[Andre Jute[_2_]]

On Apr 4, 10:49*pm, mike s wrote:
On Apr 4, 5:56*pm, Andre Jute wrote:

The only thing I don't run high is the fils,
which I starve about 10% further to tilt the distortion artifacts.
This entire scheme is to make a very, very silent amp.

Starving filaments is interesting. *It's another of those things I
think about from time to time, but there are always other more
pressing things to do. *What made me wonder about it is that my preamp
uses EF37A *(Mullard version of the 6J7) which I still have quite a
lot of despite selling many on Ebay. *EF37A pentode was also sold as
ME1400 electrometer pentode. As ME1400 it was rated with a heater
voltage of 4.5V rather than the usual 6.3V but even as a triode the
gain is formidable so NFB would seem necessary :-)

If you haven't yet, you should find the site of Steve Bench and read
him. He's the main man on starved fils, and quite a bit besides.

I expect Black, as an engineer, would have thought us all mad by the
old engineering standard of -
"An engineer does for a shilling what a fool does for a pound".

Anyone worried about the obscene waste of building a tube amp to
listen to music while people starve in Africa is too sensitive to be
an audiophile.

Andre Jute
Visit Andre's books
http://coolmainpress.com/andrejute.html

[Paul G.]

On Sat, 02 Apr 2011 14:19:58 -0500, flipper wrote:

On 1 Apr 2011 15:23:04 -0500, Newbie wrote:

It seems well known that negative feedback can reduce lower order harmonics, but
increase the higher order ones.

The oft cited source of that observation is Baxandall's experiments
with a single stage FET amplifier and what he observed is that higher
order distortion is increased for 'small' values of feedback ('worst'
being about 6dB) but turn downward when more feedback is applied, with
it taking 'more' NFB the higher the harmonic order.

Note that in figure 7 he provides both observed and calculated
distortion under the given assumptions.

Conceptually he attributes this to the fed back harmonic(s)
intermodulating with the source and summing with the original
harmonic(s). I.E. in a simplified case a 'pure' input produces an
output with fundamental plus the 2'd (and more) harmonic so the
'feedback' now contains the 2'd (and more) harmonic which
intermodulates with the 'pure' input to produce the 3'd (and more)
harmonic.

Can anyone direct me to references that show how this happens with mathematical
analysis? Perhaps a paper in the journal of the AES, or elsewhere; on the web,
perhaps?

Take it from 'the horses mouth' or, at least, one of the well known
thoroughbreds

http://www.linearaudio.nl/Documents/...p%20design.pdf

I believe it's in issue 5 about 80% of the way down.

The assumption of generated odd harmonics in a "closed loop"
amplifier that should only generate even harmonics (when open loop) is
entirely reasonable. The math makes sense. The thing that puts the
whole excercise into meaningful proportion is "alpha" (from the paper)
which is roughly the level of 2nd harmonic distortion when open loop.
On open loop the output contains the 2nd harmonic, about 20 db
(alpha) below fundamental. On closed loop, the 2nd harmonic is reduced
further by feedback as you'd expect.
On closed loop the 3rd harmonic pops up, and is approximately alpha
(20db) below 2nd harmonic.
On the next modulation product (4th) is reduced by alpha again.
This process continues until bandwidth limitations come in.
Each harmonic is reduced approximately by "alpha", looking at the
ratios of harmonic amplitudes in Table 1 of "Amplifier Design -5".
The curves in figure 7 of "Amplifier Design -5" are based on alpha
of .1 (10% distortion of 2nd harmonic in open loop parabolic
amplifier). Similiar values are obtained with a simple single FET
amplifier. In "Amplifier Design -6" figures are obtained for a
junction transistor voltage input. The consecutive harmonic levels are
not spaced as far apart (depends on amount of drive).
In both cases modulation products rapidly decrease in amplitude
with increasing order.

I set up a LTSpice simulation for the JFET circuit that Baxandall
used, and followed his protocol. With no feedback, 3rd or greater
harmonics were less than 64db below fundamental. With small amounts of
feedback (5 db gain reduction, same 3 v output), they jumped up by
10-13 db, ie, -51db instead of -64db. That's still well below 2nd
harmonic (-28 db with feedback). Applying more feedback showed the
same behaviour as in the paper (each harmonic was about 17-20 db below
the previous one). The 2nd harmonics followed the classic reduction by
feedback (1/(1+BA), and the higher harmonics were reduced by "alpha"
for each one.

With a push-pull circuit, the reduction of the 2nd harmonic will
mean that the above effects are greatly reduced. "alpha" could 40-60
db down for each harmonic.

It's worth quoting the Baxandall paper's last paragraph.....quote:
"It can thus be concluded that providing plenty of feedback is assumed
right in the beginning, the more awkward parts of the theory outlined
in this article, though academically interesting, do not need to be
taken into account for design purposes"

If you can understand that quotation, then the whole issue is a
"tempest in a teapot". It's more a curiosity of how did the 2nd
harmonic distortion give rise to much smaller amounts of high order
distortion products when feedback is applied. It also gives you the
math and a table to calculate the harmonics in a closed loop
transistor amplifier.

To conclude from Baxnadall's paper:
1- feedback doesn't create distortion where there wasn't already some.
2- new harmonics are created from existing 2nd harmonic
3- the new harmonics are much smaller than each preceding harmonic
4- they are smaller by roughly the 2nd harmonic distortion figure
5- the effect is more pronounced in bipolar junction transistors than
FETs
6- reducing 2nd harmonic (ie, push-pull) will nearly eliminate the
effect

Paul G.

[Phil Allison[_3_]]

"Paul G."


To conclude from Baxnadall's paper:
1- feedback doesn't create distortion where there wasn't already some.
2- new harmonics are created from existing 2nd harmonic
3- the new harmonics are much smaller than each preceding harmonic
4- they are smaller by roughly the 2nd harmonic distortion figure
5- the effect is more pronounced in bipolar junction transistors than
FETs
6- reducing 2nd harmonic (ie, push-pull) will nearly eliminate the
effect


** Another way of summing up the article is to conclude that NFB should be
applied around an *amplifier stage* free of gross amplitude non-linearity
and with sufficient excess gain to obtain the result needed.

A square law mixer is simply not an amplifier.

Which is stating the bleeding obvious - really.

