Feedback in audio esp wrt op-amps.

[Eeyore]

30 years of experience with bias-controlled listening tests says that
*seems* and *is* can be two different things. Intellectual logic has this
interesting tendency to rule, once the comparison is based on just
listening.


The fun begins when you level-match, time-synch and eliminate other
non-audible cues.

I'd love to see you level match the sound field from different loudspeakers !

Graham

 

[Arny Krueger]

I know it decreases overall THD numbers. I'm not one of those nuts who's
anti-NFB
per se.

What is the case AIUI is that NFB can create 'new' (higher) harmonics
that don't
exist with the open-loop situation. It's down to the maths of how feedback
works.

The actual fraction of new higher harmonic is very low in practice.

Take a really crappy power amp that has a 1% nonlinearity.

If you put in a FS 1 KHz tone, you get out a nearly FS 1KHz tone, DC that is
46 dB down, and 2 KHz that is 46 dB down. Run the output back through again,
and you get a nearly FS 1 KHz tone, DC that is still about 46 dB down, a 2
KHz tone that is about 46 dB down, and a 3 KHz tone that is about 92 dB
down.

1. Higher harmonics *are* more audible, but the additional 46 dB down is
far, far less than enough to make up the difference.

2. If the amp has 0.1 % nonlinearity (still fairly crappy by modern
standards), the numbers are 66 dB down for the DC and second harmonic, and
132 dB down for the third harmonic.

3. If the amp has 0.01 % nonlinearity (very good modern standards), the
numbers are 86 dB down for the DC and second harmonic, and 172 dB down for
the third harmonic.

Most modern power amps will be someplace between (2) and (3). Most modern
op amp applications will be closer to (3), on the good side.

 

[Mark]

I know it decreases overall THD numbers. I'm not one of those nuts who's anti-NFB
per se.

What is the case AIUI is that NFB can create 'new' (higher) harmonics that don't
exist with the open-loop situation. It's down to the maths of how feedback works.

Graham

And I am saying NFB CANNOT create new higher harmonics.

OK maybe what you are thinking aobut is this. If you take an
amplifier without feedback and overdrive it to clipping, it will
compress softly (2nd and 3rd order). If you take the same amplifier
and add NFB to it and now overdrive it to clippiong, it will remain
linear until it clips hard and this creates more high order harmoincs
comapred to the first case.
If you are talking about gutiar amp output stages where you are
overdriving into clipping intentionally, yeah feedback will make the
amp linear until it hard clips and lots of high order products are
created. An amp without feddback will overdrive with softer
compression and less higher harmonics.



But this is comparing apples and oranges. If you operate both below
clipping, the amplifer with neg feedback will have less overall
distortion and less high order distortion.


So are you talking about amps that are intentionally overdriven or
amps that are operated in their linear range?

Mark

 

[William Sommerwerck]

The source of this myth is the mistaken idea that negative feedback

regenerates the audio signal, and the nonlinearity of the amplifier
leads to higher order products of the regenerated harmonics and the
nonlinearity of the amp. This ignores the fact that the regenerated
signal is brought back in out-of-phase, and has the primary result
of reducing the high order harmonics.

I'm not sure that's right. My memory (which could be faulty) is that this
can be shown mathematically.

I'll ask around (I know a few people in high places) and see if I can get a
reference.

 

[William Sommerwerck]

What you say is intellectually logical, but it seems that post-recording

30 years of experience with bias-controlled listening tests says that
*seems* and *is* can be two different things. Intellectual logic has this
interesting tendency to rule, once the comparison is based on just
listening.

The fun begins when you level-match, time-synch and eliminate other
non-audible cues.

Arny, I was talking in general terms. The distortion in recordings does not
automatically mask distortions further down the playback chain.

 

[Bob Myers]

And I am saying NFB CANNOT create new higher harmonics.

I'd like to see what Graham is using re the "maths of
how feedback works."

The complete result for a negative-feedback amplifier,
using the configuration typically shown for an
"inverting amp" application of an op-amp (but which
may be generalized to any amplifier) winds up
looking like this (assuming high enough input
impedance at the amplifier such that current into
that path is negligible):

Vout/Vin = Rf/[Rin(A-1)]

where A is the open-loop gain of the amp and Rf and
Rin are the feedback and input resistors, respectively.
Let A get big enough, and this simplifies to the more
common

Vout/Vin = -Rf/Rin

....but I sure don't see anything in the above which
would *necessarily* result in additional harmonics.
To be sure, in the practical case, the open-loop
gain of the amplifier is non-linear, but even then
you can clearly create an amplifier employing
negative feedback which does NOT "create new
harmonics" to an appreciably greater degree than its
open-loop cousin.

As a side note, it's really, REALLY hard to find any
practical amplifier design which doesn't involve
SOME negative feedback, somewhere - although it
may not be as easy to spot as in this case.

Bob M.

 

[Arny Krueger]

I'd love to see you level match the sound field from different
loudspeakers!

Of course we can't do that with anything like the precision that is readily
obtainable with amps, preamps, converters, digital players and recorders,
etc. OTOH, it can be done well enough to be revealatory to many.

 

[Arny Krueger]

Arny, I was talking in general terms. The distortion in recordings does
not
automatically mask distortions further down the playback chain.

Since almost nothing in audio is automatic, that sounds a lot like an
excluded middle argument. ;-)

It is quite clear that the background noise in recordings is typically so
much greater than that in good equipment (other than microphones and rooms),
that it *does* mask the noise in much good equipment. True for good digital
recorders and players. True for many mic preamps, at least when used with
typical condenser mics.

