LM324 Power Supply with variable voltage and current

[mike_mike]

... Or probably I could buy 2x higher power transistors than 2N3055 and TIP35C, and replace them, for example 2SC5200, MJ15003, 2N5038 or 2N3772, which would be simpler, but for 3A output current, probably I am fine with only 2 transistors.

Another thing that I could do is to use a transformer tap switcher, using an relay and some additional components, because I have a transformer with 2 secondaries, each 15Vac... I think i will use the tap switching method

[mike_mike]

Using the schematic at pct. 3 from this website: https://sound-au.com/articles/preregulators.htm
I drew my own schematic (please find attached).
Also, using the 12V output I can power the fans that are cooling the heatsink...

If I build this tap switcher, what should I test, beside the thing that it switch the taps of the transformer ?

[xavier60]

That will work but might be tricky adjusting the voltage threshold and hysteresis.
Something more refined can be designed with opamps.
Depends on how you disable the PSUs output, directly with a high current switch or if you use a small switch on the Shutdown pin, the relay will change from 30V to 15V every time the PSU output is turned off and if the output  is allowed to go low.
Using a dual opamp, it might be possible to design something that holds the relay's state while the output is switched off.

[mike_mike]

Ok, then I will go back to the ideea to add a new transistor, TIP35C / 2N3055.
What should I test ?

[xavier60]

Ok, then I will go back to the ideea to add a new transistor, TIP35C / 2N3055.
What should I test ?

Just add the extra transistor and check the stability using 1A pulsed load. You know what the correct response looks like so no need to show the trace if it looks right. I prefer the TIP35C.

You don't want an under damped response, https://www.quora.com/What-are-over-damped-critically-and-under-damped-systems

You can always add the tap change later. This is what I have in mind. U1A is configured as a Schmitt trigger, sensing the PSUs output voltage.
One section of S1 connects the PSUs Shutdown pin to ground while the other section disconnects the output of U1A from the input of U1B which stores the relay's state.

[mike_mike]

I added the 3rd transistor (TIP35C) to my circuit.
I think that the response is almost good, probably a little bit underdamped...
1. Vout = 20V, Imax = 3.39A, Rload = 20R
111_000, 111_001
2. Vout = 10V, Imax = 3.39A, Rload = 10R
111_002

Edit: also, I used for R8 an 680k resistor in parallel with an 470k resistor, and now the max current is 3.39A.

[xavier60]

Yes, a little underdamped. nothing to worry about.

[mike_mike]

I've seen that if I set the PSU to 0V output and 0A output (so both pots at minimum) then if I press the shutdown button the green and red led's are still ON.
Also, while the PSU is with the red led ON, and if I press the shutdown button, then the red led remains ON.
Are these 2 behaviors normally ?

... and the test with 3x 2N3055. l tested with 3x 2N3055, because when I don't have any TIP35C, I will probably use 2N3055
1. Vout = 20V, Imax=3.39A, Rload=20R
113_000
2. Vout = 10V, Imax=3.39A, Rload=10R
113_001
Are the results still ok, with 2N3055 ?

[xavier60]

Yes, normal.
The preload circuit keeps a residual 10mV on the output. Assuming that the CV opamp has very little offset error, the divided 10mV  will cause the opamp's output to swing low.
The intentional input offset applied to the CC opamp causes the same thing.

[mike_mike]

I was curious and I checked the response of the PSU (3x2N3055) with Vout=20V and Rload=10R, so a 2A load:
114_000
Also, with Vout=32V and Rload=10R, so 3.2A load:
114_001

In this case, the response is also normal ? Or the voltage overshoot should be a little less ?

I know that it is normal for the voltage dip and overshoot to increase if the switched load increases ...

Edit: same test, but using 3xTIP35C:
115_000 - Vout=32V, Rload=10R,
115_001 - Vout=20V, Rload=10R

[xavier60]

You could try changing to values of R5 and C3 like before.