This article, from my friend Rod, covers the same territory and is worth a
read:

http://sound.westhost.com/articles/distortion+fb.htm


..... Phil

[Phil Allison[_3_]]

"Paul G."


To conclude from Baxnadall's paper:
1- feedback doesn't create distortion where there wasn't already some.
2- new harmonics are created from existing 2nd harmonic
3- the new harmonics are much smaller than each preceding harmonic
4- they are smaller by roughly the 2nd harmonic distortion figure
5- the effect is more pronounced in bipolar junction transistors than
FETs
6- reducing 2nd harmonic (ie, push-pull) will nearly eliminate the
effect


** Another way of summing up the article is to conclude that NFB should be
applied around an *amplifier stage* free of gross amplitude non-linearity
and with sufficient excess gain to obtain the result needed.

A square law mixer is simply not an amplifier.

Which is stating the bleeding obvious - really.

This article, from my friend Rod, covers the same territory and is worth a
read:

http://sound.westhost.com/articles/distortion+fb.htm



..... Phil

[Big Bad Bob]

On 04/04/11 15:38, flipper so wittily quipped:
What
you describe sounds like negative feedback CAUSING high harmonic
distortion, under ALL conditions.

It does, although, if you have 'enough' NFB then the increase in
(much) higher order distortion components, which are generally small
to begin with, is inconsequential.


The saving grace is that higher order harmonics manifest in ever
decreasing magnitude and eventually you reach the Cyrano Jones price
of Tribbles condition that "twice nothing is still nothing."

For those unfamiliar with that Star Trek episode, The Trouble with
Tribbles, the original offer was 1 credit and when the bartender
doubled it to 2 credits Cyrano made his well known quip.

Of course, Cyrano was not being 'mathematically correct', since 1 is
not "nothing" and neither is 2, but we understand his meaning: the
original offer was so low that doubling it was of no consequence (to
him anyway).

It's the same with NFB as long as there's 'enough' of it. OMG, you
doubled the -150dB n'th harmonic!!! Yes, well, twice nothing is still
nothing.

It is an 'increase' but we don't care about it.

ok, then that makes more sense. To suggest a theoretical increase
verses a measurable one. I can accept that.

Other cases, as described by Patrick, make sense (and these are driven
by specific design inadequacies, phase shift at high freq's and low open
loop gain at high freq's being the biggest contributing factors).

Of course they do but Baxandall's paper specifically excludes those
conditions because he was investigating NFB caused distortion and not
'lousy amplifier' caused distortion.

if someone tries to match measurements to his theory, he will probably
succeed, especially if he ignores the more practical aspects of science
in the process. In this case, apply NFB to a crappy amplifier, and
measure the "increase". Patrick's "crappy amplifier" description makes
a good case for that.

Incidentally it's a standard engineering practice to assume 5 time
constants as "discharging" a capacitor, even though a capacitor never
truly 'discharges'. It's really at 1/e^5 of the original potential but
there ya go. I suppose measuring those 'added' harmonics are the same
way. Maybe the theory says they're there but you won't be able to
measure any effect from them. With sufficient NFB and proper amp design
they'll be virtually undetectable.

[Big Bad Bob]

On 04/04/11 16:07, Andre Jute so wittily quipped:
What an offensive jerk you are, Bubba. WTF makes you think I need you
to tell me "the basics"?

there seemed to be a general misunderstanding of 'the basics' coming
from you, but NOW it seems you must already have achieved some level of
superiority over the rest of us. Well, I suppose not all of us can be
the next greatest genius in electronics engineering.

May your level of success directly reflect exactly what your skill
levels are. That's either a blessing or a curse, depending on how you
look at it.

as for me, I don't care which. Both extremes can be equally entertaining.

/me passing the bucket of popcorn to the right

[Big Bad Bob]

On 04/04/11 15:39, Andre Jute so wittily quipped:
On Apr 4, 9:19 pm, Big Bad BobBigBadBob-at-mrp3-
wrote:
On 04/02/11 05:16, Andre Jute so witilly quipped:

This article explains why added-on negative feedback is a bad thing in
tube audio if you have really good speakers; it focuses on how the
higher harmonic artifacts are added.
http://www.audio-talk.co.uk/fiultra/...dre%20Jute.htm

don't believe EVERYTHING you read on the intarweb

About half of what this guy says is pure opinion. I see no math to back
it up. basic 'servo theory' should easily disprove it.

That's a humorous article, Bubba. The math is elsewhere on the site.
But by all means bring on your math. We had a clown here called Henry
Pasternack who was last seen running away with his halls in one hand
and his MSEE from Stanford in the other. I sincerely hope you're
better prepared than he was.

bubba's your momma. who's your daddy?

[Big Bad Bob]

On 04/02/11 05:35, mike s so wittily quipped:
BTW, it is also "well known" that the heavier an object is, the faster
it falls. Of course, it's also untrue.

I like that one.

[Big Bad Bob]

On 04/02/11 20:57, flipper so wittily quipped:
And there is the article I once read from Wireless World, December,
1978. It has a graph of harmonic product levels for the amount of NFB
applied.

People seem to be having a kind of "how dare you insult my
girlfriend?" reaction but, according to Baxandall, it's just a matter
of getting 'enough' NFB.

and that's pretty much what I've been seeing/hearing for decades.

you could say the same thing about FM modulation, generating infinite
harmonics in both directions from the carrier at the modulation
frequency. Typically they're filtered out to avoid crossing over into
the next channel or exceeding licensed bandwidth [whichever]. And yet,
if you look on a spectrum analyzer, you won't see the infinite
harmonics. You'll only see the DETECTABLE ones. sure, the others are
"there" according to the theory, but they drop below the sea of chaos
known as 'background noise' and unless you're writing a demodulation
algorithm for cell phones, you can assume it's not there.

twice nothing IS nothing. I learned that in calculus class, but the
star trek reference is more fun.

You could also call this a religious war between mathemeticians and
engineers. Just don't mention accountants.

[Big Bad Bob]

On 04/04/11 18:10, Paul G. so wittily quipped:

To conclude from Baxnadall's paper:
1- feedback doesn't create distortion where there wasn't already some.

that makes the most sense yet.

[Big Bad Bob]

On 04/04/11 16:27, Andre Jute so wittily quipped:
Anyone worried about the obscene waste of building a tube amp to
listen to music while people starve in Africa is too sensitive to be
an audiophile.

sensitivity is overrated anyway. best to abandon it entirely.

