sleepy man's question

[darkfenriz]

well it is quite late in Poland and maybe my question is stupid,but:

1.we may imagine frequency compensation as a first order (RC) filter, don't we?
2. Also output devices are limiting bandwidth (filtering) the same way, mainly because capacitance (as mosfets do)
3.also small signal devices do but on higher freqs.
4.for this filter : for 6*border.freq.:
attenuation is something like -17dB
phase shift is something like 82 degrees
for 0.3*border freq.:
attenuation is less than half a dB
phase shift is about 16 degrees
5.assume the same border freqs. for output devices and freq. compensatiom-just for simplicity, and higher border for VAS
6.we get -35dB attenuation for phase shift of 180 degrees
7.negative fb. becomes positive
8.many ab class have open loop gain over 35dB
9.how the hell do they stay stable???
you may also consider it using Nyquist's stability criterion (looking at hodogram)

[darkfenriz]

i've waken up couple of hours ago, drunk coffee and still have no idea

[MikeB]

To number 9:

That's what all the caps are good for, pushing down openloopgain
below 1 at this 180°-freq. That's why it is harder to stabilize large
feedback, but is nearly always possible.
When feedback get's positive, but gain is below 1, the oscillation
is damped to death.
I already made amp with OLG of 130db, was not too difficult to stabilize.
But is still bad idea, as the amp works with uV's, so any EMI get's
audible, unless you completely shield the amp...

Mike

[janneman]

Originally posted by darkfenriz
[snip]9.how the hell do they stay stable???
you may also consider it using Nyquist's stability criterion (looking at hodogram)

Well, your assumption that all these corner frequencies are more or less the same is not correct. The trick is to have one so called dominant pole (the name speaks for itself) that rolls off before any of the others come into play. Often this is done with that famous miller cap in the Vas stage.

So, you get amps with an open loop corner frequency of a couple of kHz. Look at opamps, they have extremely high open loop gain but often have a dominant pole coming into play below 100Hz!

Jan Didden

[darkfenriz]

oh thank you very much
now it seems much brighter to me
I guess thats one of reasons why we like very fast output devices
(much 'faster' than 20kHz)
I also think another solution is to filter OUTSIDE feedbackloop (like cap between collector of differential pair)
regards

[MikeB]

Originally posted by darkfenriz

I also think another solution is to filter OUTSIDE feedbackloop (like cap between collector of differential pair)
regards

This is still inside feedbackloop, but actually i do prefer a cap between
collectors of diffamp, but more like a RC. This seems far superior to
normals cdoms.

Mike