18V/3A HY1503 PSU - weird CC/CV behavior while charging battery cells

[xavier60]

That looks good with the 100pF. I'm curious to see it with 1K and 1nF.

[iMo]

1k and 1nF..

I am going to wire up a load pulser.. I have to find here the design (the "PSU Shorter") I made years back.. 

PS: added with an OP37 in CV to compare..

[xavier60]

Maybe C2 needs to be larger. I know from a lot of experimenting that I have never been able to get good performance from Hi-Z current sourcing LDO designs like Harrison with type 1 compensation, just the capacitor with no series resistance.

[iMo]

.. and the original 100pF and 100pF in pin5 and pin 2 grounded and 100ohm to the shunt, with 741 transistorized model
The C11 (C4 in sim) is 10nF instead of the original 1u.

[xavier60]

Does the real PSU have a problem with its CV stability?

[iMo]

The problem was with slow charging batteries - see above details. Removing 100pF from 741's pin 5 to its output cured the issue fully.
No idea on the stability when pulsing the load, it needs some equipment to create, then I will look at it with o'scope.
It needs more tests on the HW, anyhow...

[iMo]

 
The story continues.. Another issue I have not identified till today. While manually switching a 10ohm load into/from the post (playing with settings and watching some nodes on the pcb) I found out sometimes the output voltage drops by aprox 0.9V after inserting the banana into the post. That was not caused by CC setting but a random process. 
So again I tracked the stuff and again it stops at the 741. The noninverting input sometimes set itself to a 0.1xx volt. There is the C11 1u (in orig schematics) at the other side of the 7k5 resistor so I thought it is perhaps a faulty capacitor, after removing it happened still the same.
There is the C1* 100n between 741's inputs in the schematics not populated on the pcb, therefore I tried to put there something and after several xx nF experiments I finished with a C3=330pF (C3 in the LTSspice) which was enough and which helped.
I also replaced the 1uF C11 with a 22nF ceramic (C4 in the LTspice).
The latest schematics of the affected part below.
I do not know - perhaps is the 741 somehow faulty or it is a clone..too much issues around it to my taste..
What still needs to be understand better is the RC in the feedback of the 741 opamp..

[gupa]

Hello, I'm ignorant but I'm learning, somewhere I saw that with four ICs, unconnected operational amplifiers are connected to the ground so that they don't make a mess. And I see that the plus pin goes from the LM324 to the problem IC. So maybe I don't drink too much water.

[iMo]

Yep, the 324 in this PSU is using only 2 opamps, the other 2 are free, not connected to anything on the pcb.
That is actually great as it saved me replacing the entire 324 as I smoked one of the opamps inside while experimenting, so I took the spare one and it works now. But sure, the not used opamps in the package have to be handled accordingly.
All the issues I had here are around the LM741 which is a single opamp in DIL8 so no spares handy
I have the chips in my junkbox somewhere, but I am too lazy to disassemble the entire PSU in order to replace the chips, so I've been trying to debug it under a "minimal effort" methodology.. 

[xavier60]

If it's because of oscillation, try changing the 2N3055 B-E resistor to 100 ohms as well as trying to find the best compensation.
I'll do some experimenting here as I get time.

[xavier60]

I wasn't having much fun at all until I realized that I shouldn't have omitted the 10 ohm Base stopper resistor.
For the compensation, a 2.2K and 150pF gave fast stable response, 3us. Just a 470pF gave a slower stable response, 10us.
As I discovered, it's possible to have oscillations in localized areas other than the whole loop.

[xavier60]

470pF and 2.2K improves tolerance to capacitive loads.
First is a load transient test with a 470uF POSCAP directly across the output.
Then with a 0.1 ohm in series with the POSCAP.
The trace is inverted as I'm looking at the negative output WRT +Ground.

