The Old "Feedback Is Bad" Lie
On Fri, 11 Jan 2008 21:22:06 -0800, isw wrote:
If you stick a signal that's beyond its bandwidth capability into an
amplifier, that amp will "outrun" its own feedback loop in an attempt to
handle the signal (i.e. some stage -- usually the input one -- will
either saturate or cut off before the correcting feedback can arrive to
prevent that). The result is distortion, but only until the feedback
*does* arrive; hence *Transient* Intermodulation Distortion.
The solution is to band-limit the signal *before* it gets to the amp, or
alternately, to use an amp with a large enough bandwidth.
In this kind of situation, an amplifier's bandwidth can vary with signal
level, so that an amp may perform fine for low-level signals, but not
for high-level ones.
There are three situations, in the modern world, where amplifiers
have to deal with high slew rate and/or possibly out of "band"
signals, but get no respect:
Phono equalizers are fed a rich diet of massively eq'd, and then
velocity-sensitively-reproduced impulses. Oh, yeah, they have to
play some music, also. Very tough gig.
Microphone preamps are another. Insanely large input dynamic range
and an expectation for similar performance at all input signal levels.
Easy to do if somebody else is signing the checks, but getting easier
over the years.
And, maybe the most relevant of all to Randy's OP: the D/A convertor
summing junction. Insanely large and fast switching impulses must
be (literally) integrated by a feedback amplifier of very little
or negative closed-loop gain (a condition that can only aggravate
the slewing issues of that integrating stage. It's a Beotch.
These before-any-possible-bandlimiting amplifiers are the fun 'uns,
and here the 1970's heads-up!s are still both relevant and often
overlooked. Engineering doesn't change, but it grows - like that.
Much thanks, as always,
Chris Hornbeck
"There's little that's impossible, but it becomes more complicated if
you move between different systems." - Mike Rivers, in another context
|