At very low frequencies you have 180 degree phase shift.
That could cause instability and the popping you mention.
Does it pop with feedback disconnected ?
That could cause instability and the popping you mention.
Does it pop with feedback disconnected ?
Last edited:
I have not been able to reproduce it since before I posted this. I'm 100% convinced it's because I used cheap crappy sockets. After things have settled in and I wiggled things around, it hasn't happened again.
Regarding the 180º shift at low frequencies, this is the part I'm still wrapping my head around. This was measured with my feedback in place, and despite it being a 180º shift, the gain was still greatly attenuated(1.7dB) compared to my normal baseline(~20dB). Can I interpret this as: yes, the circuit has a phase shift there, but as a whole, the gain at that frequency(even with the negative feedback trying to "help" it) is not enough for it to ring or oscillate? Keep in mind that when I measured this, my function generator was sitting right on top of 4Hz sine, so it had every opportunity to resonate/run away if it wanted to. Since it didn't, does this not prove that the amp as a whole is stable at this frequency?
True, there's that part. I had to take a closer look at the wave, once it gets below 15Hz it's not such a good looking sine. I don't see how the feedback would play into amplifying that well, and that may accidentally play into why the feedback isn't sustainable down there. The output gets to 180º, but it is ugly.
Gain is not in dB, is it? Shouldn't be on a log scale if it is, as dB is a log unit already.
Anyway, if gain is 20 (which I'd expect with the feedback ratio of 1/22), there's 6 dB of gain margin at both ends when phase hits 180. Can be improved by the measures I suggested earlier. Speakers are not 8 Ohm resistors, so more margin may be needed. Mechanical resonances become electrical resonances at the speaker terminals (L and C), crossover components add more reactance. A quick test for stability is to operate into open circuit, capacitor (maybe 0.47 uF), and cap and 8 Ohms in parallel. Nothing special about these loads, just easy and repeatable. Waveforms will be ugly - but if it oscillates, feedback or gain vs. frequency needs to be adjusted.
Anyway, if gain is 20 (which I'd expect with the feedback ratio of 1/22), there's 6 dB of gain margin at both ends when phase hits 180. Can be improved by the measures I suggested earlier. Speakers are not 8 Ohm resistors, so more margin may be needed. Mechanical resonances become electrical resonances at the speaker terminals (L and C), crossover components add more reactance. A quick test for stability is to operate into open circuit, capacitor (maybe 0.47 uF), and cap and 8 Ohms in parallel. Nothing special about these loads, just easy and repeatable. Waveforms will be ugly - but if it oscillates, feedback or gain vs. frequency needs to be adjusted.
Last edited:
Whomp. Good looking out.
I'm not ignoring your suggestions. As I've mentioned in other threads, I'm doing this more as a learning exercise(some of it a refresher over decades, some new). I'm trying to reach certain conclusions myself. I did end up bypassing the feedback cap today to "knock off" that small overshoot that I mentioned earlier. It took as much as a 1nF to do it. As far as the PSU, the decoupling values came from me just throwing something together to prove/disprove another very visible and strong oscillation I was dealing with. I have yet to revisit that. Same for that input cap, which was still conveniently attached from my first run. Also, I had Designing High-Fidelity Tube Preamps(Blencowe) and Valve Amplifiers(Jones), along with a few sites along the way. Every time they write about stabilization they don't really dive very deep into it. If you have any resources you can share so I can really grasp it, that would be helpful. I get the concept, but when it comes to the theory like margin and all of that, that's where it's kind of whooshing over my head.
Yep, understood. I still have to learn a bit about how to actually measure frequency response. I know that having it hooked up to my speakers is kind of a no-go for the reasons you've mentioned(not to mention annoying as hell). Is there an ideal configuration, or do people take multiple configurations(like you've suggested- pure reactive, mix, etc) and kind of work out compromises that make them all more or less as happy as they can be?
Last edited: