Variable Power Supply

[homebrew]

The zener for the current control op amp is due to the negative rail, otherwise the total voltage would exceed the rated 44V.  I didn't design this one, I'm just trying to fix some of its issues.  I've always been concerned about being so close to the max of the op amps I'm using, especially when it comes to the transient.  I get 43.7V after rectification and filtering and the max input to the TLE2141 is 44V.  Hasn't seemed to damaged anything yet during testing but I always thought that putting all three op amps behind the zener would have been a better idea given the tight output.

Hm, I guess I still don't get it ...
The negative rail should only be two diode drops below zero (approx -1.4V). With that zener, the Voltage of U1 is therefore approx. 8.6V below the supply voltage. But U1 (voltage control opamp) always sees the full supply voltage...

However, why not limiting the supply-voltage for all of the opamps to 30V with a standard tree-terminal regulator? 

[liquibyte]

The zener for the current control op amp is due to the negative rail, otherwise the total voltage would exceed the rated 44V.  I didn't design this one, I'm just trying to fix some of its issues.  I've always been concerned about being so close to the max of the op amps I'm using, especially when it comes to the transient.  I get 43.7V after rectification and filtering and the max input to the TLE2141 is 44V.  Hasn't seemed to damaged anything yet during testing but I always thought that putting all three op amps behind the zener would have been a better idea given the tight output.

Hm, I guess I still don't get it ...
The negative rail should only be two diode drops below zero (approx -1.4V). With that zener, the Voltage of U1 is therefore approx. 8.6V below the supply voltage. But U1 (voltage control opamp) always sees the full supply voltage...

However, why not limiting the supply-voltage for all of the opamps to 30V with a standard tree-terminal regulator?

The original circuit for this is rather old and from a Polish magazine if I'm not mistaken.  In the original the components were underrated by a lot so it was redesigned with modern op amps that could handle the 30V output, at least that's what I was told anyway.  The max rating for a TLE2141 is 44V so if you take the reading I get of 43.7 and add the 1.4 negative rail to U1 you'll be exceeding the op amp limit by 1.1V consistently which will lead to early failure. I've always been a little concerned about sitting so close to that limit myself but it works and I've had the thing running for hours at full load with no side effects.  I think the zener was picked as an arbitrary value to give a voltage drop and anything around 5V could also have been used as well to good effect but as I said, I didn't design this.  There have been a few versions that use a regulator both as an input voltage regulator for the op amps and as a voltage reference.  Regardless of that, the circuit still suffers from the transient and I even have a different version that I was sent for testing.

Since David Hess pointed me to the Q2 circuit in an older Tek supply, I've been looking at a few different designs to see how they dealt with things but I'm still rather new at this so it takes me awhile to grasp some of the concepts.  In the redesign, the Q2 circuit was pulled out but the supply then suffered from a turn off transient so I put it back in and, lo, the transient disappeared.  My final task in that area is to deal with the remaining startup transient.  Along the way I've learned a few things which is good but probably not enough.  This design keeps getting posted in many places so I'd like to make it as stable as possible to point people to when they ask about it.  Fun fact: the guy that did the bulk of the redesign has never built this.  To my knowledge, only hobbyists have built these and I'm starting to see why.  For all its faults though it works rather well to be honest.  I don't have enough knowledge or equipment to fully test this as far as it could be but I'm trying, as are several others.  If I could point to a different design that has been vetted I would.  I can't because nothing has ever really been put forth as a definitive go to circuit so I keep working on this one.  It's fairly simple and has current limiting that works well.  True, you won't get down to 0V without a load on the output but 37mV isn't too bad.  If I had to do this over again I'm sure that I'd still end up in the same place because no one has ever put anything else forward as a solid design.

[David Hess]

The absolute maximum ratings for operational amplifiers and other parts are just that and not the same as the operational ratings; running a part at the absolute maximum rating usually compromises reliability.  Typical operational amplifiers with an absolute maximum rating of 36 volts have specifications for +/-15 volt or 30 volt supplies and if you wanted to use a higher supply voltage like +/- 18 volts or 36 volts, then an operational amplifier with an absolute maximum rating of 44 volts should be used.

