DIY power supply/function generator

[angelo]

Cool, thanks guys. I'll be sure to document the build and post some pics to pay you guys back for all the great helpI'm considering putting an instructable up.

I'd like to stick to the LM317's or 350's because I have them on hand unless another part saves me alot of work or is substantially better suited. My problem is still how to adjust the current with the potentiometer if it can't handle the high power. is it possible to put a 1W low value resistor in parallel with the potentiometer and just get a higher value pot to deal with the new parallel setup, or is the pot still going to see too much wattage.

given that I'm most likely going to go with an external power-supply such as for a laptop because its safer, very durable, very well regulated, and has overload protected. Plus I only need it for the 4A, 24V input and nothing else. I will maximum draw 20V at 3A if I use the LM350's and 1A with the 317's.

I'm not sure what level of power the pot should be able to handle

[alm]

I think a by-pass transistor messes things up a bit unnecessarily. As mentioned on the drawing I linked to before you can use LM350 instead of the LM317 if you need higher current. It’s also possible to use the LM338 part with can put out 5A and handle 40V at its input.
Obviously if using LM350 (3A) or LM338 (5A) you must change both IC1 and IC2!
If you need more current (over 5A) then you can go for a by-pass transistor.

Sure, there's more than one way to skin a cat. It doesn't make much sense to use external pass transistors with an LM317 unless you need lots of current. I was suggesting the external pass transistor with an LM723, which is a fairly popular solution (although I prefer the two op-amp solution). One thing to keep in mind in the maximum dissipation vs. case temperature graph (you can calculate this from the max junction temperature and junction-case thermal resistance, Tj = Tc +  P * Rth junction-case). Power semiconductors often can only handle the maximum dissipation when the case temperature is below 40 degC or so, which would require almost ideal cooling. I wouldn't count on a TO-220 LM350 to dissipate 90W (3A * 30V) without extreme cooling. The advantage of external pass transistors is that you can easily parallel them to decrease dissipation and lower the temperature.

There’s also the LT1083 adj for 7,5A but it’s more expensive and cannot handle more than 25V on the input.

One thing to keep in mind is that you have to connect the adjust terminal to a potential below ground, so the actual voltage that the LT1083 sees is slightly higher than the voltage between the LT1083 input and ground. Plus the LT1083 is definitely more expensive than other solutions with power transistors like the 2N3055.

[Anders]

Yes I noticed the suggestion regarding the LM723 and I like this IC a lot even though many do not. The LM723 has to be treated with some care to work well; it has a tendency to self oscillate if not properly set up.

The power dissipation is well explained in the datasheets which everybody reads  (I hope) when constructing things. To spread the “heat” you can very nicely parallel some LM317 or any of the other named regulators for higher power if you like.

I’m also for the OP-Amp solution you mentioned but in this case to make a simple power supply the regulator solution makes a lower part count and easier wiring.

For a more advanced psu there are many ways to keep the temperature within limits such as primary triac regulating of the transformer, secondary switching between winding on the transformer or a switch mode pre-regulation.

About your comment on the LT1083, I’m not sure exactly what you mean “below ground”.
However, I agree it’s too expensive! 

[alm]

The power dissipation is well explained in the datasheets which everybody reads  (I hope) when constructing things. To spread the “heat” you can very nicely parallel some LM317 or any of the other named regulators for higher power if you like.

In my experience, it depends a lot on the manufacturer. Datasheets from National, LT and AD are usually very good, but those from ST and TI often contain the bare minimum of both specs and application info. They omit unimportant details like that the regulator will not be stable without a certain minimum ESR on the output. They will state the thermal resistance and max Tj, but it might not be apparent for a beginner how to calculate the Tj from the dissipation and thermal resistance of the system (heat sink + insulation washers + thermal compound). It may not be obvious that even though an LM317 will do 35V between input and output, and will deliver 1.5A, it won't do both at the same time.

About your comment on the LT1083, I’m not sure exactly what you mean “below ground”.

I meant that you can't set (most) three terminal regulators all the way to 0V, since they have an internal reference (often 1.25V). To work around this, you can connect the resistive divider on adjust terminal to a potential below ground, but this will decrease your max. input voltage. The same is true for the LM317, but since its max input voltage is much higher, this is rarely a problem.

[Anders]

Yes, that’s unfortunately the truth.
Some manufactures does not reveal the complete picture and this makes it a good idea always to look for a full app-note if possible no matter which manufacturer your part has.
These app-notes sometimes also explain how to parallel the regulators the way I suggested.

