made a front panel overlay, colors should have matched my electronic load (I copied the color code in inkscape) but seems that the copy shop goofed up.
I also added rubber feet and found a place for it in my bench.
    Juan

[xavier60]

Very nice, well done!  What's next I wonder?

The toroidal transformer fitted in well and doesn't cause image wobble on my DSO's screen when positioned close to it.
I have properly fitted the PCB of the simplified version for more practical testing. I have set the compensation to give 3µs load transient response.
Also fitted a noisier fan. https://au.element14.com/sanyo-denki/109r0612g402/axial-fan-60mm-12vdc-27-5cfm-39dba/dp/2768778

[xavier60]

My bench supply project developed a nasty fault. As the voltage setting was increased from minimum, the output voltage would stay at zero then jump to the full regulated voltage of 33V at the 10% setting.
The cause was because of a break in the resistance wire in the 10 turn Pot. The wiper voltage would jump from zero to full reference voltage as it passed over the break.
I don't think that this is common problem.  There is no easy way to protect against it.

[iMo]

I do not know how is your pot wired, but a resistor (large value) from wiper to GND may help. When the wiper is loosing contact the resistor pulls the input of the opamp to GND.
PS: my favorite hint with PSUs

[xavier60]

I do not know how is your pot wired, but a resistor (large value) from wiper to GND may help. When the wiper is loosing contact the resistor pulls the input of the opamp to GND.
PS: my favorite hint with PSUs

I often gave that advise also since I had a batch of 10 turn Pots all develop open wiper contact. There are actually 3 wiping contacts in series in the construction of  helical 10 turn Pots. https://partnership.bourns.com/bu/bu_prec.shtml
A few drops of PAO oil seems to have permanently fixed them.
With the break in the resistance element, the Pot acted more like a switch, selecting either 0V or the full 5V reference for the CV reference input.
I have thought of two ways to protect against this problem that can be used in combination.
One is to monitor current flow through the resistive element.
The other is to monitor for sudden increases in wiper voltage. I'm going with this one for now.

[xavier60]

To the control board I have added a dual comparator which monitors for sudden increase of wiper voltage for the CV and CC Pots via RC differentiators.
The comparators' output pins are wire ORed together and are able to immediately shut down the main regulator's output.
The control board's PIC16F88 micro-controller also gets interrupted and latches off the main regulator.
I have setup the time constant and threshold so that I can purposely cause a false trip by turning the Pot shafts very quickly with knobs removed.
With the knobs fitted, it takes many attempts to cause a false trip.

[panoss]

Just ordered 10 of them (for the same price as 5 ?) and a couple spare components. About 1 € each board including shipping. Really cheap. Let's see how it goes.
    Juan

[sorry for the off topic but] Which is this pcb service that charges so low?
(I suppose it's not Allpcb or JLCPcb because they charge shipping, not for that cheap)

[info]

It could be jlcpcb, because he said each pcb costs $1 and he got 10, which corresponds to the $2 per 10 boards and $8 for the cheapest shipping method. Just saying

[JuanGg]

Sorry for not posting in a while. I've been busy...

Very nice, well done!  What's next I wonder?

The toroidal transformer fitted in well and doesn't cause image wobble on my DSO's screen when positioned close to it.
I have properly fitted the PCB of the simplified version for more practical testing. I have set the compensation to give 3µs load transient response.
Also fitted a noisier fan. https://au.element14.com/sanyo-denki/109r0612g402/axial-fan-60mm-12vdc-27-5cfm-39dba/dp/2768778

Next comes a bit of documentation. I'll tidy things up a bit and make all final files available on an online repository.
That transformer looks supperb. Good to see it doesn't cause any problems.

To the control board I have added a dual comparator which monitors for sudden increase of wiper voltage for the CV and CC Pots via RC differentiators.
The comparators' output pins are wire ORed together and are able to immediately shut down the main regulator's output.
The control board's PIC16F88 micro-controller also gets interrupted and latches off the main regulator.
I have setup the time constant and threshold so that I can purposely cause a false trip by turning the Pot shafts very quickly with knobs removed.
With the knobs fitted, it takes many attempts to cause a false trip.

Interesting problem and solution. Luckily, not having analog pots eliminates that concern in my design .
What I found is that if the op-amps get loose on their sockets, I get 30 V out. The microcontroller notices that and outputs an error message. It only happened once after moving the supply around a bit while making the case, so it shouldn't be a problem.
    Juan

[JuanGg]

It could be jlcpcb, because he said each pcb costs $1 and he got 10, which corresponds to the $2 per 10 boards and $8 for the cheapest shipping method. Just saying

I did precisely that, plus buying a few spare components to get the most out of shipping.
    Juan

[xavier60]

The Omron relay configures the two 15VAC secondaries as parallel or series to suit the output voltage.
The changeover occurs at about 15V.   The changeover threshold can be set to another voltage by holding the power button pressed for a few seconds. The new changeover threshold becomes set to the same as the present output voltage.
A blue LED indicates when the secondaries are in series which reduces the current capability.
The blue LED warns by flashing  if the Current Control is set to higher than what the transformer can safely output.

More: The updated schematic mainly has a CV anti-windup idea which was suggested by Kleinstein. The compensation feedback for the CV op-amp is now taken from the diode OR-ing node.

Minor schematic update 31/05/19

[JuanGg]

The Omron relay configures the two 15VAC secondaries as parallel or series to suit the output voltage setting.
The changeover occurs when the output is adjusted to over 15V. If needed, the changeover threshold can be set to another voltage by holding the power button pressed for a few seconds. The new changeover threshold becomes set to the same as the current output voltage.
A blue LED indicates when the secondaries are in series which reduces the current capability.
The blue LED warns by flashing  if the Current Control is set to higher than what the transformer can safely output.

More: The updated schematic mainly has a CV anti-windup idea which was suggested by Kleinstein. The compensation feedback for the CV op-amp is now taken from the diode ORing node.

Again, that is a really neat power supply!

All my power supply's files: schematics, PCBs, firmware, 3d printed parts, front panel overlay... are available here: https://github.com/Juan-Gg/Linear-lab-bench-power-supply/tree/master

And a short description over here: https://juangg-projects.blogspot.com/2019/05/linear-lab-power-supply.html
    Juan

[xavier60]

I can now confidently say that this topology can be made to work well.

[shoorup4eg]

I can now confidently say that this topology can be made to work well.

Hello
sorry for the necroposting.
I want to make a power supply with current and voltage regulation controlled by a microcontroller. I repeated the circuit in the simulator. This scheme does not work in the simulator. What am I doing wrong? The simulation file is attached (NI Multisim 14).

[xavier60]

Could you post just the schematic?

[shoorup4eg]

Yes of course

[xavier60]

I can't see anything wrong. Ill have a closer look when I get home. There is an updated schematic on another thread Ill link later.

[shoorup4eg]

I can't anything wrong. Ill have a closer look when I get home. There is an updated schematic on another thread Ill link later.

Ok. I'm waiting

[xavier60]

There is mainly a small change in the Emitter circuit of Q1.
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664

[xavier60]

Yes of course

Just to be sure, the opamps  are upside down? I can't clearly see the pin numbers.

[shoorup4eg]

Just to be sure, the opamps  are upside down? I can't clearly see the pin numbers.

everything according to your scheme.
I can't achieve stable circuit operation in the simulator. I'm trying other designs now.