NFB network compensation help pls.

[catalin gramada]

Hello,
I trying to implement some (6db arround) nfb into my project. For some reason testing at 1 kcps with square wave is ringing a little bit too much. I tried to compensate with a capacitor over 1.5k series resistor , I usually did, but seems nothing work. Started with 500p then 1n then 2.2nf then 4,7n with no results. At 10nf it already get severe oscillations. Any thoughts how can I make it little stable,please ? Thanks



Late edit: I tried cut the high freq bypassing plate resistor into first stage with no results. The fault is from closed loop and didn't know how to compensate. The classical method bypassing series resistor didn't work for this circuit

Can be a issue of the 100uf elco and its incapacity to decouple effective in high freq, pls ?

[Chuck H]

Maybe I'm missing something, but...

It seems to me there's a DC reference missing if you have the triode cathode feeding the cathodyne PI going through a DC blocking capacitor!?! It's only DC reference is through the 1.5k and then the OT secondary winding at some minimal ohms of elevation. So ALL your bias voltage for that stage must be relative to NFB at the 1.5k/OT secondary node.

I've never seen this arrangement. It looks like nearly 100% NFB to me. The OT secondary ohms are negligible so you have to divide the 47r resistance going through the 100uf (full) bypass cap by 1.5k and THAT would seem to be your feedback ratio.

EDIT: With the "bias resistor" for the triode feeding the cathodyne going through the OT secondary the 47r resistor through the cap (AC only) would be a division from zero/0V. Not a division from the standing voltage on the feeding triodes standing bias voltage. I'm not inclined to math this out, but it seems like an unnecessarily compound circuit to me. But then, you ARE prone to such things

[catalin gramada]

Hey, thanks for input.
The bias for first stage is made over 1,5k resistor which is relative to ground through secondary winding. The feedback ratio is determined by 1,5k / 47ohm resistor which is decoupled through 100u cap. The circuit is fully functional, just this issue I mentioned above.
In a normal way I expected to can do high freq compensation with a cap over 1.5k series .For some reasons It did not work.
I thinking more the quality of electrolithic may be a problem.

[Chuck H]

Well running a simple LTSpice sim it looks to me that the 47r/100uf circuit is basically acting as a bypass cap AND a NFB shunt (Did I mention unnecessarily complicated?). If you replace the 47r with a one ohm resistor (1 frigging ohm) the standing signal on top of the one ohm resistor is 7dB @ 1kHz. (-14dB @100Hz and 14dB @ 5kHz). But that would be ALL signal standing on top of the one ohm resistor and not just NFB. I tested at full bypass (zero ohms in place of the 47r) and obviously there was no NFB (worth mentioning). So maybe try a 1%/1 ohm resistor for a nominal 7dB @ 1kHz.?. Work your way up from there.?.

[Chuck H]

Ok... Since the 47r/100uf network is acting as a bypass cap AND the 1.5k bias resistor is experiencing basically 100% NFB the gain of the stage feeding the cathodyne is affected by the bypass AND the NFB is affected by the bypass. So I think it's accurate that 1 ohm is supplying 7dB NFB, BUT, the bypass affect is simultaneously increasing gain of that feeding triode. One ohm is the balance point.

[catalin gramada]

Well, I have to keep those 1.5k resistor as bias resistor for cathode. I will play with the value of shunt resistor till desired nfb amount. Thanks

[alexradium]

Hello,
I trying to implement some (6db arround) nfb into my project. For some reason testing at 1 kcps with square wave is ringing a little bit too much. I tried to compensate with a capacitor over 1.5k series resistor , I usually did, but seems nothing work. Started with 500p then 1n then 2.2nf then 4,7n with no results. At 10nf it already get severe oscillations. Any thoughts how can I make it little stable,please ? Thanks

[ATTACH=CONFIG]56596[/ATTACH]

Late edit: I tried cut the high freq bypassing plate resistor into first stage with no results. The fault is from closed loop and didn't know how to compensate. The classical method bypassing series resistor didn't work for this circuit

Can be a issue of the 100uf elco and its incapacity to decouple effective in high freq, pls ?

that is the problem,the 100 uF has to be used between the cathode and the feedback resistor,not from cathode to ground.

[Chuck H]

I don't understand why a more typical arrangement won't work for you.?. The arrangement below adds a resistor, but you have options for a bypassing RK and full control of the feedback ratio.

[Helmholtz]

Your original circuit looks unusual but should work nevertheless. THe NFB arrangement provides both necessary DC and AC paths. If you suspect poor HF properties of the 100µ E-cap just try bypassing it with an 100nF ceramic or foil cap.

The reason for the square wave ringing is most probably a resonance of the OT and the reactive speaker load.
You might try a Zobel network (snubber) across the OT primary. Start with 1nF/2kV in series with a 4.7k/5W resistor.

Some ringing does not necessarily mean that you have a stability problem.

[catalin gramada]

Thanks for You answers..The circuit comes from an old schematic relative to applying nfb into a paraphase inverter, I remember. In any nfb circuits some bias is developed over nfb resistor, I just adapted leaving the whole bias to be developed over the nfb resistor which I don't think it is a problem btw. The problem seems to be how the cap decouple the circuit.

[catalin gramada]

The original circuit was modified to allow the whole bias over nfb resistor

[Helmholtz]

As said before, I don't think the ringing is caused by your special NFB circuit. I would most probably also show with the more conventional circuit suggested by Chuck.
IIRC have seen similar NFB arrangements in commercial circuits.

[catalin gramada]

So what are substantial diferences between those two variants pls?
Why classical arrangement with the shunt resistor in the bottom is preferred?

[Chuck H]

He did not say it was. I suggested it as a more controllable method for manipulating CK and feedback ratio independently. I wasn't focused on the ringing problem at that time. And as Helmholtz said, the ringing probably isn't a result of the feedback loop. Does the ringing disappear if you arrange for no feedback?

[Helmholtz]

So what are substantial diferences between those two variants pls?
Why classical arrangement with the shunt resistor in the bottom is preferred?

I don't think there is a significant difference.

Did you try my suggestions?