LM324 Power Supply with variable voltage and current

[Kleinstein]

One can do a few more experiment with different C1 R2 values. I would prefer output impedance in the AC mode here, as it includes all possible capacitors at the load in one simulation.  This simulation can also be used to find good values for C7/C10.

Most of the time, if the impedance is good, the transient simulations also look good.  Specific to the transients going all the way to saturation it may help to have back to back diodes across the inputs of U4. It's worth a try at least in the simulation to see if it helps.

Some ringing in case of a very difficult load (large cap with very low ESR) is normal - it is already a good sign if the regulator does not permanently oscillate under that condition. Limiting the circuit to rather slow TIP3055 and LM324 OPs means one should not expect high end performance.

[iMo]

C1 from 330p-10n.
All caps with some esr except the comp ones.

[iMo]

Output impedance with C1=10n and R2=10-100k

[iMo]

And the 28.png (1000uF 5mOhm) test with C1=4n7 and R2=100k

It oscillates with small capacitive loads, however.

It seems we need a switchable R2, 10k for small cap loads, 100k for large one

[xavier60]

The needed compensation should be about the same as what my bench supply project is happy with.
The photos show load transient tests with a pulsed 2A constant current load.
The 2nd photo shows some ringing at about 8Khz with a 470µF super low ESR polymer capacitor connected directly across the output at the regulator's PCB.
It also causes  multiple bouncing after unloading.
Common type low ESR electos don't cause ringing.
The ringing and bouncing don't occur when I connect the polymer capacitor via 20cm of 24AWG twisted pair.
I believe that this behavior is expected from current sourcing designs including Harrison topology and is not a practical problem.

https://www.eevblog.com/forum/projects/linear-lab-power-supply/msg2388873/#msg2388873

[iMo]

V12

[Kleinstein]

Output impedance with C1=10n and R2=10-100k

I think there is something wrong with the way the output impedance is simulated. Some 100 Ohms over a large range would be really bad.  I did a simulation of an older version - to get a rough idea what C4 R4 values to use. It did look much more reasonable. So I think it's more a problem with the simulation, not the circuit.

@xavier60: some ringing with a super low ESR cap at the output is hard to avoid and probably to be accepted.
This is because the typical impedance curve of the output will look like very much like an inductor, especially in the 100 Hz to 10 kHz range. One essentially needed the impedance to go down to lower frequencies - so essentially no good way to avoid this. One can and should  (R4 and C4 are good for this) deviate from the simple  Z ~ f curve in the 100 kHz range. This at least avoids ringing with more common 10-100  µF range low ESR caps.

[xavier60]

@Kleinstein, This possibly relates to what you said. When I increase the size of the compensation capacitor from 150pF, the ringing with the 470µF super low ESR  capacitor becomes much larger.
I am completely satisfied with the performance of my bench supply project.

[iMo]

@Kleinstein: would be nice to see your simulation of the output impedance (incl. the .asc). Thanks.

[mike_mike]

Is there any possibility to use an alternative for the switch across the R2 (91k) ? 

@xavier60, Can TLC072 be replaced by another op amp, in your power supply project ?

[xavier60]

I recently tried an LM358 in place of the TLC072. There was very little difference in performance. Mainly the diode fast limiting was slower at 10µs instead of 2µs which isn't important anyway.

[mike_mike]

@xavier60 Where does CC and CV led's need to be connected ? To the output of the CC and CV op amps ? (in your power supply design)

[xavier60]

@xavier60 Where does CC and CV led's need to be connected ? To the output of the CC and CV op amps ? (in your power supply design)

Yes, they connect to the respective op-amp outputs. The way my design is configured, the LED current is about 200µA which gives suitable brightness with high efficiency LEDs which are commonly found on eBay now.
With your current design using 12V control rail and the operating point of op-amps, the LED current will be something higher using 10K resistors.

Also: If for some reason the regulator's output voltage is forced higher than its setting, the CV LED will glow extra bright.

[mike_mike]

@xavier60
Can you please answer a few questions about your PSU ?
1. Can R32 and R33 be omitted ?
2. What type of diodes are D2, D3 and D10 ? 1N4148 ?
3. C9 is polarized ?
4. Where should be connected the shutdown K of the diode ?
5. What are the values for R23 and R24 ?
6. This PSU has been tested ?

[xavier60]

@xavier60
Can you please answer a few questions about your PSU ?
1. Can R32 and R33 be omitted ?
2. What type of diodes are D2, D3 and D10 ? 1N4148 ?
3. C9 is polarized ?
4. Where should be connected the shutdown K of the diode ?
5. It is good to build this PSU ?