[AG6QR]

No, I'm not giving up. But I'm still looking for an approachable Power Supply schematic, the one posted by homebrew is too overwhelming for me. I'm looking for something like the MAKE Powe Supply but safer. 

The Elenco XP-720K is a very simple variable supply offered in kit form.  The instructions contain a schematic.  It has three supply rails.  Two of them are positive and negative variable, using LM317 and LM337 regulators, and the third is fixed 5V, 3A, using LM7805 plus an additional pass transistor.  The design is nothing special, and the positive variable rail (the one using the LM317) is almost straight out of the LM317 data sheet. 

The mechanical construction of this kit is good, with the regulators mounted on a big heat sink that is directly exposed to outside ventilating air.  The case is sturdy metal, and grounded.  The mains input is fused.  Unlike the MAKE supply, there are no obvious safety flaws that I can see.

This supply suffers the problem of only going down to 1.25V, like many LM317-based designs.  Another obvious shortcoming is that it does not have adjustable constant current limiting -- a feature which is extremely useful.

I've built one, and while it's not an innovative design, it is a fairly well-implemented and sturdy version of the standard LM317 type variable supply.  If you're thinking of building a 317-based variable supply of your own, this might be a source of ideas.  If you search around the web, you can see where different people have used this as a base and added modifications, including meters, 10-turn pots, and other enhancements.  The case has a fair amount of room inside for custom additions.

The assembly instructions and schematic are here:

http://www.elenco.com/admin_data/pdffiles/XP720Kweb.pdf

[homebrew]

Regardless of that, the circuit still suffers from the transient and I even have a different version that I was sent for testing.

Maybe a better simulation model might help to debug the circuit. I have attached you a full simulation of your supply in LTSpice. Feel free to mess around :-)

It has quite good regulation performance in voltage mode! However, in current mode it starts to oscillate ...

[liquibyte]

The absolute maximum ratings for operational amplifiers and other parts are just that and not the same as the operational ratings; running a part at the absolute maximum rating usually compromises reliability.  Typical operational amplifiers with an absolute maximum rating of 36 volts have specifications for +/-15 volt or 30 volt supplies and if you wanted to use a higher supply voltage like +/- 18 volts or 36 volts, then an operational amplifier with an absolute maximum rating of 44 volts should be used.

I know but to be honest I think the 2141 is a quality part and the ones I've got in the two supplies I built seem to be chugging along nicely.  Once I figure out how to smooth all this out I plan on fixing that part of it because I'm not comfortable running them at just under the limits.  I do get conflicting advice on that.  Most of the folks over there have said that they've been running these like this for years to no ill effect and just about everyone over here says don't do it.  It may be that a regulator is in order for at least the supply to those.

Maybe a better simulation model might help to debug the circuit. I have attached you a full simulation of your supply in LTSpice. Feel free to mess around :-)

It has quite good regulation performance in voltage mode! However, in current mode it starts to oscillate ...

I've been trying to get that done forever and thought I was just dense.  I really, really appreciate you taking the time to do that and I'll certainly put it to use and see what I can come up with by playing with it.  One question, why did you pulse a mosfet on the output?  So far in all my dealings with LTSpice IV I've only put a load on the output but to be honest I'm not very good with it yet.  I did try an simulate out a version of this but couldn't get it to work so I guess I was doing something wrong.  Good work, this is going to come in rather handy.  Thanks for doing this.

[homebrew]

I've been trying to get that done forever and thought I was just dense.  I really, really appreciate you taking the time to do that and I'll certainly put it to use and see what I can come up with by playing with it.  One question, why did you pulse a mosfet on the output?  So far in all my dealings with LTSpice IV I've only put a load on the output but to be honest I'm not very good with it yet.  I did try an simulate out a version of this but couldn't get it to work so I guess I was doing something wrong.  Good work, this is going to come in rather handy.  Thanks for doing this.