[Big Bad Bob]

On 04/03/11 00:09, Don Pearce so wittily quipped:
A small amount of negative feedback (e.g. 6dB) in a single ended
stage, though reducing the second-harmonic distortion, and also the
total (unweighted) distortion, by about 6dB, will increase the
higher-order distortion and the quality of reproduction may well
become worse as judged subjectively.

hmmm... so the test conditions are specific, and not general (as was
originally implied). My argument all along has been that a high gain
amplifier with a large amount of negative feedback would not exhibit
this particular problem due to its operation being closer to that of an
operation amplifier, assuming no phase shift or stability issues.

The magnitude of harmonics of extremely high order will be increased
by the application of negative feedback, no matter what practical
amount of negative feedback is employed, but this is of no consequence
if, when thus increased, they are, say, 120dB below the fundamental.

thanks for that part. It is the common sense application of the theory
where NFB isn't generically bad any more. Now it does what I think all
of use have assumed it would do, to bring distortion down to a 'below
detectability' level.

He also explains the mechanism producing the 'new' distortion as the
modulator/mixer action of the non-linear amplifier, which we know is
non-linear or else we would not be wrapping NFB around the thing
trying to 'improve' it's linearity, acting on the fed back
harmonic(s).

that would be IM distortion, and that's very very bad for the reason
stated here.

Let me put this as few words as the subject will allow. Global NFB is
capable - through intermod - of generating tiny amounts of higher
harmonics where none existed. And I do mean tiny. These higher
harmonics are, in practice always swamped massively by naturally
occurring higher harmonics. These higher harmonics respond normally to
the application of NFB and are reduced proportionately to the ratio of
open loop to closed loop gain.

agreed.

So no, NFB does not cause higher harmonics to increase.

also agreed, to the extent that some other 'bad design' factor like
clipping below max power or drawing current on a grid or using a sharp
cutoff pentode instead of triodes in the pre-amp or using a B+ voltage
that isn't significantly higher than expected voltage swing on the
preamp tubes doesn't throw a wrench into the works. Those other things
might make the original premise correct, sorta like what Patrick said
early on.

[Big Bad Bob]

On 04/04/11 14:49, mike s so wittily quipped:
Starving filaments is interesting. It's another of those things I
think about from time to time, but there are always other more
pressing things to do.

I would think that 'use the right component for the job' might be a
better way to go, rather than forcibly de-rating the device by starving
the filament. It's also more likely to give you unpredictable results,
or require tweaking each unit. Tube characteristics change over time,
after all, though I suppose lower heater voltage might slow that process
down a bit.

Now, something else comes to mind with reduced heater potential, and
that is a cooler cathode, which for a triode might reduce the noise
level. If you can squeeze out enough current to operate properly, the
reduced cathode temp might produce less entropy in the electron cloud.
That might result in lower noise, so for a 1st stage in a preamp it
might be a good idea on that basis. It's worth an experiment. But then
again, 'right component for the job' is probably the better choice.

But reducing heater current on the power output stage? I think I'd shop
for a better tube instead. Maybe one of the high end tube makers would
spec one out specifically for low distortion power amps.

[Patrick Turner]

On Apr 3, 1:57*pm, flipper wrote:
On Sat, 2 Apr 2011 20:08:50 -0700 (PDT), Patrick Turner





wrote:
On Apr 3, 5:34 am, flipper wrote:
On Sat, 2 Apr 2011 05:35:07 -0700 (PDT), mike s

wrote:
On Apr 1, 9:23 pm, Newbie wrote:
It seems well known that negative feedback can reduce lower order harmonics, but
increase the higher order ones.

Can anyone direct me to references that show how this happens with mathematical
analysis? Perhaps a paper in the journal of the AES, or elsewhere; on the web,
perhaps?

In most things it pays to start at the beginning and proceed as far as
you are interested or able -

http://www.ece.ucdavis.edu/courses/n...B/feedback.pdf

BTW, it is also "well known" that the heavier an object is, the faster
it falls. Of course, it's also untrue.

And, apparently, some people don't believe in gravity at all.

NFB does, indeed, increase higher order harmonics, depending on the
order of the harmonic and how much NFB is applied.

Take it from 'the horses mouth' or, at least, one of the well known
thoroughbreds

http://www.linearaudio.nl/Documents/...p%20design.pdf

This adress took me to a 3.3M .pdf file which has an enormous wealth
of useful information.

There is a 'page 56' about 80% down the scroll....

Yep, that's it.



And there is the article I once read from Wireless World, December,
1978. It has a graph of harmonic product levels for the amount of NFB
applied.

People seem to be having a kind of "how dare you insult my
girlfriend?" reaction but, according to Baxandall, it's just a matter
of getting 'enough' NFB.

Ah, Once Upon A Time, when I was about 26, I rooted my girlfriend so
often she walked about bow legged and cross eyed in a delerious
romantic swoon from one day to the next.

Neither of us could get enough of each other. NFB didn't get thought
about very much. But after the wedding
it sure was considered, and enough of lots of things were also
considered. After the divorce the gettering within both of us kinda
faded, and we couldn't get enough.......

But between then and now, girls of 62 have become sad old chooks and
won't ****, can't ****, and since I am a fiddle with many a tune still
able to be played if there was a girl who could use a bow, then the
old chooks can't look at me, and think a BJ is a Blue Jumper, and a
bicycle doesn't have enough "class", and makes an old man look
distortedly young, and strange.

I happen to think the business of enough NFB is a fair enough Idea,
providing the amp is fairly linear, quiet, and has sufficient BW and
can be listened to without displeasure before FB is applied.
Adding NFB usually vastly improves bass and creates low output
resistance which ensures the speaker crossover filters all work as
intended and that the response level is flatter than if the Rout was
higher than say 1 ohm.

To me, the NFB enables the tubes to display their grandure with music.
If you want good music, or good sex, you must bring the performer in
from the street where its cold and windy, and have them relax and
focus while comfortable, in the absense of noise and distractions. NFB
kinda takes care of things.

But many of my customers don't mind tube amps without a stitch of or
global NFB. All these amps are triode amps. Myself, I prefer the
shiela without make-up, and wearing only a birthday suit, au-natural.
No need for lingerie. But for music some NFB is OK.

