My Project: Modern Linear Power Supply

[Kleinstein]

I'd consider doing 4 x 9v-12v AC secondary windings to allow for switching between low voltage / high current or high voltage / default current  (parallel 2 / series 4 )  and also be able to use relays to switch between 1-4 windings to have less heat produced.

There's reasonably cheap solid state relays that can handle up to 7-10A of current and AC voltages which you could use instead of mechanical relays if you're worried about longevity.

With relay switching more winding steps would make sense. Quite some commercial supplies use 3 or 4 steps. This gives smaller steps and thus also smaller current spikes. There is still relay wear with high power and large capacitors.

SSRs could work, but the peak current is really high - the SCR simulation showed some 70 A and this may not yet be the worst case.
Another option would be MOSFET switching after the filter capacitors. This would need extra filter capacitor (2 x for half the voltage), but it would nearly eliminate the current spike and allow less hysteresis in the switching.

[B0B45]

SSRs could work, but the peak current is really high - the SCR simulation showed some 70 A and this may not yet be the worst case.
Another option would be MOSFET switching after the filter capacitors. This would need extra filter capacitor (2 x for half the voltage), but it would nearly eliminate the current spike and allow less hysteresis in the switching.

Careful, the simulation is just a proof of concept—so ESR is just made up. Measurement shows about 40A max for one half-wave. I can later check the real ESR.

For now, I'm going to get some SCRs and diodes for a test. All parts I selected should be able to handle 100A+ on short term easily. 

[B0B45]

Think Off-On slew rate is very fast. Combined with long cables and specific loads, this can cause unpleasant effects. Consider slowing down the slew rate.
My power supply used to respond about the same as yours, but then I reduced the slew rate by about a factor of 10 to minimize overshoot.

Btw. how did you reduced the slew rate?

[temperance]

Btw. how did you reduced the slew rate?

Strange question. Who developed this power supply?

[sorin]

Btw. how did you reduced the slew rate?

No offense, but if you're not familiar with how to reduce the slew rate, you probably shouldn't be designing a power supply.

[xavier60]

The slew rate limiting of my bench PSU is due to it going into current limiting while charging the output capacitor.
Voltage over-shoot is greatly minimized by taking the CV compensation feedback from the OR-ing node rather than the output of the opamp.
The idea was originally suggested by Kleinstein some years ago. The idea also solves over-shooting when transitioning from CC to CV.
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664

[temperance]

The slew rate limiting of my bench PSU is due to it going into current limiting while charging the output capacitor.
Voltage over-shoot is greatly minimized by taking the CV compensation feedback from the OR-ing node rather than the output of the opamp.
The idea was originally suggested by Kleinstein some years ago. The idea also solves over-shooting when transitioning from CC to CV.
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664

And another thing to "tame" the overshoot when switching mode is to use op amps able to quickly recover from saturation.

[B0B45]

I was simply asking d-el about his changes to his own power supply.

In the first post, I already wrote that I used an old design and redesigned it with modern parts. So it's the first time I'm dealing with power supply regulations and stuff. That's why I posted all the info for suggestions and maybe some tips.

[sorin]

And another thing to "tame" the overshoot when switching mode is to use op amps able to quickly recover from saturation.

I don't think this is necessary. You can prevent it from going into saturation in the first place (with D11 and D12).

I was simply asking d-el about his changes to his own power supply.

In the first post, I already wrote that I used an old design and redesigned it with modern parts. So it's the first time I'm dealing with power supply regulations and stuff. That's why I posted all the info for suggestions and maybe some tips.

You can't simply modify an old design, because if you do, you need to analyze and tweak it for stability.

[xavier60]

If the output is being enabled/disable via switching of the reference voltage, it should be safe to slow the rise time of the reference voltage.

[B0B45]

To get everyone back on the same page: the overshoot d-el was talking about comes from this screenshot and the post on the "Turn on - Turn off test" topic. It shows that Ch1, with the ADA4522 as a CC regulator, has a very high slew rate at this test: about 10-15 µs.

Reason for this: The ADA4522 is slower than the OPA182 (ch. 2), so the time until the CC regulations kick in is higher.
The simulation shows that it takes some time for the op-amps to fall from 12V until the regulation is active. With higher current settings the overshoot is gone. (See screenshots on post #3)


Other overshoots with the ADA4522 are already dealt with by adjusting the feedback cap.

[Kleinstein]

To get everyone back on the same page: the overshoot d-el was talking about comes from this screenshot and the post on the "Turn on - Turn off test" topic. It shows that Ch1, with the ADA4522 as a CC regulator, has a very high slew rate at this test: about 10-15 µs.

Reason for this: The ADA4522 is slower than the OPA182 (ch. 2), so the time until the CC regulations kick in is higher.
The simulation shows that it takes some time for the op-amps to fall from 12V until the regulation is active. With higher current settings the overshoot is gone. (See screenshots on post #3)


Other overshoots with the ADA4522 are already dealt with by adjusting the feedback cap.

