Lab power supply design opinions

[starplayer]

Hey guys.

I'm looking to build my first lab power supply, and I ran across the one in this thread http://www.electronics-lab.com/forum/index.php?topic=7317.0. Hope there is no problem linking to other forums.

However, the original version had lots of problems, so the guys there modified it heavily, and after more than 100 pages of thread, they came up with this final version, which can be seen on page 129.






It's supposed to have variable voltage from 0 to 30 v, and limited current from 0-3 Amps.

What do you experts think about this design? Do you think it would make a sturdy lab power supply?

Thanks a lot for your expertise.

The BOM is attached.

[T3sl4co1l]

That looks a lot like another thread on this forum, which also has a lot of problems...

Personally, I'd just redo it from the ground up...

Tim

[Alex Eisenhut]

Why would you build one?

http://www.ebay.com/itm/HP-6284A-Power-Supply-0-20V-0-3A-/231362951150?pt=LH_DefaultDomain_0&hash=item35de4e77ee

Fix this one and you'll get all your jollies that way and still end up with a case and some knobs and stuff.

[Thor-Arne]

However, the original version had lots of problems, so the guys there modified it heavily, and after more than 100 pages of thread, they came up with this final version, which can be seen on page 129.

There is another thread continuing that design here.

[void_error]

That looks a lot like another thread on this forum, which also has a lot of problems...

I assume that would be mine...

[starplayer]

Thank you for your answers.

I'd love to do a power supply from the ground up, but if I did, it'd probably be much worse than that, considering that was revised several times with ideas from several people who know more than me.

As for buying one, here in Europe it's not that easy to buy one second hand, plus, I already have many of the parts needed.

I'd already seen that design continuation, but I though it was weird that he replaced a few complicated parts of the circuit with much simpler components, I mean, either the original designers don't have any idea or it's him who doesn't have any idea, but I could be completely wrong.

Here's his schematic, is it better? I mean, he replaces a an opamp reference voltage circuit with just two zeners, even though they have inverse temperature coeficients. Is it that easy?



Thank you for your expertise.

[Jay_Diddy_B]

Hi starplayer and the group,

The design does not look too bad. I have the following comments:

1) Using two Zener diodes in series with opposite temperature coefficients will help. You could always replace this arrangement with a temperature compensated Zener. Examples would be from the 1N821 thru 1N829 family. You could also use a shunt reference IC.

2) The dissipation in the main pass transistor is very high. If the power supply is doing 0V (shorted) the dissipation would be the set current multiplied by the input voltage (40-45V). Modern designs use two or three taps on the transformer and use relays to the select different input voltages based on the output voltage. This is to minimise dissipation in the output transistor.

3) I have not examine the circuit in detail, but I would build a model to check for:

a) over-shoot on the output voltage at turn-on and turn-off.
b) I would look at the smoothness of the transient between CC and CV operation.
c) The transient response to step load current changes.

Regards,

Jay_Diddy_B

[David Hess]

3) I have not examine the circuit in detail, but I would build a model to check for:

a) over-shoot on the output voltage at turn-on and turn-off.

This was a big problem with this design that we discussed on EEVBlog not long ago without successful resolution.

It's supposed to have variable voltage from 0 to 30 v, and limited current from 0-3 Amps.

What do you experts think about this design? Do you think it would make a sturdy lab power supply?

The output voltage and current objectives are ambitious for the reason Jay_Diddy_B identifies.  I would add thermal protection and/or fold-back current limiting unless the output stage was derated enough to handle the worst case conditions.

[Alex Eisenhut]

I may be working from obsolete principles, but isn't C1 underrated voltage-wise?

[Andreas]

It's supposed to have variable voltage from 0 to 30 v, and limited current from 0-3 Amps.

What do you experts think about this design?

You make the same error that I did with my own power supply:
You can cook water on the transformer after some minutes with full power.

You cannot get  30V/3 Amps out of a 28V/120VA transformer.
Its only around 2 A because of the non sinuoidal current with capacitive load and bridge rectifier.

The factor for current needed is around 1.8 at minimum -> 30V/5.6A = 180VA would fit better.

With best regards

Andreas

[dannyf]

Here's his schematic, is it better?

It is a little bit too complicated - I think you can get it done with 2 - 3 opamps, depending on your design.

Similar designs have been discussed here.

The biggest challenge, I think, is the builder's expectation - those are truly DC power supply and they coupe poorly with a dynamic load.

[David Hess]

I may be working from obsolete principles, but isn't C1 underrated voltage-wise?