No worries ...
I use the mosfet to switch the output on and off repetitively. The voltage-source at the gate is programmed to do this every 200ms.
You could also use a current source. However you can not test current limiting with such a current source, as it WILL draw its programmed current ...
I also added a minimal load of 1K to discharge the output cap and another 470uF to simulate an average electronic project that might have a bulk capacitor on it. Capacitive loads are often a bit of a thread to regulation as they put another pole into the Bode-diagram and can hence cause instabilities.

One immediate thing I noticed while simulating your supply: The negative rail is very unstable and would really benefit from another 330uF cap after the clamping diodes ...

[liquibyte]

I've been trying to get that done forever and thought I was just dense.  I really, really appreciate you taking the time to do that and I'll certainly put it to use and see what I can come up with by playing with it.  One question, why did you pulse a mosfet on the output?  So far in all my dealings with LTSpice IV I've only put a load on the output but to be honest I'm not very good with it yet.  I did try an simulate out a version of this but couldn't get it to work so I guess I was doing something wrong.  Good work, this is going to come in rather handy.  Thanks for doing this.

No worries ...
I use the mosfet to switch the output on and off repetitively. The voltage-source at the gate is programmed to do this every 200ms.
You could also use a current source. However you can not test current limiting with such a current source, as it WILL draw its programmed current ...
I also added a minimal load of 1K to discharge the output cap and another 470uF to simulate an average electronic project that might have a bulk capacitor on it. Capacitive loads are often a bit of a thread to regulation as they put another pole into the Bode-diagram and can hence cause instabilities.

You've given me a great tutorial with this one.  I've been playing with it and fixed a few connections and a couple of values and have been inputting ESR values and such to see how it affects the simulation and to see if I can make it as accurate as possible based on the datasheets.  I've also tried changing a few things such as just using R=limit(10,V(n002)**2/90,10) for the load and adding a pulse voltage to simulate the spike.  I'm still working on trying for realistic results but what I'm seeing looks promising.  3roomlab's simulations gave me an idea so I changed the value of C6 from 100n to 2.2u to see what it would do and the output ramped up slowly hitting full voltage after the spike passed.  I'm going to play around with this in the real world to see if it has an effect without being deleterious to the rest of the circuit in unexpected ways but I think it could be a quick fix, at least for now.  I'd still like to figure out why it's doing what it is but this is a great start.  I should apologize to the member that I got snippy with about LTSpice because the more I learn, the more I realize how wrong I was about it.  It's still confusing but it's becoming less so.

Edit: I'm told this will screw up the current regulation.  Can anyone tell me why a non-inductive sense resistor would be better than the wirewound aluminum one I have now?  It has been proposed that this could be the cause of the initial spike but I'd like opinions before I spend $7 / resistor as replacements and still have the same issue.  Is there a way to minimize the effects of the inductance of the sense resistor on the circuit?

[David Hess]

have any circuit in real actually use PMOS to do this? many linear supply circuits are always BJT as limiting component for current, is PMOS thermal run-away the main problem that nobody use it?

The common source configuration shown in your schematic using a p-channel MOSFET (or PNP bipolar transistor) has voltage gain which depends on the load so it is more difficult to frequency compensate the error amplifier resulting in poorer transient response.  P-channel MOSFETs are also more expensive than n-channel MOSFETs for a given performance.

Edit: I'm told this will screw up the current regulation.  Can anyone tell me why a non-inductive sense resistor would be better than the wirewound aluminum one I have now?  It has been proposed that this could be the cause of the initial spike but I'd like opinions before I spend $7 / resistor as replacements and still have the same issue.  Is there a way to minimize the effects of the inductance of the sense resistor on the circuit?

If the inductance is high then it will affect the response of the feedback circuit but that does not seem likely with such a low value of resistance.

If you want to test it, short out the sense resistor and test for the startup surge again.

[liquibyte]

If the inductance is high then it will affect the response of the feedback circuit but that does not seem likely with such a low value of resistance.

If you want to test it, short out the sense resistor and test for the startup surge again.

I shorted the resistor and it had no noticeable effect on the transient so I guess that's not where it's coming from.  I've been playing around with the simulation and I notice that the output of the current regulator oscillates like the input does but that the other two op amps don't.  I haven't scoped this in the real world but it's certainly interesting as an experiment.