About the LT1083; well this (internal reference voltage) is valid for most regulators, I misinterpreted your statement and thought you meant that the LT1083 was different from the rest. 

[angelo]

hey guys, I've got all the parts set out and I will start taking pictures, but the only thing I'm missing is what wattage pots I need. if I want to have 3A and 20V max what wattage should they be,  is there a way to determine wattage from voltage and current besides multiplying (60W) as I believe I'm missing something.
thanks

[Anders]

It depends on which solution you go for, which circuit you’re planning to build.
In any case for voltage adjustment a standard 0.25W LIN will do fine, for current it’s another story. You can use two potentiometers for the voltage if you like, one low value for fine adjustment. Do you need continuous variable current limit or switched ranges?

[alm]

What's wrong with using a standard pot and a few external components instead of a high-power pot? It's probably easier to get and less expensive.

One thing to keep in mind when using potentiometers for high-current applications in that the maximum power is usually with the pot on its highest resistance setting. If you set it halfway, only half of the surface is used, so the maximum power is approximately halved. Since the current will increase linearly with the inverse of the resistance in this circuit, and the power is increased with the square of the current, it's really easy to burn a pot at high current settings.

I'd change the circuit so the pot is not in the high-current path and use a separate shunt resistor in series with the current. I believe there's an example in the LM317 datasheet (either the National or the Onsemi one).

[kc1980]

Does anybody have a schematic for a DIY switching power supply with voltage and current limiting?  Ideally, I'd like to get about 7A out of it.  Too ambitious maybe? 

[alm]

Does anybody have a schematic for a DIY switching power supply with voltage and current limiting?  Ideally, I'd like to get about 7A out of it.  Too ambitious maybe? 

Quite ambitious, I believe. A switching power supply with a variable voltage output is quite complex, since a regulator usually only has a limited voltage range without changing the inductor. Agilent probably has full schematics in some service manual. Commercial designs are likely to be primary side switchers (so they don't need a heavy, expensive 50/60 Hz transformer), I'd probably give up some efficiency and go for a secondary side switcher, I believe there are some schematics floating around on the internet, for example this one from ELV (german, PDF, with full schematics), and Elektor recently published one (will cost you money, up to 3A, not sure how scalable the design is). Something like this probably wouldn't be suitable as a first SMPS project, I'd start with something simpler.

The specs are also attainable with a linear power supply, just use ten or so power transistors in parallel, and a large heatsink with active cooling to cool them all. Larger and less efficient that a switching design, but much simpler. Basically just a regular linear variable power supply, just with more pass transistors.

[Anders]

The ELV supply is nice but you have to buy the thing (very expensive!) since they probably won’t hand out the source code and a few other things. And it’s not really a full SMPS it’s a secondary switched (as mentioned) which means you still need a large and heavy transformer.
Why not make yourself a primary switch mode 0 – 30V 0 – 12A almost for free? Find an old ATX computer supply and make a few modifications almost without spending anything. Sounds good? well this guy knows all about it: http://www.chirio.com/switching_power_supply_atx.htm

[alm]

The ELV supply is nice but you have to buy the thing (very expensive!) since they probably won’t hand out the source code and a few other things.

I didn't look at it in depth, but I assumed that the micro controller firmware would be easy to replace since the regulation seemed all analog, so the only thing it would do is control a few DAC/PWM signals, and measure some voltages.

And it’s not really a full SMPS it’s a secondary switched (as mentioned) which means you still need a large and heavy transformer.

True. The fact that none of the electronics magazines or websites (that I'm aware of) have published a variable-voltage/current primary switched power supply says something about the complexity.

Why not make yourself a primary switch mode 0 – 30V 0 – 12A almost for free? Find an old ATX computer supply and make a few modifications almost without spending anything. Sounds good? well this guy knows all about it: http://www.chirio.com/switching_power_supply_atx.htm

Not sure how much easier a good modification is compared to building from scratch. You need to confirm that the loop will remain stable under all conditions, since the duty cycle will be outside of the normal specifications, and I wouldn't trust them to have enough margin for a 200%+ increase in output voltage either. He also disabled the over voltage protection. Many ATX power supplies need a minimum output current at several voltages to remain stable, a load resistor would screw up your current regulation (you'd need a constant current sink). I'm not a big fan of reusing ATX power supplies, except when you need a high-current 12V output without much protection, and don't demand a clean signal. If you want to try it, I'd use a high-quality name-brand one, since they tend to be the most stable and are better protection.

