I was hoping this test would show something obvious! Are the feedback networks identical? I'm running out of ideas now.
I will measure the values of their components.
Do you think the channels track well enough? It looks pretty decent to me
Ok, the feedback resistors measure identically. Oddly enough, one channel actually has about 0.7 the gain of the other when the feedback is connected.
Sweeping the frequency didn't reveal anything odd, both channels rolled off thanks to Cdom, although the "faulty" channel had a SLIGHTLY less "perfect" rolloff, but the phase and amplitudes were very close over a wide bandwidth.
I'm stumped!
That probably means one of them is using the wrong secondary. Double-check the switching around the OPT secs and feedback networks. Is it the 'low gain' one which oscillates?
It is the low gain one which oscillates, but the gains match nearly perfectly open loop, which may or may not be a coincidence.
There may have been a mistake in the leadouts on the transformers
Wait, when applying the same signal directly to each transformer primaries with the sig gen, both transformers had an identical voltage output...
Edit: I currently just have the transformer hard wired for 8 ohms.
Edit: I currently just have the transformer hard wired for 8 ohms.
Equal gain open-loop and equal feedback components would necessarily give equal gain closed loop. Something is wrong. 0.7 just happens to be the voltage ratio between 4 and 8 or 8 and 16. Putting it another way, the low gain channel has 1.4 times too much feedback which is why it is oscillating. If you used the 4ohm feedback for the 8 ohm secondary this would do it. (or 8 ohm feedback for 16ohm sec). Double check that area.
WHOOPS, my bad, the closed loop gains are identical. I must have been doing something silly originally.
I'll take a picture of the closed loop square wave response of both channels.
I'll take a picture of the closed loop square wave response of both channels.
How absolutely bizarre. With the feedback capacitor reconnected, and feedback on both channels reconnected, the square wave responses now match.
Maybe there was something conductive under the PCB effectively increasing the feedback???
Maybe there was something conductive under the PCB effectively increasing the feedback???
Bad behaviour with no load is always possible, because the OPT will have nothing much to damp its HF resonance. This causes a sudden change in feedback phase. A Zobel network should cure that. The important thing is that it behaves OK with a load. Most people probably never test their amp without a load, because of the danger to the OPT.
26 dB GNFB is quite demanding for OPT. I am quite sure that when you reduce the NFB to 20...22 dB, there is no oscillation anymore.
Sorry for seemingly ignoring your advice on this - I am slightly reluctant to reduce the feedback because I'm trying to reproduce the amplifier as exactly as possible.
As it happens, a 220n + 10R zobel cures the no load oscillation. Sadly, 100n isn't quite enough (though the oscillation doesn't hit full amplitude)
It seems to clean up the square waves a bit too.
Decent enough for what? Square wave testing tells you as much about the square wave as it tell you about the amp. There seems to be some asymmetry, so maybe there is a little slew rate limiting.
One more thing to eliminate as a possibility: Also check the values of the feedback "load" resistors (the lower leg of the NFB circuit). If these were off significantly, it would not show up much at all under OL gain tests, but it sure would under CL gain tests!
Dave
Edit: Sorry -- I see the CL gain checks are now showing the channels to be equal.
Dave
Edit: Sorry -- I see the CL gain checks are now showing the channels to be equal.
Last edited:
- Status
- Not open for further replies.