A question to mr Pass about the document; Audio, distortion and feedback.

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
This was done by simulation. In this case, I have arbitrarily
picked 7 frequencies of equal amplitude from a fundamental
of 1 up to 29 and calculated the distortion through a device
having 1% 2nd and 1% 3rd harmonic distortion and displayed
the input waveform vs the distortion.

You could do the same thing with 7 generators and a low
distortion subtractive circuit (given the high numbers, any
decent op amp would do)

There is nothing special about the test - I simply wanted to
illustrate how badly this stuff piles up.

:cool:
 
Papa..... Point well made......

Although I don't think I could have gotten through college physics without my K&E 'rule...... But then again, I think we only had three elementry particles to contend with, "back in the good old days"....!

May you have a great holiday season, Mr Pass. And may your 2011 be filled with only the most enjoyable products of the Napa and Sonoma Vallley.....!!
 
Hello mr. Pass

Again about your document: Audio, distortion and feedback.

There is this graphic, I include an image of it.

In this graphic and theoretically the best sound are at both end of it, so; an amp with no feedback or with 40 db feedback will have a very low distortions in the higher thd harmonics, and so a better sound.

After tests and listening tests on various amps circuits, I found out that if I use a high level of global negative feedback the amp sound much less good than if I use a low level of global negative feedback.

From your experiences and knowledges, do you have any explanations of why, even if according to the graph, a 40 db feedback will give very low distortions in the higher thd harmonics, in reality it sound much less good than if I use a low level of global negative feedback ?

Thank

Bye

Gaetan
 

Attachments

  • thd vs gnfb-2.gif
    thd vs gnfb-2.gif
    147.3 KB · Views: 357
Last edited:

jcx

Member
2003-02-17 7:38 pm
..
a low feedback amp with largish low order distortion does have substantial IMD – IMD products are a consequence of multiple frequency signal mixing in a nonlinearity

lower order nonlinearities do produce fewer IMD product frequencies for the same input but they still produce IMD

the article's implication is by Nelson's own admission "exaggerated" for effect - real high gain feedback circuits early stages contribute very little distortion as a consequence of the high loop gain - the signals in the input stages are very small because of the high gain, so the stages are operating "deep Class A" using a small portion of their load curve which generates low levels of low order distortion
this is “in addition” to the feedback distortion reduction – which is larger with higher loop gain

I would still recommend "stealing" every one of Nelson's low distortion device selection, bias tricks to make the stages of a high feedback amp as linear as possible open loop

but well implemented high feedback audio amplifier circuits are not displeasing to the ear because of something they add to the audio signal like the hypothesized 7th order distortion or complex IMD products - it is perfectly feasible to design high feedback audio amplifiers with all measurable audio frequency distortions below the noise floor, with complex multitone signals, "complex" speaker loads, measuring THD, IMD, TIM, PIM...

this is the 2nd decade of the 21st century, medical ultrasound, telcom A/DSL demands have greatly upped the performance of op amps, ADC/DACs, improved circuit design, widely diffused the knowledge to analog design EEs – it is only “Audiophile” designers that get away with strawman arguments about 1970's amplifiers and the “evils” of feedback, ignoring broad advances in technology and nonlinear control theory in particular
 

jcx

Member
2003-02-17 7:38 pm
..
you can build circuits with open loop distortion below the starting 10% 2nd order nolinearity in the graph

you can add 20-40 dB more loop gain in many audio amp circuits for 60-80 dB loop gain at all audio frequencies - at which point the harmonics above the 3rd are below -120 dB - ie in the noise - assuming you start with 10 % open loop distortion

Cherry shows the same math in his JAES paper with more insight into distortion contributions by individual devices in a audio amp

Estimates of Nonlinear Distortion in Feedback Amplifiers
Volume 48 Number 4 pp. 299-313; April 2000
 
Hello

I usually do my amps with an open loop of 60-70 db and a feedback of around 30 db.

Phase intermodulation distortion seem a promising type of measurement, anyone use it, what do we need to do it ?

Few years ago there was lot of talk about the cheever thesis document and there was quites polarizes opinions, what can we say today about this document ?

Thank

Bye

Gaetan
 

tessier

Member
2008-05-27 7:14 pm
This was done by simulation. In this case, I have arbitrarily
picked 7 frequencies of equal amplitude from a fundamental
of 1 up to 29 and calculated the distortion through a device
having 1% 2nd and 1% 3rd harmonic distortion and displayed
the input waveform vs the distortion.

You could do the same thing with 7 generators and a low
distortion subtractive circuit (given the high numbers, any
decent op amp would do)

There is nothing special about the test - I simply wanted to
illustrate how badly this stuff piles up.

:cool:

Hi

It's very interesting.

Which frequencies did you chosed to get 7 non-harmonically related tones to do your IMD test ?

Tanx

Paul
 
Last edited:
Hi, Nelson:

Since IMD is due to an amplifier's non-linearity, as is THD, do you find there is a usefully high correlation between an amplifier's IMD and THD performance, as one might then expect? In other words, do you find harmonic distortion spectrum tests alone sufficient to judge amplifier non-linearity, or is it uniquely valuable to also perform an intermodulation specific test? If a IMD specific test is of some unique value, is there then a general approach in which you go about improving an amplifier's IMD performance, yet which does not necessarily significantly improve it's measured THD performance as well?

-Ken