Amplifier Capacitance

Bombaywalla, sorry for my tardy response; obviously you got it right.

There is one missing element.

That is that the power supply has a timing constant. That is to say, there is a certain time period that will elapse if the power transformer is unplugged, where the voltage will sag to a certain point while the supply is under load.

Then there is a timing constant in the amplifier itself. This is the -3db point of the amplifier.

If the amplifier has a -3 db point that is a frequency lower than that of the timing constant of the power supply, then the amplifier can modulate the supply, which results in IM distortion amongst other things.

This is why an amplifier should never be direct-coupled from input to output! Otherwise, the amount of capacitance needed to get the timing constant of the supply low enough goes towards infinity.

This is why an amplifier can 'motorboat' (repeated thump) if a filter capacitor fails in the supply- the timing constant has become so high that the amplifier exhibits low frequency instability.

I apologize if I went a little too esoteric here, but there is obviously more to it than just inrush currents and the like. If something is not clear let me know.

Kijanki, I thought you might find this interesting. Years ago I worked on a popular tube preamp. I found that it was a copy of the Marantz 7 circuit, sans tone controls.

This preamp tended to have a problem- which was if you played a bassy track too loud, it could thump and do some weird stuff- especially if you were able to watch the woofer cones.

Anyway, it turned out that the timing constants of the preamp section went lower than that of its power supply. I spoke to the designer (a Brit) who was not at all happy to hear of my diagnosis, despite admitting to me seconds before that the bass instability was a problem.

The take-away is this is not an issue limited to amplifiers...

That's interesting. I would have thought the Roland stuff to go lower. We set our poles at 2 Hz so there would be no phase shift at 20Hz. We use servos too- in fact its part of our patent. It was a trick making them be stable with cutoff frequencies that low. One thing is sure- you can't use a servo to define the LF -3db point (the result is that every component in the servo circuit will have an audible artifact, and that is something a servo should never do).

If the rumble is going to be there, whether you have 2 Hz or 5 Hz cutoff is not a lot of difference as long as you have good LF stability. If there is instability rumble will drive you crazy!