The most common cause of audible distortion in audio gear is clipping due to
importune gain setting.

 

[Arny Krueger]

I'm not sure that's right. My memory (which could be faulty) is that this
can be shown mathematically.

I have just laid out the math results in two other posts. I've confirmed it
with simulations.

 

[Arny Krueger]

And I am saying NFB CANNOT create new higher harmonics.

But it can. However, if the NFB is working reasonably, the higher harmonics
are at vanishing low levels.

OK maybe what you are thinking aobut is this. If you take an
amplifier without feedback and overdrive it to clipping, it will
compress softly (2nd and 3rd order).

That avoiding NFB necessarily leads to soft clipping is a myth.

NFB does do a nice job of making such clipping as might happen have sharper
corners.

Tubes saturating can be a little soft. Transistors saturating or cutting off
give pretty sharp corners.

If you take the same amplifier
and add NFB to it and now overdrive it to clippiong, it will remain
linear until it clips hard and this creates more high order harmoincs
compared to the first case.

If the equipment is clipping, there are going to be plenty of higher
harmonics like 5,7 (presuming no P-P), even without NFB.

But this is comparing apples and oranges. If you operate both below
clipping, the amplifer with neg feedback will have less overall
distortion and less high order distortion.

Agreed.

 

[William Sommerwerck]

The most common cause of audible distortion in audio gear

is clipping due to importune gain setting.

Actually, clipping is not "in" the gear.

I do wish you'd glom onto a Crown K1. You really need to hear this
amplifier -- and run it through some blind tests.

 

[William Sommerwerck]

The complete result for a negative-feedback amplifier,

using the configuration typically shown for an
"inverting amp" application of an op-amp (but which
may be generalized to any amplifier) winds up
looking like this (assuming high enough input
impedance at the amplifier such that current into
that path is negligible):

Vout/Vin = Rf/[Rin(A-1)]

where A is the open-loop gain of the amp and Rf and
Rin are the feedback and input resistors, respectively.
Let A get big enough, and this simplifies to the more
common

Vout/Vin = -Rf/Rin

...but I sure don't see anything in the above which
would *necessarily* result in additional harmonics.

Because the mathematical model you're using doesn't include distortion!

To be sure, in the practical case, the open-loop
gain of the amplifier is non-linear, but even then
you can clearly create an amplifier employing
negative feedback which does NOT "create new
harmonics" to an appreciably greater degree than its
open-loop cousin.

How do you know that?

As a side note, it's really, REALLY hard to find any
practical amplifier design which doesn't involve
SOME negative feedback, somewhere - although it
may not be as easy to spot as in this case.

Many amplifiers include negative feedback, if only to stabilized local gain.

 

[Arny Krueger]

Actually, clipping is not "in" the gear.

Please explain.

I do wish you'd glom onto a Crown K1. You really need to hear this
amplifier -- and run it through some blind tests.

Your'e externalizing again William. It is you who need to run your K1 nd a
QSC USA 400 or Dyna ST-120 through some level-matched bias-controlled
listening tests.

 

[Scott Dorsey]

And I am saying NFB CANNOT create new higher harmonics.

Sure it can. Just put something that creates harmonics in the feedback
path.

Negative feedback relies on the feedback path being linear and having
low group delay. If these aren't the case, bad things can happen.
--scott

 

[Eeyore]

And I am saying NFB CANNOT create new higher harmonics.

And it seems you are incorrect (at least when the amplifier having the feedback applied
has some non-linearity).

Graham

 

[Eeyore]

Actually, clipping is not "in" the gear.

I do wish you'd glom onto a Crown K1. You really need to hear this
amplifier -- and run it through some blind tests.

What's the K1 like then ?

Graham

 

[Arny Krueger]

What's the K1 like then ?

Something like The Second Coming?

 

[Mark]

And it seems you are incorrect (at least when the amplifier having the feedback applied
has some non-linearity).

Graham

take something like crossover distortiuon for example...

in an open loop amp, crossover dist. creates lots of harmonics.

add neg feedback and they are all reduced. The high order ones are
not reduced AS MUCH as the low order ones, but they are certainly not
increased (assumming a proper design not on the verge of instability
and assuming the feedback componets themselves are linear, resistors
are usually linear for our purposes).

so after you add neg feedback the proportion of high order to low
order will change and realative to the low order there will be more
high order, but in absolute terms that are all reduced. Someone else
already said this so I am repeating...

I don't know how else to say it...
Mark

 

[Phil Allison]

"Scott Dorsey Notorious Charlatan"

It does, because a stage which is audibly blameless by itself may turn
into
a sonic disaster when it appears a few hundred times in the signal path.

** Huh ??

A few HUNDRED times ???????

The colossal fool must be on LSD.

"C'est un Nagra. C'est suisse, et tres, tres precis."

** C'est Dorsey - so you know it is total crapology.



......... Phil

 

[Eeyore]

take something like crossover distortiuon for example...

in an open loop amp, crossover dist. creates lots of harmonics.

add neg feedback and they are all reduced. The high order ones are
not reduced AS MUCH as the low order ones,

That's because of the falling loop gain with frequency of the amplifier. Not what I was
referring to.

but they are certainly not
increased (assumming a proper design not on the verge of instability
and assuming the feedback componets themselves are linear, resistors
are usually linear for our purposes).

so after you add neg feedback the proportion of high order to low
order will change and realative to the low order there will be more
high order, but in absolute terms that are all reduced. Someone else
already said this so I am repeating...

I don't know how else to say it...

You've missed the point I was making entirely. Other posters have explained it better than
myself however.

Graham