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

[xavier60]

Maybe I'm seeing it differently. The compensation RC should be wired the same as in the original schematic. R21 from the 2N3055's Emitter to the Inverting input. C7 from the CC opamp's output to the Inverting input.

[iMo]

Ok, with 10k in series with the 1k5 sense resistor and 6n8 ceramics or 10n ceramics (3mm long leads directly wired to in- and out of the opamp) is the amplitude at the CC opamp's output 600mVpp and period 600ns (10n ceramic).
Touching the opamp's output with my finger decreases the amplitude to 400mVpp.

I boosted the ground pcb from the probe point to the point near the 741's ground with a thick short wire with none change.

[xavier60]

With 10K and 10n, it's unlikely that the 1.7MHz is coming through the opamp in any normal way. What signal level is there on both the opamp's inputs?
Try touching parts of the circuit to see if anywhere changes the frequency.

[iMo]

The signal at the in- is the ~same as the output (10n fb cap), in+ is clean.

Touching the CC's in- or output decreases ampl to 400mVpp and period increases to 800ns.

Other 324 pins incl 741's pins do not show any relevant impact.

Transistors - the touching of the base of the first transistor - the same change as with in-/out of the CC opamp, other points around the transistors - no relevant impact.

[xavier60]

The signal at the in- is the ~same as the output (10n fb cap), in+ is clean.

That should mean that the signal is being imposed directly onto the output somehow or directly onto the Inverting input rather than via R21.
I can't imagine how.
Edit: try removing D12 and carefully testing CC mode.

[xavier60]

The signal at the in- is the ~same as the output (10n fb cap), in+ is clean.

Touching the CC's in- or output decreases ampl to 400mVpp and period increases to 800ns.

Other 324 pins incl 741's pins do not show any relevant impact.

Transistors - the touching of the base of the first transistor - the same change as with in-/out of the CC opamp, other points around the transistors - no relevant impact.

Have you tried another opamp?

[iMo]

With the D12 disconnected the amplitude at the CC opamp's output is 800mVpp period 750ns.

The replacing the 324 will be the last step
Nope, I will do it now, I put there a socket..

[iMo]

So - I've desoldered the 324, it is TI's LM324.
Soldered in the gold plated socket. I have not found the 324 in my junkbox, so now using TL084 (TI).

CV - as in my schematics.

CC - 100pF in the fb (orig one from top side) and 1k5 to shunt sense, added a 100nF decoupling

The circuit behaves differently:

1. with CC engaged no oscillation, noise below 20mVpp at CC's opamp output
2. with CC disengaged and loaded (2.2A, 18.7V):
a. when PSU switched on there is 30-40mVpp 270ns period oscillation at CC's output, it slowly decreases to 20mVpp, sometimes disappears (noise below 10mV at CC's opamp output) and appears, it disappears when touching output of the 741 or the feedback of the 741
b. when the load is reinserted (2.2A 18.7V) the oscillation starts again at 30-40mVpp, then within aprox 15secs its amplitude goes down and disappears (noise below 10mV at CC's output), but still sensitive..

3. after 40minutes when the board is already hot the CC's opamp output shows 20mVpp 150ns period oscillation CC engaged or disengaged..

The TL084 gets pretty hot (+12/-12V).

A really mysterious stuff, indeed

I have here a different 084 and a TLC274, I will also try to find a 324 somewhere as well..

[xavier60]

Keep in mind that the opamp for the inactive mode should have its output at maximum + voltage, about 11V and not capable of producing any output signal.

[iMo]

It might come from the 741 via the diode's capacitance (the 084's output is not zero impedance), even though those frequencies are rather odd for that circuit. Later on I will continue (perhaps I would socket the 741 too) - it is enough for today.. Thanks!

PS: next step - to handle the inputs of the two not used opamps in the package, perhaps they mess up..

[iMo]

So, I have wired together the in- with its out at both spare opamps in the TL084, and their in+ to ground (the ground in the schematics).
The CV and CC wiring remains as the latest above.
The load is the ww resistor and it gives 2.27A at 18.7V (the max PSU voltage here) with CC disengaged.

After power up the PSU I can see aprox 10mVpp 300ns period oscillation (probe 1x, AC, BW limit off) at the CC opamp's output, which is slowly dropping down (within 5-10secs) into the noise of my Rigol 1062CA (perhaps 1-2mVpp). That is with the CC disengaged. With engaged CC I see clean output (<2mVpp noise).