At 63, at my favourite cafe the waitress of 20 rushes over to me with
a smile and brings me edible and drinkable treats which are my humble
idea of what is enough. She outshines the others among the cafes
around the square, and if she was a vacuum tube, she'd be a little
triode that was musical, engaging, warm, smooth, erotic, emotional,
practical, and vivacious and plain damn desirable. When I have
finished and paid I say "I'll catch you later..." Trouble is I am 30
years too late already. Nobody can be young enough. If I was a vacuum
tube I'd probably be big tired old 845, but capable of sweeping a girl
off her feet in her dreams if only I was directly coupled to a good
little triode. A good butler called Mr NFB would make the experience
better if he worked in the background to take care of details which
romance forgets.

Patrick Turner.

[Patrick Turner]

On Apr 3, 9:04*pm, mike s wrote:
On Apr 2, 1:16*pm, Andre Jute wrote:





On Apr 1, 9:23*pm, Newbie wrote:

It seems well known that negative feedback can reduce lower order harmonics, but
increase the higher order ones.

Can anyone direct me to references that show how this happens with mathematical
analysis? *Perhaps a paper in the journal of the AES, or elsewhere; on the web,
perhaps?

Here's an article to help you distinguish between added-on negative
feedback and natural negative feedback.http://www.audio-talk.co.uk/fiultra/...dre%20Jute.htm

This article explains why added-on negative feedback is a bad thing in
tube audio if you have really good speakers; it focuses on how the
higher harmonic artifacts are added.http://www.audio-talk.co.uk/fiultra/...dre%20Jute.htm

More articles on Harmonic Distortion on the KISS Amps site, available
through those pages.

This in intriguing in that what Jute has re-invented or re-created is
pretty much exactly the type of amplifier that Harold Black was faced
with in the 1920s when he invented negative feedback. *So it would be
interesting to hear from someone with such an amplifier as to what
happens if negative feedback is employed.

Certainly well designed and built valve amplifiers can perform
superbly as domestic audio systems without negative feedback. *Though
building two with exactly matched characteristics for stereo can be a
(not unrewarding) challenge. *Of course before the signal reaches the
domestic system it's been through many other higher performance
amplifiers, all of which will employ NFB.

My own system has no global NFB in the output stage, but it uses push
pull triodes, driven by another pair of triodes, so it's hardly worth
it. *And I like the "traditional" sound.http://mike.wepoco.com/Home/retro-ge...ers/Building-a...


What exactly was "traditional" sound?

In 1925, reproduced sound was mostly pretty damn awful mainly because
transducers such as microphones and speakers and recording discs were
so woefully limited in their ability to create hi-fi. Triode
amplifiers could always be easily made without global NFB and to be
able to perform flawlessly compared the other crappy gear which was
used.

But now we have had hi-fi discs since vinyl days. The mics and
speakers and record cutters, FM transmiters, and digital technology
etc have quite low distortions, so the amp needs to perform better
than the worst generic crappy tube stuff which was used in 1925. There
were crappy triode amps. The WE movie theatre amps were not so hot
IMHO.

There has always been plenty of damn awful tubed audio electronics
available at a cheap price for those who wouldn't know hi-fi even it
it bit them on the arse.

Patrick Turner.

[Ian Bell[_2_]]

flipper wrote:

I already tried 'argument', including quoting his exact words, but you
are immune to a rational discussion, and I see no reason to re type
the entire paper when the original is just sitting there for you to
read,

I have not read this paper. Can you provide me with a link to it please?

Cheers

Ian

[Paul G.]

On Mon, 04 Apr 2011 20:57:58 -0500, flipper wrote:

On Mon, 04 Apr 2011 22:10:11 -0300, Paul G.
wrote:
......[snip!]


With a push-pull circuit, the reduction of the 2nd harmonic will
mean that the above effects are greatly reduced. "alpha" could 40-60
db down for each harmonic.

Modulation effects from the second will certainly be reduced but
harmonics from the rest, usually dominated by the 3'd, will do the
same thing.


It's worth quoting the Baxandall paper's last paragraph.....quote:
"It can thus be concluded that providing plenty of feedback is assumed
right in the beginning, the more awkward parts of the theory outlined
in this article, though academically interesting, do not need to be
taken into account for design purposes"

Yes, I've explained the 'enough' NFB point till I'm blue in the face.

If you can understand that quotation, then the whole issue is a
"tempest in a teapot"

Not entirely because people in this forum don't always employ 'tons of
NFB."

In fact, I think in the tube genre there's a following who believes in
'modest' (global) NFB, which Baxandall's paper suggests might not be
as good a thing as they imagine. On the other hand it might be because
they're often thinking about clipping effects and Baxandall wasn't
addressing that issue.

. It's more a curiosity of how did the 2nd
harmonic distortion give rise to much smaller amounts of high order
distortion products when feedback is applied. It also gives you the
math and a table to calculate the harmonics in a closed loop
transistor amplifier.

To conclude from Baxnadall's paper:
1- feedback doesn't create distortion where there wasn't already some.

A rather moot point since there's no such thing as a distortionless
amplifier and if there were you wouldn't using NFB to reduce what
wasn't there.

2- new harmonics are created from existing 2nd harmonic

There's nothing 'special' about the 2'd that make it 'uniquely' a
modulator and new harmonics will be created from any existing
harmonic. 2'd just happens to be the first and largest produced by a
single square law device (his FET approximating)

3- the new harmonics are much smaller than each preceding harmonic
4- they are smaller by roughly the 2nd harmonic distortion figure

Depends on which harmonic is doing the modulation.

5- the effect is more pronounced in bipolar junction transistors than
FETs
6- reducing 2nd harmonic (ie, push-pull) will nearly eliminate the
effect

It'll reduce 2'd harmonic modulation but not the 3'd, which will then
modulate when fed back. They all will.