[kxenos]

The simulation has a difference with the original schematic in cathode of zener D9 the original has a supply there, the sim doesn't. Also, why do you need that 10k in the non-inv input with op-amp's output?

[liquibyte]

Actually, this is what I've been working with.  I've corrected a few issues and rearranged things to look more like the original schematic I used to build these so I could make sure things were hooked up right.  I'm attaching the .asc file as well if someone wants to play with it.

[David Hess]

I've been playing around with the simulation and I notice that the output of the current regulator oscillates like the input does but that the other two op amps don't.  I haven't scoped this in the real world but it's certainly interesting as an experiment.

This behavior is understandable and the reason I said earlier that I am not a fan of the cascading the current limit loop with the voltage limit loop.  If they are separate, then each can be frequency compensated for best performance.  If the current limit loop controls the voltage limit loop, then the added phase shift requires the current limit loop to be slower to prevent oscillation or instability.

[liquibyte]

This behavior is understandable and the reason I said earlier that I am not a fan of the cascading the current limit loop with the voltage limit loop.  If they are separate, then each can be frequency compensated for best performance.  If the current limit loop controls the voltage limit loop, then the added phase shift requires the current limit loop to be slower to prevent oscillation or instability.

After messing with this more, I understand better what you're talking about how it's not an ideal situation.  Since the only real issue that I can see is the transient issue, how would I go about separating the two and still have the same functionality it has now?  Also, does anyone have an idea on how to inject a spike on the input of this?  I've been trying to introduce a delay but when I use the way I came up with in the mosfet thread, the pulse stretches out in time with the delay which is really frustrating.

[kxenos]

Can you please remove C5 to see what happens?

[liquibyte]

Can you please remove C5 to see what happens?

Looks like things get wonky.  Sorry for the big pic but resizing it makes it look crappy.

[TunerSandwich]

"0 – 24 Volt, 2 Amp Bench Top Power Supply"

"Mini Volt Meter (ID: 460 from Adafruit Industries NOTE: These operate from +3.2V to +30V DC and are powered by the supply being measured) (1)"

sounds like a very accurate instrument...who needs buffering and isolation anyway ?

 

[David Hess]

This behavior is understandable and the reason I said earlier that I am not a fan of the cascading the current limit loop with the voltage limit loop.  If they are separate, then each can be frequency compensated for best performance.  If the current limit loop controls the voltage limit loop, then the added phase shift requires the current limit loop to be slower to prevent oscillation or instability.

After messing with this more, I understand better what you're talking about how it's not an ideal situation.  Since the only real issue that I can see is the transient issue, how would I go about separating the two and still have the same functionality it has now?

i did thought about the cascading problem. i dont quite understand how a separation can be done, in my mind, it seems both forms of regulation must happen at the same time and yet are separate entities in circuitry

The usual solution is to wire the outputs of the two error amplifiers together with some diode OR logic; then either amplifier can pull the output down to limit either the current or the voltage.  This causes the amplifier which is not currently controlling the output to saturate but recovery time is generally fast enough for this not to be an issue and the output transitions seamlessly between constant current and constant voltage mode.

[BradC]

(2) Ridiculous case! Wood is definitively not recommended to build enclosures that house components operating on the mains...

Wow, better tell almost every guitarist in the world they are carrying around a dangerous mains powered object. For that matter, most TV's pre late 60's and almost all antique radios and amplifiers.

Wood is perfectly satisfactory. Now the other points I agree with, but this one is lunacy of the highest order.

[homebrew]

(2) Ridiculous case! Wood is definitively not recommended to build enclosures that house components operating on the mains...

Wow, better tell almost every guitarist in the world they are carrying around a dangerous mains powered object. For that matter, most TV's pre late 60's and almost all antique radios and amplifiers.

Wood is perfectly satisfactory. Now the other points I agree with, but this one is lunacy of the highest order.

No, it isn't ...
There is a reason, why these amps are built on a metal chassis. There is also a reason, why domestic installations are not allowed in wood enclosures. And there is also a reason why today's TV's are not housed in wood anymore ...