[mike_mike]

I tried first to modify the value or R5, leaving C3 at 1nF. Also, I modified R2 from 3k9 to 4k7, to reduce the output voltage from maximum 32V to maximum 27V.
1. R5=3k9, 3xTIP35C
Vout=27V, Imax=3.39A, Rload=10R
116_000
2. R5=4k7, 3xTIP35C
Vout=27V, Imax=3.39A, Rload=10R
116_001
3. R5=4k7, 3x2N3055
Vout=27V, Imax=3.39A, Rload=10R
116_002

Please have a look and tell me what you think ?

[xavier60]

They all look good now. I would have expected it needing a combination of reducing R5 and increasing C3.
Was R5 3.3k?

[mike_mike]

Yes, R5 was 3.3k.

[xavier60]

Yes, R5 was 3.3k.

I think that means the 1nF was a bit too small with 3.3k. Or likely something going on that I don't understand.
The main thing is it's working right.

[mike_mike]

I also tested the PSU with the PSU Shorter, with zero ohm load, to check it's stability when there is a short on the output.
Please have a look and tell me what you think.

1. LM358, 3x TIP35C, Vout=27V, Imax=2.98A
121_000, 121_001
2. LM258, 3X2N3055, Vout=28V, Imax=3.29A
122_000, 122_001
3. On the shunt resistor, LM358, 3x TIP35C, Vout=28V, Imax=2.98A
120_001

The main thing that I don't understand is the waveform marked by the red arrow (121_000_arrow).

[xavier60]

That would be recovery overshoot followed by some undershoot. Are you able to get TLC072 opamps?

What I'm not certain about is the long delay of the voltage recovering in some of them, as the one with the red arrow.

[xavier60]

Forget that last bit about the delay. It's just looks that way because of the expanded vertical.
All looks fine except for the slow opamp which will not be a problem in use anyway.

[mike_mike]

Are you able to get TLC072 opamps?

Sadly, I am not able to get TLC072, because I can't find them at my electronic parts dealer, and if I will get them from another dealer, then it would be very expensive...
I would like to stick to LM358/258.

[mike_mike]

I replaced the C3 = 1nF MLCC (multilayer ceramic capacitor) by a ceramic disc capacitor (also 1nF), and I got the following results:
1. 3x2N3055
117_001
2. 3xTIP35C
117_002
This response look like a little bit overdamped, does it is still OK ?

This was the response with 1nF MLCC capacitor:
117_000

[xavier60]

That possibly explains why C3 capacitance seemed low. Most types of MLCC are used for bypassing and have large tolerances although low values are usually more accurate because they are more likely to be used where accuracy is more important.
 Some overdamping is fine.

[mike_mike]

I checked this time the power supply using the 3x TIP35C transistors... this PSU is already installed into a case, with the transformer, rectifier, heatsink, voltmeter. The voltmeter and fans were disconnected while running the tests. The fans are powered using a separate PSU, with fixed voltage.
What I found is that even if there is R5=4k7 and C3=1nF and op amp is LM358P from TI, there is a little underdamped response.
Please have a look and tell me what you think.
I uploaded only the screenshots were the overshoot is more visible...
1. Vout=27V, Rload=10R
126_013
2. Vout=27V, Rload=47R
126_005
3. Vout=20V, Rload=10R
126_012
4. Vout=4.7V, Rload=4.7R
126_009

[mike_mike]

... by increasing R5 to 6k8 and leaving C3 at 1nF, I managed to reduce the overshoot:
Vout=27V, Rload=10R

[xavier60]

With my PSU, I put a 1nF cap in parallel with C3, 120pF with R5, 10K. The response went  from 8uS to 50uS with a small overshoot bump.
I now think that the small overshoot bump may not indicate underdamped response, so ignore it for now.
The affect of the 1nF on my PSU's response means that there is a big difference, could be because of the different output transistors.

You could just leave yours as it is or do more experimenting if you have nothing else to do.
I'm a bit curious to know if it is still unstable with R5, 10K.

[mike_mike]

You could just leave yours as it is or do more experimenting if you have nothing else to do.
I'm a bit curious to know if it is still unstable with R5, 10K.

Should I leave the PSU as it is with R5=4k7 or R5=6k8 ? and C3=1nF  ?

I will try an experiment with R5=10k, for curiosity...