The issue is now whether with push-pull or symetric distortion (all
odd products ie., 3rd,5th,7th, etc) will have the same problem of
feedback increasing the amplitude of upper harmonics.
I would say "no" because the harmonic that would be fed back would
be the 3rd, and that would be modulated by the fundamental to become
the 4th harmonic. The modulation mechanism is the distortion itself.
Now the 4th harmonic won't get through the amplifier because of the
symmetry - it's push-pull.
To verify this, I made up a circuit in LTSpice in which an ideal
opamp drove a circuit with 2 diodes and a few resistors so the
clamping wan't too extreme. The diodes (part of the shunt to ground)
were in opposite polarity to give symmetry. I measured the distortion
products, then wrapped an opamp around the circuit, and varied the
internal gain of the opamp, and watched tha amplitude of the
harmonics. In all cases I forced the output signal to stay at a
constant level.
The results of the simulation showed no even harmonics under any
condition. The ratios of the harmonics to each other was about 5db
less for each consecutive odd harmonic (3rd..5th..7th..etc) when open
loop. With small amounts of feedback, the ratios varied between 3-10
db. With larger amounts of feedback the ratios varied between 3-7 db,
with most of the variation occuring between 3rd and 5th harmonics.
The absolute values of the harmonics roughly decreased in accordance
with the feedback (loopgain).
Feedback is decreasing the distortion (all harmonics), but there is
variation with respect to each harmonic. For an overall decrease of
20db in all distortion, the individual harmonics will vary plus-minus
3-4 db with respect to each other.
Some of the variation will be from the FFT, I chose a signal
frequency of 1 KHz, and on the time domain simulation I forced a max
timestep of .2us in order to give a reasonable results.
The opamp I used was the "Universal Opamp", and I set GBW to
100MHz. That would cause frequency response issues for open loop gains
greater than 10000, or very large amounts of feedback. The result of
limited frequency response (GBW=100MHz is a VERY fast audio amp) gave
a limit to the distortion reduction (60-70db) for large amounts of
feedback.

The verdict? Feedback still works..... for small amounts of feedback
the ratios of the harmonics vary with respect to each other, but much
less than what the overall reduction from feedback does.

Paul G.

[Big Bad Bob]

On 04/05/11 03:24, Patrick Turner so wittily quipped:
On Apr 3, 9:04 pm, mike wrote:
My own system has no global NFB in the output stage, but it uses push
pull triodes, driven by another pair of triodes, so it's hardly worth
it. And I like the "traditional" sound.http://mike.wepoco.com/Home/retro-ge...ers/Building-a...


What exactly was "traditional" sound?

that's a very good question. gramophone maybe? "his master's voice"
indeed.

There has always been plenty of damn awful tubed audio electronics
available at a cheap price for those who wouldn't know hi-fi even it
it bit them on the arse.

I refurbed one of those 'damned awful tube audio' units back in the
70's, 'cause it had a reasonably nice wood cabinet. The phono used a
crystal cartridge (and diamond a needle, but it had trouble with 1970's
vinyl), the amplifier had an open loop SE (6BQ5 or similar) with output
transformers that saturated due to current flowing in one direction for
too long. I reversed polarity on them to compensate for the hysteresis
problem, added a bit of negative feedback (not too much since gain
sucked) to improve freq. response a bit, then later changed it out for
an external 10W/channel tube power amp that I found at the local swap
meet and fixed up a bit. It used better components and sounded pretty
good. Oh, yeah, I built matching speakers for it too. I still have
those speaker cabinets, with newer speakers, better [Peavey style] grill
cloth, modern components.

[Big Bad Bob]

On 04/05/11 11:51, flipper so wittily quipped:
You could also call this a religious war between mathemeticians and
engineers. Just don't mention accountants.

No, it's an 'argument' created by people who don't read the paper, nor
what's written about it, and misrepresent what's said... which is a
neat trick seeing as how they haven't read it.

I actually didn't read it myself, except for the quotes that were posted
here [who has that much time, right?]. But the argument being made
sounded so ridiculous that I rejected it outright based on my own
experience and a few mental calculations.

Also picked up a couple of trolls. that's always fun.

[Big Bad Bob]

On 04/05/11 19:29, Paul G. so wittily quipped:
The issue is now whether with push-pull or symetric distortion (all
odd products ie., 3rd,5th,7th, etc) will have the same problem of
feedback increasing the amplitude of upper harmonics.
I would say "no" because the harmonic that would be fed back would
be the 3rd, and that would be modulated by the fundamental to become
the 4th harmonic. The modulation mechanism is the distortion itself.
Now the 4th harmonic won't get through the amplifier because of the
symmetry - it's push-pull.

well, that's an interesting point, but the even harmonic cancellation is
for ADDED distortion, and if the IM distortion (creating the 4th
harmonic) is exactly the same for both legs, well then yeah. It cancels
out. But the real point is that the amplitude of the 'fed back'
distortion is so small as to be negligible, and inverted with respect to
the 'generated' distortion. The IM components that are of concern are
between the odd harmonics and one another, and the rest of the audio
(usually not a pure tone).

But of course large NFB brings these levels to "below detectable" which
is essentially zero. And that was the question, whether NFB is always
resopnsible for more harmonic distortion at the higher harmonics, or
whether it does what it claims to do, reducing THD and not creating more.

To verify this, I made up a circuit in LTSpice in which an ideal
opamp drove a circuit with 2 diodes and a few resistors so the
clamping wan't too extreme. The diodes (part of the shunt to ground)
were in opposite polarity to give symmetry. I measured the distortion
products, then wrapped an opamp around the circuit, and varied the
internal gain of the opamp, and watched tha amplitude of the
harmonics. In all cases I forced the output signal to stay at a
constant level.
The results of the simulation showed no even harmonics under any
condition.

that's actually very interesting. Can you inject multiple tones and see
what THAT does?

Feedback is decreasing the distortion (all harmonics), but there is
variation with respect to each harmonic. For an overall decrease of
20db in all distortion, the individual harmonics will vary plus-minus
3-4 db with respect to each other.

real world vs theory maybe? Or just an artifact of the nature of
harmonic distortion, because the harmonics aren't necessarily in phase
with the original tone. In fact they probably aren't.


Some of the variation will be from the FFT, I chose a signal
frequency of 1 KHz, and on the time domain simulation I forced a max
timestep of .2us in order to give a reasonable results.

FFT's are fine, so long as your sample width is a power of 2. BTW I
have a nice multi-thread DFT algorithm that you can use if you want.
Works great on multi-core machines. Each harmonic is calculated as a
separate work unit, sort of like those distributed projects like dnetc
or seti@home. I suppose anyone could code it now that I gave my
'secret' away...

The verdict? Feedback still works..... for small amounts of feedback
the ratios of the harmonics vary with respect to each other, but much
less than what the overall reduction from feedback does.

thanks for experimental confirmation.