Of course you can use woodern cases, if the mains components are inside a metal or plastic enclosure within the wood case or at least mounted on a proper chassis.

But hey, you can do whatever you want ... 

[rolycat]

(2) Ridiculous case! Wood is definitively not recommended to build enclosures that house components operating on the mains...

Wow, better tell almost every guitarist in the world they are carrying around a dangerous mains powered object. For that matter, most TV's pre late 60's and almost all antique radios and amplifiers.

Wood is perfectly satisfactory. Now the other points I agree with, but this one is lunacy of the highest order.

No, it isn't ...
There is a reason, why these amps are built on a metal chassis. There is also a reason, why domestic installations are not allowed in wood enclosures.

Then perhaps you should inform companies such as Tivoli Audio who still build domestic radios and hi-fi into wooden enclosures.

And there is also a reason why today's TV's are not housed in wood anymore ...

And that would be because making wooden housings is a labour-intensive, high cost process, and wood is a relatively heavy material unsuitable for large, lightweight flat-panel televisions. It has nothing to do with safety.

[rolycat]

guess what the tivoli EE will say about his PSU?

That it's poorly designed, incompetently constructed, and has pathetically inadequate heat sinking and ventilation?

[liquibyte]

guess what the tivoli EE will say about his PSU?

That it's poorly designed, incompetently constructed, and has pathetically inadequate heat sinking and ventilation?

This!  With a complicated design you're almost guaranteed to make mistakes or odd decisions and then have to work around that to modify things.  With a circuit as simple as this, there's no excuse for building it as sloppily as has been done.  I'm going to guess this guy doesn't do much in the way of electronics though his bio says that he does.

Michael Colombo

In addition to being an online editor for MAKE Magazine, Michael Colombo works in fabrication, electronics, sound design, music production and performance (Yes. All that.) In the past he has also been a childrens' educator and entertainer, and holds a Masters degree from NYU's Interactive Telecommunications Program.

Maybe I should start writing for MAKE, they don't seem to require much in the way of experience or knowledge apparently.

[jlmoon]

(2) Ridiculous case! Wood is definitively not recommended to build enclosures that house components operating on the mains...

Wow, better tell almost every guitarist in the world they are carrying around a dangerous mains powered object. For that matter, most TV's pre late 60's and almost all antique radios and amplifiers.

Wood is perfectly satisfactory. Now the other points I agree with, but this one is lunacy of the highest order.

No, it isn't ...
There is a reason, why these amps are built on a metal chassis. There is also a reason, why domestic installations are not allowed in wood enclosures. And there is also a reason why today's TV's are not housed in wood anymore ...

Of course you can use woodern cases, if the mains components are inside a metal or plastic enclosure within the wood case or at least mounted on a proper chassis.

But hey, you can do whatever you want ...

Last time I checked most appliances today are housed in poly carbonate structures that when heated emit some pretty nasty stuff!  So which is better?.. lol Paper(wood family)  or Plastic ?   I remember the old ("Works in a Drawer" ~ "Quasar"  a division of Motorola) TV's and yes the drawers were made of wood (pressed wood fibers - some call MDF) 

[homebrew]

Then perhaps you should inform companies such as Tivoli Audio who still build domestic radios and hi-fi into wooden enclosures.

Why? I don't have any knowledge how these radios are constructed internally. Maybe the PSU is housed in a metal compartment shielding it from the wooden case. Who knows ...

To put it as simple as possible:
Mains wiring and components have to be either in a mains earth grounded enclosure or they have to be double insulated. (At least in most parts of the western world!)

Wood is not an approved insulator as its properties vary with the sort of wood and moisture etc. etc. and are hard to control.

In the described project, I do not see any form of double insulation within the case.

So yes: For that purpose: Wood is definitively not recommended!

For god sake, you can enclose ANY enclosure (conforming to your local norms) inside a wood box if cooling is sufficient. But to the best of my knowledge, wood should never be used to insulate mains voltages (As it has been in this case!). Maybe however, to block access ...