[iMo]

I disconnected the 330p at 741's inputs and wired 100ohm in parallel to the 1k at BE of the 2N3055.
It did not help with jumping voltage at switching ~10ohm load on/off.
I returned the 330p between the 741's inputs and the jumping voltage is gone.

PS: as you are well equipped - could you try the 100pF between pin5 and pin6 of the 741 as in the original - how it affects stuff, plz??

PPS: below the transient at the posts when I switch in a resistor, 15V/1.9A.

.. and the 741's output, the oscillation might come from a longer lead.

[xavier60]

I'm using an LM358. Even a 1000pF across the input only caused a small amount of ringing.
With a 470pF and 2.2K, mine settles in 10us.
That high frequency ringing could be coming from the wiring to your shorter.

[xavier60]

Your LM741 might be faulty if it's very old. I have seen old opamps randomly develop large input offsets.

[xavier60]

I'm using an LM358. Even a 1000pF across the input only caused a small amount of ringing.
With a 470pF and 2.2K, mine settles in 10us.
That high frequency ringing could be coming from the wiring to your shorter.

The test isn't valid as I had omitted R6. With R6 fitted, ringing becomes noticeable with 470pF across the inputs, so not a problem anyway.
This capacitor is likely to prevent RF getting into the opamp, rather than compensation related.

Edit: I can't see any useful purpose of putting a capacitor between pins 5 and 6 on an LM741.

[iMo]

Spent a half day with the CC. When it engages the output of the 324 oscillates 0.5Vpp at aprox 2MHz.
I have not found a way how to calm it down..
PS: talking real hw..

[xavier60]

 CC loops are usually well behaved. What compensation values are you using and what has been tried?

[iMo]

The aprox 2MHz signal at the psu's output is below 40mVpp with active CC, and the CC basically works.
There is a clean signal when CC is set off. Oscillation is visible in entire active CC region.
The original is 120pF in the fb.Tried blocking the inputs, various RC combinations in the feedback (C from 330 to 6n8, R 0/1k/4k7/10k), 22 ohm resistor at the opamp's output, messing with wiring layout, etc. The amplitude at the output of the opamp decreases perhaps by 30% with larger values of caps. The third and fourth opamp in the package is clean, supplies clean. The oscillation is of course visible everywhere with low amplitude as it is boosted by the transistors.
PS: the 1k or 100ohm in the BE of the 2N3055 has no impact on the amplitude.

[xavier60]

That's puzzling. The CC opamp is configured as a Miller Integrator which is a low pass filter. With say 1nF and 10K, the attenuation would make a 2MHz signal barely visible if at all at the output of the opamp.

[xavier60]

I added CC to my mock-up. Works fine with 10K for the compensation resistor and even with no capacitor.
I by chance noticed that the 470pF across the CV opamp inputs was causing a 200KHz oscillation at higher currents. No problems with 120pF.

[iMo]

This is 324's CC output with 120pF feedback (in mid of the CC range).
CV wired as in the above schematics.
When I disconnect its output from the 1N4007 output gets clean.

[xavier60]

Try the 6n8 and 10K. The attenuation will be about 1000:1 at 2MHz. The only possible reason I can think off so far is that somehow there is too much inductance in series with the CS resistors.

[xavier60]

I tried putting 200mV p-p of 2MHz through my output stage. Most of it gets blocked by the 10 ohm resistor at the Base of the 2N3055. I see about 10mV at the output.
What are you loading with?

[iMo]

The 10k ser 6n8 increases the amplitude almost twice (500mVpp).
The load is a 10W ww (a white ceramic brick) resistor on aprox 80cm lab wires.

The only thing which lowers the amplitude significantly (10x) is when I put my finger at the 324 output.. 
So there is a HF coupling somewhere.

My o'scope is isolated via a trafo, but connecting its ground with psu's earth does not help.

The probe is a Tek probe, 10x, grounded to the ground of the schematics.

The tracks around the opamp were cut, the 3 red wires are in+ in- and out of the CC opamp. Rearanging them does not help.