[Big Bad Bob]

On 04/06/11 16:44, flipper so wittily quipped:
I actually didn't read it myself, except for the quotes that were posted
here [who has that much time, right?]. But the argument being made
sounded so ridiculous that I rejected it outright based on my own
experience and a few mental calculations.

And what argument is that?

heh, did I leave out too much information?

going back to the beginning, it was being (incorrectly) argued that NFB
always increased higher level harmonic distortion. Has the thread
gotten so 'deep' that the original post has been forgotten? Not
surprising really, what with trolling and self-promotion from a couple
of posters and common sense from the rest.

it's beer-o-clock anyway and I still need to get work done. My boss
(me) is a real slavedriver and keeps yelling at me for wasting time on
USENET instead of working.

[Big Bad Bob]

On 04/05/11 03:05, Patrick Turner so wittily quipped:
People seem to be having a kind of "how dare you insult my
girlfriend?" reaction but, according to Baxandall, it's just a matter
of getting 'enough' NFB.

Ah, Once Upon A Time, when I was about 26, I rooted my girlfriend so
often she walked about bow legged and cross eyed in a delerious
romantic swoon from one day to the next.

Neither of us could get enough of each other. NFB didn't get thought
about very much. But after the wedding
it sure was considered, and enough of lots of things were also
considered. After the divorce the gettering within both of us kinda
faded, and we couldn't get enough.......

But between then and now, girls of 62 have become sad old chooks and
won't ****, can't ****, and since I am a fiddle with many a tune still
able to be played if there was a girl who could use a bow, then the
old chooks can't look at me, and think a BJ is a Blue Jumper, and a
bicycle doesn't have enough "class", and makes an old man look
distortedly young, and strange.

I happen to think the business of enough NFB is a fair enough Idea,
providing the amp is fairly linear, quiet, and has sufficient BW and
can be listened to without displeasure before FB is applied.
Adding NFB usually vastly improves bass and creates low output
resistance which ensures the speaker crossover filters all work as
intended and that the response level is flatter than if the Rout was
higher than say 1 ohm.

To me, the NFB enables the tubes to display their grandure with music.
If you want good music, or good sex, you must bring the performer in
from the street where its cold and windy, and have them relax and
focus while comfortable, in the absense of noise and distractions. NFB
kinda takes care of things.

But many of my customers don't mind tube amps without a stitch of or
global NFB. All these amps are triode amps. Myself, I prefer the
shiela without make-up, and wearing only a birthday suit, au-natural.
No need for lingerie. But for music some NFB is OK.

At 63, at my favourite cafe the waitress of 20 rushes over to me with
a smile and brings me edible and drinkable treats which are my humble
idea of what is enough. She outshines the others among the cafes
around the square, and if she was a vacuum tube, she'd be a little
triode that was musical, engaging, warm, smooth, erotic, emotional,
practical, and vivacious and plain damn desirable. When I have
finished and paid I say "I'll catch you later..." Trouble is I am 30
years too late already. Nobody can be young enough. If I was a vacuum
tube I'd probably be big tired old 845, but capable of sweeping a girl
off her feet in her dreams if only I was directly coupled to a good
little triode. A good butler called Mr NFB would make the experience
better if he worked in the background to take care of details which
romance forgets.

heh, full quote simply because it was a great read

[Big Bad Bob]

On 04/05/11 11:23, flipper so wittily quipped:
Let me put this as few words as the subject will allow. Global NFB is
capable - through intermod - of generating tiny amounts of higher
harmonics where none existed. And I do mean tiny. These higher
harmonics are, in practice always swamped massively by naturally
occurring higher harmonics. These higher harmonics respond normally to
the application of NFB and are reduced proportionately to the ratio of
open loop to closed loop gain.

agreed.

Except what he wrote is wrong, as the measured data in the paper
unquestionably shows.

that would be the theory, actually, and mathematically you could prove
something like that happening.

So no, NFB does not cause higher harmonics to increase.

also agreed,

Also wrong, as the measured data in the paper unquestionably shows.

should qualify as 'above the detectability point' but I thought that was
implied.


I think you must be a myers-briggs 'SJ' (guess ESTJ) type, where I'm a
'ENTP' type. You appear to see things in terms of details and direct
observations, whereas I look at the general case based on real world
experience and my understanding of the theories behind it all (besides
the details being boring). Problem is I see why you make these
statements and I'm just as likely to agree with you as I am the other
side [from your perspective]. And so I look back at every NFB audio
circuit I've ever made, and what the o-scope traces looked like, and how
they performed, and things of that nature, and intuitively understand
that yes, NFB [when done right] does NOT increase any harmonic
distortion, at least not in a way that's detectable.


In any case, the elephant really isn't like a tree, he's like a snake.
Or not. Heh.

In the real world, ESTJ's pair up nicely with ENTP's as part of an
engineering team because input from each perspective helps make a better
design. ENTPs are mad scientists, while ESTJs are bit fiddlers. An
ENTP will keep an ESTJ from running in an infinite loop to solve an
unnecessary problem [such as refactoring computer code 3 dozen times or
moving components around in circles on a circuit board], wherease an
ESTJ will keep an ENTP from doing something that generally violates the
laws of physics or doesn't coincide with the goals of the department.

[Big Bad Bob]

On 04/05/11 10:35, flipper so wittily quipped:
Incidentally it's a standard engineering practice to assume 5 time
constants as "discharging" a capacitor, even though a capacitor never
truly 'discharges'. It's really at 1/e^5 of the original potential but
there ya go.

Really? "Really" it is? Is that a 'theory' or a crappy capacitor?

a crappy capacitor would discharge itself due to poor insulation. but I
digress.

I suppose measuring those 'added' harmonics are the same
way. Maybe the theory says they're there but you won't be able to
measure any effect from them. With sufficient NFB and proper amp design
they'll be virtually undetectable.

If you had read the paper you'd know his whole point is to use
'enough' NFB.

details, details. That was covered in a different post anyway. It's
back to theory vs real world vs measureable vs what really matters. I
thought I was posting in English but I guess not. Anyway the day's
getting too long and I need more caffeine. Or alcohol. Or both.

[Big Bad Bob]

On 04/06/11 17:24, flipper so wittily quipped:
going back to the beginning, it was being (incorrectly) argued that NFB
always increased higher level harmonic distortion.