R32 is a 100mA polyswitch and isn't important if there is no concern about the regulator breaking down.
R33 is a 100mA polyswitch which protects the preload circuit from reverse polarity. The whole preload circuit can be replace by a resistor.
Diodes  D2, D3 and D10 can be 1N4148.
C9 can be a polarized electrolytic capacitor. I like to use leaded monolithic ceramic capacitors.
Shorting the shutdown K of the diode to ground turns off the regulator's output.
My design and the one imo is developing are very similar. You should stay with that design with some changes.
The under voltage lockout idea with D9 should work well. It might not need D6 and R16. The gain of Q1 is easy to change by changing the value of R25 if necessary.
My design does not have a reliable under voltage lockout to suit other op-amp types. 
The CC buffer can go and there is no need for the op-amp for the CC LED, just copy what I have done.
Ill stop at this point. Draw up a new circuit, then we will sort out more details.

R23 and R24 are 10K 10 turn pots.

[mike_mike]

I would like to built your PSU circuit  ... if it is tested and working good.
Are the testing results any good ?
I need something that does not need so much testing ... something that is already tested and it is working good 

Later Edit: what can go wrong if the under voltage lockup is missing ?

[Kleinstein]

@Kleinstein: would be nice to see your simulation of the output impedance (incl. the .asc). Thanks.

Here is a simulation file for the Z_out simulation. Its based on an older Version and with slightly changed OPs / transistors to use models that I had on the PC. The response that version looks reasonably, though still with a relatively slow regulation.
The curve is the result, with the Zout still in dB (rel 1 Ohm), so 0 dB is 1 Ohms and -60 dB is 1 mOhms

[iMo]

@Kleinstein: thanks, I wanted to see how you did the .ac analysis. I did it exactly the same way, but with the recent version of the PSU schematics, which is way different than yours. Thus the results are pretty different too..

[xavier60]

I would like to built your PSU circuit  ... if it is tested and working good.
Are the testing results any good ?
I need something that does not need so much testing ... something that is already tested and it is working good 

Later Edit: what can go wrong if the under voltage lockup is missing ?

If the ULVO doesn't work properly, the output voltage can go higher than the set voltage after the mains is switched off.
My design will work well if built the same as the schematic. I understand that some of the parts you already have are different.
Testing can't be avoided.
You should not want to completely abandon another member's work after they have made so much effort.

[mike_mike]

You should not want to completely abandon another member's work after they have made so much effort.

I will not abandon it.
I just wait with impatience for @imo and @Kleinstein to find the final version, of the design.

Does anybody has the final version ?

[iMo]

The version 12 could be taken as a solution worth of implementing in HW.
You should implement it into the real HW with care.
1. you do not need the switch nor the 91k resistor, unless you plan to feed systems with large input capacitors with <0.01ohm esr
2. as others wrote you may not need the CC's pot buffer
3. you may also get rid of the CC_LED opamp, as an LED could be wired directly to to CC opamp's output (via a 10k resistor for example against Vcc)
4. on real HW you may want to optimize the compensation capacitor's values as described earlier
5. my favorite hint - wire a large resistor (ie 1Meg) from the pot's wiper to the gnd, in case your pot is a cheap plastic crap
6. there is not such thing like final design, all designs include at least one issue, and in the moment as a design is implemented in the HW it becomes obsolete

[mike_mike]

3. you may also get rid of the CC_LED opamp, as an LED could be wired directly to to CC opamp's output (via a 10k resistor for example against gnd)

The LED should be wired from the op amp's output to gnd, in series with a 10k resitor ?
 
Later Edit: if I will remove the CC pot buffer and the led notification op amp (CC led op amp), then what should I do with them ? LM324 has 4 op amps and I use only 2 of them in this design ...

I removed the CC pot buffer and CC led op amp.
I added the 1Meg resistors.

Please have a look at the schematic and tell me if it is good.

[Kleinstein]

The last version is not that much different from the Version from Xavier. It is the same type of circuit with more minor differences

The circuit is about ready to build in real hardware, e.g. on a perf.-board or similar.
The LM324 has 4 OPs, so one could keep the OP for the mode LED. The 4 th OP could control a fan or over-temperature off.
Alternative there are dual OPs like LM358.
If really build for a high current, I would consider a better (lower drift, e.g. TLE2021) OP for the CC mode. This could permit a smaller shunt resistor and less trouble with self heating of the shunt. With the LM324 class OP one would have quite some drift at low currents or heat problems with the shunt at high currents.

The possible 2 diodes in parallel to R30, I already mentioned earlier would be the main addition I would consider.
For a first test, I would start with only one power transistor and thus less maximum current. I would add the other only after a first test.

[iMo]

3. you may also get rid of the CC_LED opamp, as an LED could be wired directly to to CC opamp's output (via a 10k resistor for example against gnd)

The LED should be wired from the op amp's output to gnd, in series with a 10k resitor ?

I fixed above, it should be wired from CC opamp's output to Vcc via 10k resistor. The LED current will be from 10uA to 440uA, I think enough for the indication.

[iMo]

Later Edit: if I will remove the CC pot buffer and the led notification op amp (CC led op amp), then what should I do with them ? LM324 has 4 op amps and I use only 2 of them in this design ...

I would cut the LM324 package in half and use the spare opapm's for a second PSU..