Stable global positive feedback loop?

[maxkracht]

I'm about finished rebuilding an amp and I forgot which way around the OT primary went, I also happened to have a NFB switch installed. Turn on the feedback loop and got a big volume boost. It is stable with the volume and tone control all the way up and actually sounded OK, until I turned my output attenuator all the way off and it squealed like crazy. I didn't know you could accidentally balance a positive feedback loop so well...

[R.G.]

We had several lectures on feedback stability. The fundamentals of GETTING something to oscillate is it needs feedback from the output that has;
> a signal that is in phase with the input at some frequency
and
> a total gain from the input through the amplifier, through the attenuation of the feedback path to the input of one or greater.

If any of those are not true, you can't make it oscillate.

The forward gain of the amplifier includes any rolloffs of response from input capacitors and lack of frequency response within the amplifier itself. The feedback path starts at the output pin/node goes through any resistors/capacitors/inductors/attenuation to ground/etc. all the way back to the input pin. So the amp boosts the signal bigger, and any losses in the feedback path make it smaller. If the total of +X db forward gain and - Y db feedback losses comes out to over 1 (that is, 0 db), it oscillates. If the total of Xdb - Ydb comes out less than 1, it ... can't oscillate.

Your amp seems to have either low enough forward gain or big enough stray, unintentional feedback losses after the intentional feedback is removed, that even flipping the polarity doesn't make it have enough to oscillate.

[martin manning]

Might be some phase shift involved so that the output, and therefore the feedback, isn't exactly in phase with the input.

[cdemike]

That's really interesting that it changed when your attenuator was turned off. Maybe one of the more knowledgeable members can shed some light on this, but I'm wondering if the OT's interaction with the attenuator impacted the frequency response enough that it avoided oscillating at higher attenuation levels.

[maxkracht]

It's certainly a lower gain amp and I accidentally had the wrong speaker plugged in (16 ohm speaker on an 8 ohm tap). My bench attenuator is just an L pad. It was completely stable until the last notch on the dial. No idea why the attenuator would have that much effect.

[R.G.]

The output tubes use the speaker loading (as transformed to them by the OT) as part of their voltage gain. Tubes have a transconductance - for a small change in grid-cathode voltage, they induce a small change in the plate current. The voltage gain of a transconductance stage is the transconductance (di/dv) times the load resistance/impedance (dv/di) to make a pure gain, with no units. So a lower load on the output stage translates to a higher voltage gain in the output stage. Removing an attenuator, or even changing the setting, changes the loading, and that changes the forward voltage gain of the amplifier. With feedback connected, the overall gain from input to output is mostly determined by the feedback network, not by the variable-gain tubes and resistances. When you remove the negative feedback, the loading on the output stage has free reign to vary the forward gain; the higher open loop gain can work with the unintentional feedback to make it oscillate as the gain goes up.

[pdf64]

It's certainly a lower gain amp and I accidentally had the wrong speaker plugged in (16 ohm speaker on an 8 ohm tap). My bench attenuator is just an L pad. It was completely stable until the last notch on the dial. No idea why the attenuator would have that much effect.

the last notch on the dial

Is that nearly 0 attenuation (hence full power to the speaker) or nearly maximum attenuation (hence 0 power to the speaker)?

[maxkracht]

Is that nearly 0 attenuation (hence full power to the speaker) or nearly maximum attenuation (hence 0 power to the speaker)?

Full power to the speaker caused the squeal.

Amp is all boxed and ready to ship out so I can't test that. Measuring the attenuator didn't show much difference in resistance between where it squealed and where it didn't. It stays somewhere in the 14-18 ohm range through the full sweep (standard rotary l-pad doing its job). When measuring inductance, full attenuation is around 8uH @ 1kHz (the attenuator's series inductance) then it goes up fairly linearly to around 600uH, until the point at which the amp squealed, and it jumps to 1200uH (my speaker's inductance).

[pdf64]

Ok, the flip into oscillation may have been due to a largely resistive load (L pad) being switched (at its 0 attenuation setting) to a highly reactive load, ie speaker cab.
eg at bass resonance cab impedance might be 10x nominal.
Higher load impedance = higher output stage gain = sufficient gain to meet the requirements for oscillation.

[bepone]

I'm about finished rebuilding an amp and I forgot which way around the OT primary went, I also happened to have a NFB switch installed. Turn on the feedback loop and got a big volume boost. It is stable with the volume and tone control all the way up and actually sounded OK, until I turned my output attenuator all the way off and it squealed like crazy. I didn't know you could accidentally balance a positive feedback loop so well...

positive feedback sounds "lively".. it is possible to do yes

[maxkracht]

I have barebones 100w 8 ohm and a 16 ohm attenuators in a box with various dummy loads, mostly to avoid pissing off the neighbors while troubleshooting/testing things. I assumed a mostly resistive load was fine for this purpose, but I guess I found an exception. I always test on a real speaker, so I would have caught it eventually, but now I feel like I should upgrade my setup to avoid problems like this. Also makes me want to play with adding some controlled positive feedback to things. It really didn’t sound bad…