I've found the service manual for an Agilent primary switched power supply (3MB PDF). Reading the theory of operation section may give you a rough idea, although the implementation of a high-power switcher is quite hard. Note that circuit layout and magnetics are likely to be very critical.

I'd say that the extra complexity does not outweigh the lower costs for a DIY project. 30V, 7A is well within the range of linear supplies, you just need good cooling. Of course there's nothing wrong with doing it for fun, but I would suggest you to become very familiar with simple buck/boost converters and fixed-voltage primary switchers before you try to make something like this.

[kc1980]

Thanks for all the suggestions!  It sounds way more complicated than I had previously thought.  I've been meaning to get into switching power supply design but it sounds like I'm still too much of a novice.  In the meantime, I'll crack open my copy of Switching Power Supply Design by Pressman to get my feet wet.  I will definitely build a linear power supply very soon.

BTW, I have a old HP 6255A linear power supply that I purchased used on Ebay.  I've owned it for about 7 years, but i have no clue how old it really is.  Do you think I can still trust this power supply or is it likely that the electrolytic capacitors have degraded over time?

thanks again guys!

[Anders]

Kc1980:>
A visual check of the capacitors can sometimes reveal if things are going bad. You might as well have real good look at everything while in there. Look for discoloring, suspect soldering and anything else that doesn’t look right. Another thing you can do is to test the supply with different loads and measure the ripple voltage which should be within specs.
If you have access to cap/ESR meter you can carefully disconnect and measure the caps.

 

[Anders]

The ELV supply is nice but you have to buy the thing (very expensive!) since they probably won’t hand out the source code and a few other things.

I didn't look at it in depth, but I assumed that the micro controller firmware would be easy to replace since the regulation seemed all analog, so the only thing it would do is control a few DAC/PWM signals, and measure some voltages.

This might seem simple enough for a professional programmer but there sure are many traps and pitfalls that will jeopardize the reliability regarding stability and voltage control unless you know exactly what you are doing. This also applies to the electronics of a supply of curse.
I’m not saying it cant be done just that it’s not for everyone to manage, the interaction between electronics and software are crucial.

And it’s not really a full SMPS it’s a secondary switched (as mentioned) which means you still need a large and heavy transformer.

True. The fact that none of the electronics magazines or websites (that I'm aware of) have published a variable-voltage/current primary switched power supply says something about the complexity.

I agree the complexity is high compared to linear supply’s however the principle is not that far from the secondary switchers except maybe for the very dangerous high voltages involved.

Why not make yourself a primary switch mode 0 – 30V 0 – 12A almost for free? Find an old ATX computer supply and make a few modifications almost without spending anything. Sounds good? well this guy knows all about it: http://www.chirio.com/switching_power_supply_atx.htm

Not sure how much easier a good modification is compared to building from scratch. You need to confirm that the loop will remain stable under all conditions, since the duty cycle will be outside of the normal specifications, and I wouldn't trust them to have enough margin for a 200%+ increase in output voltage either. He also disabled the over voltage protection. Many ATX power supplies need a minimum output current at several voltages to remain stable, a load resistor would screw up your current regulation (you'd need a constant current sink). I'm not a big fan of reusing ATX power supplies, except when you need a high-current 12V output without much protection, and don't demand a clean signal. If you want to try it, I'd use a high-quality name-brand one, since they tend to be the most stable and are better protection.

The less it’s modified the better, you will benefit form the original well tested supply and the proposed mod is not very complex at all. As the author points out there are only a few controllers which are suitable for this kind of mod and this tells me that he has been there, tried other controllers and more. I’m sure you can appreciate the way he extended the voltage range with minimum effort and still kept the original parts in the high current section. And of curse it was necessary to disable the over voltage, haven’t seen many of these implemented in variable PSU’s.
I don’t believe you need a minimum load for this to work but if so it could be done without interfering if it’s put before the current shunt.

I've found the service manual for an Agilent primary switched power supply (3MB PDF). Reading the theory of operation section may give you a rough idea, although the implementation of a high-power switcher is quite hard. Note that circuit layout and magnetics are likely to be very critical.

That’s why I think it’s a great idea to use an already made and well tested base like the ATX supply to start on.
The Agilent supply is very advanced and that’s just not a project for homebrew. If you need a supply like this I believe you have to buy it.

I'd say that the extra complexity does not outweigh the lower costs for a DIY project. 30V, 7A is well within the range of linear supplies, you just need good cooling. Of course there's nothing wrong with doing it for fun, but I would suggest you to become very familiar with simple buck/boost converters and fixed-voltage primary switchers before you try to make something like this.

Yes, it’s possible to build analog supply in any range but even here you can make mistakes. One issue in analog technique is self-oscillation which suddenly under some condition can blow things up for you. Also depending of what you are using it for there are hazards involved like ham radio equipment for instance.

[alm]

BTW, I have a old HP 6255A linear power supply that I purchased used on Ebay.  I've owned it for about 7 years, but i have no clue how old it really is.  Do you think I can still trust this power supply or is it likely that the electrolytic capacitors have degraded over time?

I'd test the ESR (if you have an ESR meter or can improvise one), or just measure the ripple. It's quite likely that it still works fine, or will work fine after replacing a few capacitors.

The less it’s modified the better, you will benefit form the original well tested supply and the proposed mod is not very complex at all. As the author points out there are only a few controllers which are suitable for this kind of mod and this tells me that he has been there, tried other controllers and more. I’m sure you can appreciate the way he extended the voltage range with minimum effort and still kept the original parts in the high current section. And of curse it was necessary to disable the over voltage, haven’t seen many of these implemented in variable PSU’s.
I don’t believe you need a minimum load for this to work but if so it could be done without interfering if it’s put before the current shunt.

The issue that I have is that certain choices like magnetics are quite critical for the stability of a switcher, and they where chosen for 12V 1-20A and 5V 0.5-10A operation (or whatever the specs are), plus they should be as cheap as possible. If you're going to change it to work under significantly different conditions, it's possible that it won't be stable. Some ATX supplies are barely stable in their original configuration. Without knowing the margins of the original design, this is hard to confirm. It may be stable with well-behaved loads, but will it stable under all conditions (as a lab supply should)? The energy levels involved are high enough to be dangerous or fry something. Is the author an experienced SMPS designer, or just someone who studied the datasheet, hooked up his multimeter, and found out that the voltage increased? It's clear that he did spend a fair amount of time on this, but does he know what he's doing? Even if he does, you'll have to confirm that your supply has the same amount of headroom as his supplies, what if they skimped on the magnetics or capacitors?

That’s why I think it’s a great idea to use an already made and well tested base like the ATX supply to start on.
The Agilent supply is very advanced and that’s just not a project for homebrew. If you need a supply like this I believe you have to buy it.

That's what it takes to do it right (except GPIB control, you can obviously skip that), and why I consider it hard. This supply is sure to be very stable, and will be protected against everything you can throw at it (as a lab supply should). And yes, it does have over voltage protection .

Yes, it’s possible to build analog supply in any range but even here you can make mistakes. One issue in analog technique is self-oscillation which suddenly under some condition can blow things up for you. Also depending of what you are using it for there are hazards involved like ham radio equipment for instance.

Sure, plenty that can go wrong with analog circuits (do you consider a SMPS digital?). But the components involved with a linear power supply are usually very slow (eg. LM358, often bandwidth limited, 2N3055), so that makes them less likely to oscillate. The frequencies and current changes (dI/dT) of a SMPS are much higher, which makes things more critical and harder to get stable.

[kc1980]

Great info!  Thanks guys.  So, I've decided just to build a linear power supply.  There are times when my 1.8A bench supply just doesn't cut it.  Can you suggest any good schematics for voltage and current regulated linear power supplies?  I found this one -- any thoughts on it?

http://www3.telus.net/chemelec/Projects/MC1466/MC1466.htm

Also, I may incorporate an old variac that I have laying around to improve efficiency.

Thanks again guys.

[kc1980]

Who here is capable of analyzing circuits like this?  I'm reluctant to build something that I don't understand so I've been staring at this schematic off and on for a few hours.  I'm not necessarily asking for someone to explain this to me -- I'm just wondering if this type of circuit analysis should be trivial for a seasoned professional EE. 

http://www3.telus.net/chemelec/Projects/MC1466/MC1466-1.png

[Anders]

The circuit you refer to is partially (almost all of it) a replacement for the obsolete MC1466 IC as the author explains. It might look complex but it’s not too bad to assembly if you make a PCB for it.
Perhaps if you study the datasheets for the original circuit (MC1466) you can grasp its internals better. There are even some example circuits which might shed some light on how the IC operates.

http://www.datasheetarchive.com/pdf-datasheets/Datasheets-21/DSA-409990.pdf

http://www.datasheetarchive.com/pdf-datasheets/Databooks-5/Document249256.pdf

http://www.datasheetarchive.com/pdf-datasheets/Databooks-2/Book278-1125.pdf

[kc1980]

Thanks Anders!  I just wanted to understand it better.  My original approach was to slap it all into LTSpice and see what each functional section does.