It does, in Baxandall's FET experiment with 'how high' the order
depending on the amount of NFB.

ok - I'll split the difference with ya on this one. 'ALWAYS' vs
'depending upon'. Your statement counteracts the 'ALWAYS', as in "find
one counterexample and 'always' no longer applies." That's what I
implied by the 'always' part. So yeah, the theory would suggest that
you get an increase [based on the experiment] but with sufficient NFB it
drops below the point where you can detect it. Or something like that.
The math would suggest an infinitesimal increase in a higher harmonic
somewhere 'out there' regardless, but the reality would show it to be so
small as to become like the star trek 'twice nothing is still nothing'
example.

I'm out of breath. did I miss anything? No, don't say it (I know
you're thinking it) or we'll end up in a causality loop.

[Patrick Turner]

On Apr 6, 12:29*pm, Paul G. wrote:
On Mon, 04 Apr 2011 20:57:58 -0500, flipper wrote:
On Mon, 04 Apr 2011 22:10:11 -0300, Paul G.
wrote:

.....[snip!]







* With a push-pull circuit, the reduction of the 2nd harmonic will
mean that the above effects are greatly reduced. "alpha" could 40-60
db down for each harmonic.

Modulation effects from the second will certainly be reduced but
harmonics from the rest, usually dominated by the 3'd, will do the
same thing.

* It's worth quoting the Baxandall paper's last paragraph.....quote:
"It can thus be concluded that providing plenty of feedback is assumed
right in the beginning, the more awkward parts of the theory outlined
in this article, though academically interesting, do not need to be
taken into account for design purposes"

Yes, I've explained the 'enough' NFB point till I'm blue in the face.

* If you can understand that quotation, then the whole issue is a
"tempest in a teapot"

Not entirely because people in this forum don't always employ 'tons of
NFB."

In fact, I think in the tube genre there's a following who believes in
'modest' (global) NFB, which Baxandall's paper suggests might not be
as good a thing as they imagine. On the other hand it might be because
they're often thinking about clipping effects and Baxandall wasn't
addressing that issue.

. It's more a curiosity of how did the 2nd
harmonic distortion give rise to much smaller amounts of high order
distortion products when feedback is applied. It also gives you the
math and a table to calculate the harmonics in a closed loop
transistor amplifier.

*To conclude from Baxnadall's paper:
1- feedback doesn't create distortion where there wasn't already some.

A rather moot point since there's no such thing as a distortionless
amplifier and if there were you wouldn't using NFB to reduce what
wasn't there.

2- new harmonics are created from existing 2nd harmonic

There's nothing 'special' about the 2'd that make it 'uniquely' a
modulator and new harmonics will be created from any existing
harmonic. 2'd just happens to be the first and largest produced by a
single square law device (his FET approximating)

3- the new harmonics are much smaller than each preceding harmonic
4- they are smaller by roughly the 2nd harmonic distortion figure

Depends on which harmonic is doing the modulation.

5- the effect is more pronounced in bipolar junction transistors than
FETs
6- reducing 2nd harmonic (ie, push-pull) will nearly eliminate the
effect

It'll reduce 2'd harmonic modulation but not the 3'd, which will then
modulate when fed back. They all will.

* *The issue is now whether with push-pull or symetric distortion (all
odd products ie., 3rd,5th,7th, etc) will have the same problem of
feedback increasing the amplitude of upper harmonics.
* *I would say "no" because the harmonic that would be fed back would
be the 3rd, and that would be modulated by the fundamental to become
the 4th harmonic. The modulation mechanism is the distortion itself.
Now the 4th harmonic won't get through the amplifier because of the
symmetry - it's push-pull.
* *To verify this, I made up a circuit in LTSpice in which an ideal
opamp drove a circuit with 2 diodes and a few resistors so the
clamping wan't too extreme. The diodes (part of the shunt to ground)
were in opposite polarity to give symmetry. I measured the distortion
products, then wrapped an opamp around the circuit, and varied the
internal gain of the opamp, and watched tha amplitude of the
harmonics. In all cases I forced the output signal to stay at a
constant level.
* *The results of the simulation showed no even harmonics under any
condition. The ratios of the harmonics to each other was about 5db
less for each consecutive odd harmonic (3rd..5th..7th..etc) when open
loop. With small amounts of feedback, the ratios varied between 3-10
db. With larger amounts of feedback the ratios varied between 3-7 db,
with most of the variation occuring between 3rd and 5th harmonics.
The absolute values of the harmonics roughly decreased in accordance
with the feedback (loopgain).
* *Feedback is decreasing the distortion (all harmonics), but there is
variation with respect to each harmonic. For an overall decrease of
20db in all distortion, the individual harmonics will vary plus-minus
3-4 db with respect to each other.
* *Some of the variation will be from the FFT, I chose a signal
frequency of 1 KHz, and on the time domain simulation I forced a max
timestep of .2us in order to give a reasonable results.
* *The opamp I used was the "Universal Opamp", and I set GBW to
100MHz. That would cause frequency response issues for open loop gains
greater than 10000, or very large amounts of feedback. The result of
limited frequency response (GBW=100MHz is a VERY fast audio amp) gave
a limit to the distortion reduction (60-70db) for large amounts of
feedback.

* The verdict? Feedback still works..... for small amounts of feedback
the ratios of the harmonics vary *with respect to each other, but much
less than what the overall reduction from feedback does.

Paul G.- Hide quoted text -

- Show quoted text -

I think It may have been Ian Hinkman or some name like that who re-
visited the NFB causes distortion idea in a 1990s article in
Electronics World. He used a circuit where a resistor plus diode was
used to make a simple circuit produce copious even number H which
would be then converted to additional second ordrr products as NFB was
added, but which then diminished when a larger amount of NFB was used,
typically over 14dB. PP amps like the Williamson or Mullard 520 have
SE input stages so any FB will create IMD even No products to appear
even though the OP stage is PP. But these products are usually at very
low levels because the amp input stages usually produce far less THD
than the OP stage which may only make 2% open loop just under
clipping.

So all the concerns ppl have about NFB creating more mud than it
washes away is largely un-founded, and the measurements end to confirm
all this.

Time and time again, many people find they don't mind if global FB is
either used or not used with a good triode amp.

Patrick Turner.

[Paul G.]

