[Phil]

Sander deWaal wrote:
Phil said:


Well Andre, it looks like it's you and me, since everyone else has
either pussied out on me (Phil Asshole, Graham), dealt with other
issues, although probably honestly (Scott), or is waiting for further
developments (Patrick). I went by the UT library today, and looked
through several years of JAES, 1980 onward, but only found one article
by Matti Otala. I think maybe he published some papers in IEEE, I'll
have to check. However, in the Jan. 1980 JAES issue, there is an article
by Vanderkooy and Lip****z called "Feedforward Error Correction in Power
Amplifiers" that looks *very* interesting! I haven't thoroughly looked
at it yet, but they review all the various types of feedforward schemes,
starting with the one invented by Harold Black in 1923! They also review
the Quad 405, which is a different type of feedforward, which I can
state from personal experience sounds pretty bad compared to any decent
tube amp.



That you didn't find much after 1980, is because mr. Otala published
his findings in the early '70s.

At that time, the problems as described by him, were a reality in may
commercial amplifiers, and we've learned a lot since then.

Later, people like Daugherty and Greiner proved that (large factors
of) feedback isn't the evil that may seem to think it is, and that it
doesn't necessarily generate additional distortion, when applied
correctly.

I happen to think that Otala has played a major role in getting more
insight in what happens in an amplifier stage with feedback.

Others after him corrected and modified his findings.


There are more ways that lead to Rome.


BTW a modified 405-II can sound very good, at least to these ears.

The Quad I heard was not modified, so I couldn't say, although the one I
heard sounded very good unless compared to a good PP tube amp.

I want to emphasize that the original subject here is not what Phil
Allison implies, namely TIM or SID, but rather the question of whether
negative feedback causes audible problems even when there is no TIM.
I'll go ahead and quote the review of Otala's paper I gave in the other
thread, from "The Audio Critic," Vol 2, #2, p 37, regarding Matti
Otala's analysis of feedback (made after he, the editor Peter Aczel,
Mitch Cotter, Stew Hegeman, Andy Rappaport, Max Wilcox, and Bruce Zayde
had a "BS" session in TAC);

"The paper presents rigorous mathematical proof, for the most
generalized, all-inclusive case, that feedback cannot make amplifier
distortions go away; all it can do is to change one kind of distortion
into another. By the application of feedback, the amplitude
nonlinearities of the open loop are converted into phase nonlinearities
of the closed loop. That's all. The garbage cannot, by definition, be
made to disappear; it's simply swept into another corner. In the typical
feedback amplifier, the amplitude of the audio signal phase-modulates
the high-frequency components of the signal. Furthermore, any amplitude
intermodulation distortion in the open loop is converted into phase
intermodulation distortion in the closed loop. What about TIM, alias
SID? It turns out that it (he?) is a limit case of this
feedback-generated phase modulation effect, with all shades of gray
possible before the actual black eruption occurs. None of this shows up
on standard tests."

I still can't find this paper, despite several trips to the UT library,
but a little thought shows that it actually is consistent with much of
what you and the others ae saying. When an amp with, say, 40 dB of
feedback is hit with a step, the output initially has an "error" of 100
x, *independent* of any gain or load non-linearities, which must be
"corrected" by the feedback loop. For every single change in the input
voltage, the gain is off by a factor of 95 to 105, depending on gain and
load non-linearities, and this error must be corrected by the feedback
loop. Intuitively, it seems obvious that Otala's proof must in *some*
way be correct, that this constant "correction" must play havoc with low
level and high frequency signals. I don't think anyone would deny that,
given an amp with variable feedback followed by a pot to equalize the
overall gain, turning up the feedback will eventually make an amp that,
like the Crown preamp, will "bite your ears off," even if the amp never
gets into TIM territory or other obvious problems. The question is how
much of an effect does Otala's "dynamic phase shifting" have. Here
again, it seems obvious that part of the problem was the S-L-O-W power
transistors of the late '70's, when Otala's various articles were
written. I suspect that high speed devices reduce the problems created
by feedback, the amount of phase distortion produced, and of course
Otala himself came up with several ideas to reduce these effects in his
Citation XX design, although I also haven't been able to find any
literature on that design.

Nevertheless, it is a given, in my mind, that a very high open loop
gain, with its need to constantly "correct" every input signal by 99%
(in the case of 40 dB feedback), *regardless* of the inherent linearity
of the amp's devices and circuit, MUST cause problems for signals 60 dB
to 80 dB below the main signal, and perhaps also phase shift the high
frequency components, as Aczel's summary of Otala's paper states,
thereby robbing the circuit of much of its "life" and "air," the
criticisms one normally hears about high feedback amps, and also solid
state amps, in which the solid state capacitances and high thermal
variations also interfere with low level signals. This will not show up
as TIM or SID, unless the amp has been very poorly designed, and I'm
still not sure how one would measure it. My best guess has been to use a
20 Hz signal and a much smaller (-60 to -80 dB) 10 KHz signal, filter
out the 20 Hz signal from the output with a notch filter plus high pass
filter, and either look directly at the 10 KHz signal for signs of
distress, or filter it out with another notch filter, and see if phase
shifting causes "sidebands" to appear and disappear when the 20 Hz
signal is put in and out of the test.

Assuming that normal feedback causes problems -- and as Patrick says,
with low feedback and tubes it isn't too bad, but SS amps have more
problems and generally need more feedback -- it would be nice if we
could figure out a way to tremendously reduce the need for feedback to
"correct" every normal signal by 99% even when there is no device
distortion, meaning allow the feedback to "focus" *only* on actual
device and load non-linearities. Here is where Black's "feedforward"
circuit may allow for a real advance in SS amps, especially if tubes,
with their (generally) superior ability to handle a mix of high and low
level signals without messing up the low level information, are used to
provide the error signal. Properly applied, Black's feedforward scheme
(but not the feedforward designs by many others!) does exactly this, it
allows feedback to appear and affect the signal *only* when actual
deviations caused by device or load non-linearities appear. It may even
be possible to correct the effects of a typical transistor's parallel
capacitances -- which, being made of silicon, are of *very* poor quality
-- and thermal variations! Normally, this would be a "why waste the
time, just use tubes," situation, but good output transformers are
heavy, big, and expensive, and if the amps used in compact disk players,
as well as TV's, could be considerably improved, that would be nice! And
of course, inexpensive amps that sound very good are always in demand.

Phil