30 volts AC will produce about 40 volts DC after rectifier losses.  Voltage derating is normally less for bulk input capacitors.

[Alex Eisenhut]

I may be working from obsolete principles, but isn't C1 underrated voltage-wise?

30 volts AC will produce about 40 volts DC after rectifier losses.  Voltage derating is normally less for bulk input capacitors.

Fair enough. How about ripple current and temperature? If the transformer is undersized and gets very hot...

[David Hess]

I may be working from obsolete principles, but isn't C1 underrated voltage-wise?

30 volts AC will produce about 40 volts DC after rectifier losses.  Voltage derating is normally less for bulk input capacitors.

Fair enough. How about ripple current and temperature? If the transformer is undersized and gets very hot...

The transformer is a different matter.  As far as the input capacitors on 50/60 Hz AC power supplies, usually if you have enough input capacitance to support the ripple level at the maximum output current, their ripple current rating is more than enough.

At 60 Hz with a full wave rectifier, the rule I use is 8333 microfarad volts per amp.  So the input capacitance is proportional to the maximum output current and inversely proportional to the allowed input ripple voltage.

[starplayer]

Thank you very much for your answers and tips.

Given what was said, I believe I will derrate the amps to 2. Probably won't need more anyway.

3) I have not examine the circuit in detail, but I would build a model to check for:

a) over-shoot on the output voltage at turn-on and turn-off.

This was a big problem with this design that we discussed on EEVBlog not long ago without successful resolution.

Isn't the big overshoot a big problem? I mean, it could fry some ICs if we're not careful. Do you have an idea how to get to that discussion?

Do you know any lab PSU schematics that don't suffer these problems so bad?

By the way, a vga cooler that is supposed to handle 140 Watts tdp should suffice for this, correct?

Oh, and why doesn't C7 (at the output) correct the overshoot? Even if it needed to be bigger, wouldn't it absorb the voltage spike?
Thank you for your time

[liquibyte]

Here we go again.  I've got two of these working and they work ok for the most part.  I do experience an overshoot at turn on and have been working on a way to mitigate it but haven't had much time lately due to other obligations so I haven't done real world testing of the mosfet circuit that I came up with.  I have the parts but just haven't gotten around to it yet.

I really, really wish that someone with more test equipment and the knowledge to use it would build one of these and put it through its paces because the few people that are actively working on this are hobbyists with minimal knowledge and equipment.  This design pops up far too often with the same question as to whether or not it's fit for a hobby supply build.  If you ask me, I'd say yes with the caveat that the power transient is enough of an issue that it needs fixed but other than that, it performs well enough.

One of the things I will say negative is the cost.  If you already have most of the components you need to build this, go ahead.  If you're going to buy everything from scratch and are on a tight budget, save your money and get something that is recommended for use here pre-built.  I've sunk several hundred into this project and am still not satisfied but have learned quite a bit for that investment and it's not like I can't reuse everything again later if I want to.

The transformer I'm using is a VPS28-4600 at 130VA and the caps are 15000uF.  My transformers stay at an acceptable temperature at full load for what it's worth.  I don't really have the experience to fully put the thing through its paces though so take what I say with a grain of salt.  I'll send anyone here a board to test out if they'd like to volunteer but I only have two left so supplies are limited.   The design is modular enough that most of the most of the more expensive components are off board and can be recycled easily enough.

First big thread > https://www.eevblog.com/forum/beginners/power-supply-design-%282%29/
Second thread with the mosfet circuit > https://www.eevblog.com/forum/beginners/will-this-mosfet-circuit-work/

[Alex Eisenhut]

The transformer is a different matter.

I was thinking more along the lines of the cap being near the transformer in the enclosure and getting hot that way.

[liquibyte]

The transformer is a different matter.

I was thinking more along the lines of the cap being near the transformer in the enclosure and getting hot that way.

I took that into consideration myself, see attached.

[Alex Eisenhut]

The transformer is a different matter.

I was thinking more along the lines of the cap being near the transformer in the enclosure and getting hot that way.

I took that into consideration myself, see attached.

Looks awesome.

[liquibyte]

It's changed a bit since I took those but the transformers and filter caps are still in the same place.  One thing I did was to add active cooling to the aluminum power resistors because they do get hot.  I'm still not done with the periphery circuitry but I wanted to wait until I had the transients taken care of.  I've got an LM35 on the underside of an aluminum plate that the resistors are bolted to that controls the fan speed based on the temperature there so as to keep quiet when things aren't heating up.  I'm also designing LED displays for voltage and current for each side but am not very far along in that due to the other issues.