On Wed, 06 Apr 2011 18:54:24 -0500, flipper wrote:

.......[snip ! ]........

Interesting observation about PP cancellation.

That won't happen on a multistage amplifier if the input is SE,
though, because the modulation becomes part of the signal at that
stage.

I realize that's not what you tested and just mean it can't be
universally extrapolated to just 'any ole' PP amplifier.

Might be a good argument for using a differential input, though.

The verdict? Feedback still works.....

That's never been in dispute.

for small amounts of feedback
the ratios of the harmonics vary with respect to each other, but much
less than what the overall reduction from feedback does.


Although the front end of a power amp can be single ended, the
signal levels (if designed properly) are quite low compared to the
final section. You'd expect 2nd harmonic distortion to be 1% or less
at that stage. Using 1%, would give the 3rd harmonic at roughly .01%
(-80db), and the 5th at -120db as per the Baxandall paper. The odd
harmonics DO pop up, but they're pretty small.
A bit off topic, if you ARE using a single ended topology
throughout, you can bias the driver stage in such a way to partially
cancel the 2nd harmonics, and thereby reducing some of the odd
harmonic component as a side effect. I built a single ended tube amp
some years ago to see what all the fuss was about. Using the above
"trick" and optimum feedback, it both measured and sounded well.
It had about 15w/ch, under .1% at full power, and way less than .1% at
lower power. Using level matched A/B testing by changing the amps
only, I found it very difficult to hear the difference between it and
a good modern amplifier. I suppose that's heresy.

Paul G.

[Big Bad Bob]

On 04/07/11 07:57, Paul G. so wittily quipped:
A bit off topic, if you ARE using a single ended topology
throughout, you can bias the driver stage in such a way to partially
cancel the 2nd harmonics, and thereby reducing some of the odd
harmonic component as a side effect. I built a single ended tube amp
some years ago to see what all the fuss was about. Using the above
"trick" and optimum feedback, it both measured and sounded well.
It had about 15w/ch, under .1% at full power, and way less than .1% at
lower power. Using level matched A/B testing by changing the amps
only, I found it very difficult to hear the difference between it and
a good modern amplifier. I suppose that's heresy.

sounds like a good design to me

I usually bias for 'middle of the spec' operation. Seems to work ok.

[Big Bad Bob]

On 04/06/11 20:38, flipper so wittily quipped:
I think you must be a myers-briggs 'SJ' (guess ESTJ) type, where I'm a
'ENTP' type. You appear to see things in terms of details and direct
observations, whereas I look at the general case based on real world
experience and my understanding of the theories behind it all (besides
the details being boring).

Uh huh. Well, if you aren't 'observing' the real world then what makes
you think you've got a handle on it?

It's a combination of lots of experience, extremely good insight,
incredibly high intelligence, and the attitude of a hacker.

thanks for confirming my observation. keep in mind that some of us can
see the entire alphabet after only 2 letters. Others have to go through
all of the letters before it becomes clear. I'd rather not be bored by
the detail of the other 24, so I'll stop after 'B' and say "alphabet".
Occasionally there may be a missing letter, and the detail guy will pick
up on that. But it's not interesting enough for me to care about 1
letter missing. The 'alphabet' concept is sufficient.

[Big Bad Bob]

On 04/08/11 03:18, flipper so wittily quipped:
On Fri, 08 Apr 2011 01:40:44 -0700, Big Bad Bob
wrote:

On 04/06/11 20:38, flipper so wittily quipped:
I think you must be a myers-briggs 'SJ' (guess ESTJ) type, where I'm a
'ENTP' type. You appear to see things in terms of details and direct
observations, whereas I look at the general case based on real world
experience and my understanding of the theories behind it all (besides
the details being boring).

Uh huh. Well, if you aren't 'observing' the real world then what makes
you think you've got a handle on it?

It's a combination of lots of experience, extremely good insight,
incredibly high intelligence, and the attitude of a hacker.

thanks for confirming my observation. keep in mind that some of us can
see the entire alphabet after only 2 letters. Others have to go through
all of the letters before it becomes clear. I'd rather not be bored by
the detail of the other 24, so I'll stop after 'B' and say "alphabet".
Occasionally there may be a missing letter, and the detail guy will pick
up on that. But it's not interesting enough for me to care about 1
letter missing. The 'alphabet' concept is sufficient.


I guess that explains why you haven't been seeing what's actually
said, only 1/13 of the letters get through.

more likely I'm just explaining things inadequately (or leaving out
explanations for things that appear intuitively obvious to me) for
someone that sees things the way you do. So the letters get through,
but I would view it as an explanation of a concept and not a set of details.

[Big Bad Bob]

On 04/08/11 15:33, flipper so wittily quipped:
On Fri, 08 Apr 2011 15:21:05 -0700, Big Bad Bob
wrote:

On 04/08/11 03:18, flipper so wittily quipped:
On Fri, 08 Apr 2011 01:40:44 -0700, Big Bad Bob
wrote:

On 04/06/11 20:38, flipper so wittily quipped:
I think you must be a myers-briggs 'SJ' (guess ESTJ) type, where I'm a
'ENTP' type. You appear to see things in terms of details and direct
observations, whereas I look at the general case based on real world
experience and my understanding of the theories behind it all (besides
the details being boring).

Uh huh. Well, if you aren't 'observing' the real world then what makes
you think you've got a handle on it?

It's a combination of lots of experience, extremely good insight,
incredibly high intelligence, and the attitude of a hacker.

thanks for confirming my observation. keep in mind that some of us can
see the entire alphabet after only 2 letters. Others have to go through
all of the letters before it becomes clear. I'd rather not be bored by
the detail of the other 24, so I'll stop after 'B' and say "alphabet".
Occasionally there may be a missing letter, and the detail guy will pick
up on that. But it's not interesting enough for me to care about 1
letter missing. The 'alphabet' concept is sufficient.


I guess that explains why you haven't been seeing what's actually
said, only 1/13 of the letters get through.

more likely I'm just explaining things inadequately (or leaving out
explanations for things that appear intuitively obvious to me) for
someone that sees things the way you do. So the letters get through,
but I would view it as an explanation of a concept and not a set of details.

Perhaps, but you'll never know till you read what's said because not
everything that starts with AB is a rendition of the English alphabet,
and it only seems that way to you because you don't bother to go past
the AB.

uh, no. not even close.