One could reduce the time it takes for the CC regulator (and similar the CV part) to engage by using a lower supply for the amplifier. This way there is less time for voltage change times regulator slew rate (this is usually less than the OP-amps slow rate). Normally the slew rate is set by the feedback capacitor and related network, not by the OP-amp speed. With the relatively large resistors at the inputs there is little need for the very low noise ADA4522, as the resistor noise is higher.  There should also not be the need of an extra fast amplifier. The basic lab supply circuits get away with slow types like OP07 or LM358 (though not ideal with the additional cross over delay). The relatively fast output transistors used may allow a rather fast setting, but it than gets tricky with parasitic inductance and similar.

For more freedom to tweek a resistor in series to C6 (and similar at the voltage regulator) can help.
Another point is having output capacitance with controlled / additional series resistance. This helps with variable output load.

[temperance]

I don't see those diodes in the schematic you posted. But a modern op amp with very fast overdrive recovery, costing a lot less than the op amps you are using now is the TLE2141. It takes only 150 ns to recover and has a slew rate of 27V/μs.

[B0B45]

One could reduce the time it takes for the CC regulator (and similar the CV part) to engage by using a lower supply for the amplifier. This way there is less time for voltage change times regulator slew rate (this is usually less than the OP-amps slow rate).
...
For more freedom to tweek a resistor in series to C6 (and similar at the voltage regulator) can help.
Another point is having output capacitance with controlled / additional series resistance. This helps with variable output load.

I had the same idea and reduced the supply voltage for the op-amp to 4V. Now the time until the CC regulation kicks in is about 2 µs. Guess this is fast enough for most applications.

I'll add an extra resistor in series with the feedback cap in the PCB update. Controlled resistors on the output capacitance sound interesting, but at this moment, it's maybe a little bit too much. 

I don't see those diodes in the schematic you posted. But a modern op amp with very fast overdrive recovery, costing a lot less than the op amps you are using now is the TLE2141. It takes only 150 ns to recover and has a slew rate of 27V/μs.

I only got one ADA4522 as a comparison test to the OPA182 I originally used. So I replayed the op-amps on ch1 and did some screenshots. The schematics I posted in my first post also show the OPA(2)182.

The TLE2141 is an interesting idea, at least for the CC-Reg. If I get my hands on one, I will test it.

[temperance]

I found a lower cost op amp with a recovery time of 200 ns: TLV2197

[B0B45]

I found a lower cost op amp with a recovery time of 200 ns: TLV2197

Today I tested the OPA189 with an overload recovery time of 320 ns. See screenshots: 8V, 40mA, 50 Ohm.

The last days I was updating the schematic and the PCB. It's not 100% final yet, but it shows the main changes.

Happy New Year.

Bob

[B0B45]

The slew rate limiting of my bench PSU is due to it going into current limiting while charging the output capacitor.
Voltage over-shoot is greatly minimized by taking the CV compensation feedback from the OR-ing node rather than the output of the opamp.
The idea was originally suggested by Kleinstein some years ago. The idea also solves over-shooting when transitioning from CC to CV.
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664

Hallo Xavier,

I tested your idea of rearranging the feedback capacity after I got the LTspice simulation running. It really worked. Although, I may lose a little of the phase margin.

[xavier60]

Im not too sure about that. In my case, the mod is applied to the CV feedback to control voltage over-shoot when the PSU transitions from CC to CV, for example when a short-circuit is removed from the output.
From experience, the idea can't be applied to the CV and CC sections without causing instability at the CV/CC transition.

[temperance]

I hate schematics with rectangles to represent op amps. They are just unreadable.

[B0B45]

Little update for my power supply with PCB 1.2:

- Relay got replaced with SCRs
- Regulation has some changes -no more overshoots and fast CC-CV switching

I attached the last schematic and some screenshots.

For now it works pretty fine, but I'll test it a few days till I reassemble the casing.

Edit: minor change in the schematic.

[B0B45]

For now I can say:

Voltage control works fine (pic 1).

Current control works fine.

The CV to CC transition (and back) works fine.

The only issue or glitch I found is that in CC-mode there is a little 50Hz ripple (around 5mV AC) when using a load above ~20 Ohms (pic 2). If I switch the tip and ground clip of my scope on the output connections, the waveform gets lower but doesn't disappear (pic 3).

I did the same test on my Statron power supply, and I got even worse results (pic 4).

Not sure if this comes from the ground clip itself. Same results by using the short ground-spring. What I can say its perfectly matches the mains voltage.   

Does somebody have an idea where this effect might come from?

Kind regards,
Bob

[PCB.Wiz]

The only issue or glitch I found is that in CC-mode there is a little 50Hz ripple (around 5mV AC) when using a load above ~20 Ohms (pic 2).

How does that ripple change with current set point?

What I can say its perfectly matches the mains voltage.   
Does somebody have an idea where this effect might come from?

Any current sense regulator is going to be susceptible to magnetic fields.
PSRR issues usually appear at rectified power frequencies and waveforms (100Hz, ~sawtooth),

It cannot actually be perfectly matches the mains voltage, but it will have some phase relationship, plus some high pass elements.
If you have low voltage AC available, you can try simple RC coupling to give to a similar phase and harmonics