[David Hess]

The transformer is a different matter.

I was thinking more along the lines of the cap being near the transformer in the enclosure and getting hot that way.

Exposing the input capacitors to high temperature from the transformer will lower their life but I have never seen it become a significant problem unless the transformer was overheating to start with.

[David Hess]

3) I have not examine the circuit in detail, but I would build a model to check for:

a) over-shoot on the output voltage at turn-on and turn-off.

This was a big problem with this design that we discussed on EEVBlog not long ago without successful resolution.

Isn't the big overshoot a big problem? I mean, it could fry some ICs if we're not careful. Do you have an idea how to get to that discussion?

I consider it a fatal problem in the design and would never put up with it.

The earlier discussion is here:

https://www.eevblog.com/forum/beginners/power-supply-design-%282%29/

Do you know any lab PSU schematics that don't suffer these problems so bad?

We discussed the Tektronix PS501 and PS503 designs and maybe some of the HP designs but they are more complicated or at least not as easy to understand.  I think part of the problem in finding a schematic for a good design is that power supply designs like these are considered a solved problem.

By the way, a vga cooler that is supposed to handle 140 Watts tdp should suffice for this, correct?

You still need to derate the transistor for its expected operating temperature.  Multiple transistors may be used in parallel simply to distribute the heat making it easier to handle.  A VGA or CPU cooler is optimized for cooling a single device.

Oh, and why doesn't C7 (at the output) correct the overshoot? Even if it needed to be bigger, wouldn't it absorb the voltage spike?

That was one of the mysteries.  Even with the output capacitor, the voltage spike was present.  To me it seemed like a control circuit problem but we never found a cause.

[liquibyte]

That was one of the mysteries.  Even with the output capacitor, the voltage spike was present.  To me it seemed like a control circuit problem but we never found a cause.

The thinking is that loss of control at the negative supply is causing the issue.  An idea for eliminating the negative supply altogether has been proposed but it's going to be awhile before I can test out the theory.  Someone else may get there before I can but the thread is over here if anyone wants to talk about it with the main actors of the redesign.  One of the concerns is getting things down to zero volts without the negative rail.

[David Hess]

That was one of the mysteries.  Even with the output capacitor, the voltage spike was present.  To me it seemed like a control circuit problem but we never found a cause.

The thinking is that loss of control at the negative supply is causing the issue.  An idea for eliminating the negative supply altogether has been proposed but it's going to be awhile before I can test out the theory.  Someone else may get there before I can but the thread is over here if anyone wants to talk about it with the main actors of the redesign.  One of the concerns is getting things down to zero volts without the negative rail.

As you pointed out, all it would take is someone with the right equipment to track down the exact cause.  A digital storage oscilloscope, even a low performance one, would be ideal.  In the past before I had a DSO or analog storage oscilloscope, I either used a voltmeter which supported a peak and hold function or built a peak and hold circuit to probe various nodes.

I do not think there is any significant problem with designing a single supply version but it tends to restrict the types of operational amplifiers which can be used and makes the circuit more complicated.

One thing I would avoid which this design does is having one control loop drive the other.

[liquibyte]

That was one of the mysteries.  Even with the output capacitor, the voltage spike was present.  To me it seemed like a control circuit problem but we never found a cause.

The thinking is that loss of control at the negative supply is causing the issue.  An idea for eliminating the negative supply altogether has been proposed but it's going to be awhile before I can test out the theory.  Someone else may get there before I can but the thread is over here if anyone wants to talk about it with the main actors of the redesign.  One of the concerns is getting things down to zero volts without the negative rail.

As you pointed out, all it would take is someone with the right equipment to track down the exact cause.  A digital storage oscilloscope, even a low performance one, would be ideal.  In the past before I had a DSO or analog storage oscilloscope, I either used a voltmeter which supported a peak and hold function or built a peak and hold circuit to probe various nodes.

I do not think there is any significant problem with designing a single supply version but it tends to restrict the types of operational amplifiers which can be used and makes the circuit more complicated.

One thing I would avoid which this design does is having one control loop drive the other.

I will probably end up looking into the proposed solution eventually and I know your last point has merit just by studying other proven designs.  Nothing I've encountered to date has a setup quite like this one but, for its faults, it does seem to work really well.  We just have to get past this one issue and then I'd say it would be a good first supply project to learn from.  There are good people working on a solution but given the limited knowledge of some of us (me) I'm not sure how long it's going to take.  I've hit a snag in my finances so I won't be doing much work on this for a few months at least.  Here's a screenshot we got on the other forum showing the startup transient.