[saturation]

This was how things used to be in the absence of good ICs.  Darlington, Sziklai transistor pairs powered by a full and half wave bridge rectified circuit are good for lessons in schools.  There are psuedo totem pole transistor setups to source current, but most transistors are wired in parallel.   There are a lot of diodes used for fixed voltage drops to bias transistors.

Its obsolete today, you could still build it, but the parts count would cost more than using any established DC regulating IC. 

Even simple circuits like an oscillator using a LM555 with a low parts count instead of a PIC12F675 isn't cost effective, the PIC albeit more involved to get started with and tools to program, costs about the same per chip as the LM555.  However, once you learn how to use PICs the overall versatility gives more flexibility in total design than working with discrete chips.

Should you still build it to own and use? you can if you have the parts free lying about so no expense there.

Who here is capable of analyzing circuits like this?  I'm reluctant to build something that I don't understand so I've been staring at this schematic off and on for a few hours.  I'm not necessarily asking for someone to explain this to me -- I'm just wondering if this type of circuit analysis should be trivial for a seasoned professional EE. 

http://www3.telus.net/chemelec/Projects/MC1466/MC1466-1.png

[kc1980]

I would love to get my hands on something like the MC1466L.  It's described as a precision wide-range voltage and current regulator.  With external components, the current and voltage limits can be set and the whole thing functions like a lab power supply.

http://pdf1.alldatasheet.com/datasheet-pdf/view/87107/MOTOROLA/MC1466L.html

Are there any modern ICs that do the same thing?  I'm aware of linear regulators (e.g. LM317) that can be configured to for constant voltage or constant current, but can it be configured to do both (i.e. limit the voltage and current as a lab power supply would do)?

[alm]

The first two pages of this thread were about that topic .

In short:

• LM723: All-in-one, only requires some passive components and an external power transistor. Good as a voltage source, not a great current source (but adequate). Quite a lot of cheap commercial lab supplies use this IC, and it's probably the closest to what you asked (not sure how it compares to that obsolete part).
• L200: Voltage regulator that will do current limiting, but not usually with a variable current (useful for eg. a fixed 5V source).
• Two LM317's, one as voltage, one as current regulator.
• 'Discrete' solution with one (or more) power transistor(s), two op-amps, and a bunch of passives, diodes and small transistors. The most complex, and the best performance (many high-quality lab power supplies use this solution in some form or shape).

[Anders]

The LM723 has some limitations compared to the more accurate MC1466L, both the current and the voltage (max/min) ranges are limited. The almost equivalent L146 was (now unfortunately also obsolete) better in terms of the maximum input voltage (80 V instead of only 40 V for the 723 part). The MC1466L goes all the way down to zero volts with no hassle, where the 723 requires extra effort. The current regulation is also better and simpler to use with the MC1466L.
Still; the winner is the 723 but only because it’s still easily available and very cheap, just don’t forget to take measures to prevent it from self-oscillating!   

The L200 is very rugged and close to indestructible, adding a single op-amp it will also perform continuous variable current limit. The L200 also takes some effort if down to zero volts are required.

A discrete solution opens for more possibilities and higher accuracy but also adds to the parts count and complexity.

[kc1980]

alm,anders -
Thanks for your replies!  Just to clarify, my original question was regarding a MC1466L replacement that requires very little external components.  It would be fun to spend time learning about designing with discrete BJTs, but I think that I need to focus more on subject matter that is not so obsolete in today's world.  After reading all the input and researching online, I really think an op-amp-based design will work the best for me. 

Regarding the first two pages of this thread, I don't like the idea of passing too much current through pots -- especially because I'm looking to output up to 7A.  I like the idea of using some level of switching to improve efficiency at the main transformer, but that could get way over my head very quickly.  Instead, I'll probably use a variac transformer and manually adjust the secondary voltage to a reasonable level.

I will post my schematic as soon as I finish.  It is basically a tweaked version of the German ELV design.