The output of the 741 is clean.

In the lower portion of spectra I can see 5mVpp saw tooth ripple w/ 10ms period at the CC opamp's output, and 7.5mVpp at the CV opamp's output. The ripple increases a bit (30mV) when going to CC's zero current output.

The above oscillation is superimposed on this ripple.

The 10mVpp 300ns oscillation at the CC opamp's output appears always when doing abrupt changes, it slowly disappears into the background noise after couple of seconds.

So I will leave it now and will have a look at it tomorrow, what is new there as it lives its own life, it seems

[iMo]

Some shots from this morning..
Immediately after the cold start the oscillation (2.3A/18.7V CC disengaged) was as below. It lasted for about 30-40min without disappearing. Afterwards after re-inserting the load the effect of slow disappearing started again.
The 741 CV output is clean - see below the ripple.
With CC engaged I do not see oscillation.
The CC output is DC 11.56V with disengaged.
I replaced the original 100p capacitor in the CC fb with 120p C0G, but it did not help. Cleaned with IPA couple of times.
The oscillation is visible at the output posts as well.
Probe Tek 1x, BW off, AC, no oscillation or ripple seen with probe's gnd and input put on the same node at PSU.

It seems the rising edge of the ripple starts the oscillation burst (it reminds me on my experiments with superregeneration ), the oscillation somehow depends on the temperature, the effect of disappearing with 5-10sec time constant leads to some thermal effect or leakage or??

[xavier60]

Post a schematic of how everything is. The LM741 has been replaced?
 
Also, a load transient test might show if there is loop instability.

[iMo]

Below the mod v5 schematics.
There is one thing which may cause the instability - the 200mA/3A switch is mounted on the front panel, shorting the 4R7 for 3A (aprox). The leads are thick, but aprox 15cm (!!) long coming to the shunt on the pcb. That may cause some HF feedback.. I will doublecheck that later on..
Ideally I planned to place a relay near the shunt with the 4R7, but there is small space for it and of course the laziness.. 

[xavier60]

If you plan to do the load transient test, go ahead.
If not, remove the 330pF   from across the 741's inputs. Put a 1nF capacitor or higher between its output and Inverting input.
The 10K between the Inverting input and GND must be left in place.

[iMo]

The transient test will come next week, I need some low L higher power resistors for the load at the shorter and some rugged construction for it as well..

[xavier60]

I did more experimenting. The 120pF that I though was ok to put across the inputs, isn't.
And it has just occurred to me that even with a large capacitor between output and Inverting input of the CV opamp, the response can never be purely Integral. The AC or proportional gain can never be lower than one.
 

[iMo]

Here is the CC and CV out with only 1n ceramic in the CV fb only (330p at 741's inputs removed).
10k in 741 in- is there.

[xavier60]

Maybe it's oscillation in the output stage. If so, the signal should be a bit stronger on the anode of D12.

[iMo]

 

I've removed the rather long leads (my really bad idea) from the shunt (the 0.33ohm placed in rear part of the box) to my new switch on the front panel (creating a loop antenna) and all oscillations are gone.. (with 1n only in the 741 feedback). What is left are the couple of mVpp 100Hz ripples at the CV and CC opamp's outputs.
I will make the transient test next weeks.
Also I have to solve the 200mA/3A switching as well as I need the lower CC range as well..
Thanks!

PS: ok.. when not having an oscilloscope handy I would consider my idea with the switch pretty 
 

[iMo]

Here is the "PSU Pulser" (aka "PSU Shorter" in past)..
Rather hot at 18.7V and 1.7A..
40ms active, 300ms off. The load 68ohm/6 active, 330ohm off.
Rising edge on the mosfet's drain 200ns, falling edge 60ns at 18.7V/1.7A.
So we may finetune the PSU's CC/CV now.. 

[iMo]

After a final exercises with ltspice and PSU pulser/shorter below the o'scope pictures when pulsing with 1.65A (set 18.7V max voltage, CC off).
The 470uF original output capacitor replaced with 47uF one..
The pos/neg peaks at the PSU's posts are less than 100ns wide (when zooming in)..

[iMo]

..and the final schematics.. case closed..

[xavier60]

The narrow spike is likely due to the way the DSO probe is connected to the PSU.
The real response is 4us, a little fast for that design resulting in the slight ringing. It will be fine.