Millwood, there have been studies of higher order distortion and its 'annoyance' factor since the 1930's. You could read the 1941 'Radiotron Designers Handbook', like I did, and know that even before WW2, higher order odd distortion was considered very bad for audio quality. Some of the references in the Boyk link would give you even deeper insight.
Can Old Dogs learn New Tricks?
the repetitive and uninformed nature of this "discussion" frustrates me - the fact that a few people new to the field are overly impressed by a few old papers that happen to reinforce their preconceptions about willful ignorance of the objectivist engineering community, bad negative feedback, "solid state" sound ect keeps this ridiculous game of "telephone" going - everyone repeats the same old stuff
people only referencing Otala's cheering section would appear to be pushing an agenda by selective reference - andy_c nailed the flaw in the Otala/Gilbert analysis quickly as did Cherry in "Amplitude and Phase of Intermodulation Distortion" JAES V31#5 May 1983
distortion analysis has advanced substantially, P Wambacq & W Sansen “Distortion Analysis of Analog Integrated Circuits” 1998 is representative of where engineering analysis can go when the demands (= money) of a major market like DSL/ADSL engage the research community, Cherry, “Estimates of Nonlinear Distortion in Feedback Amplifiers” JAES V48#4 2000 is another jewel that clarifies and simplifies distortion analysis without the obscuring Volterra math
why not try reading a truly thorough and relatively recent review article: "Multitone Testing of Sound System Components - Some Results and Conclusions, Part 1: History and Theory" JAES V 49#11 nov 2001 by Czerwinski et al at Cerwin Vega - 119 references!
the paper sets a high standard in historical and engineering analysis of distortion and audibility - the art advances, why don't these "discussions"?
the repetitive and uninformed nature of this "discussion" frustrates me - the fact that a few people new to the field are overly impressed by a few old papers that happen to reinforce their preconceptions about willful ignorance of the objectivist engineering community, bad negative feedback, "solid state" sound ect keeps this ridiculous game of "telephone" going - everyone repeats the same old stuff
people only referencing Otala's cheering section would appear to be pushing an agenda by selective reference - andy_c nailed the flaw in the Otala/Gilbert analysis quickly as did Cherry in "Amplitude and Phase of Intermodulation Distortion" JAES V31#5 May 1983
distortion analysis has advanced substantially, P Wambacq & W Sansen “Distortion Analysis of Analog Integrated Circuits” 1998 is representative of where engineering analysis can go when the demands (= money) of a major market like DSL/ADSL engage the research community, Cherry, “Estimates of Nonlinear Distortion in Feedback Amplifiers” JAES V48#4 2000 is another jewel that clarifies and simplifies distortion analysis without the obscuring Volterra math
why not try reading a truly thorough and relatively recent review article: "Multitone Testing of Sound System Components - Some Results and Conclusions, Part 1: History and Theory" JAES V 49#11 nov 2001 by Czerwinski et al at Cerwin Vega - 119 references!
the paper sets a high standard in historical and engineering analysis of distortion and audibility - the art advances, why don't these "discussions"?
A little story about Dr. Cherry. In the late '60's, Cherry and Hooper wrote a textbook on amplifier design. I bought and used the book for years. It had about 1000 pages. NOWHERE, in the book, was any mention of slew rate, 'slope distortion' or rate of change distortion effects.
We found it ironic that in the '80's when Dr Cherry attempted to take over TIM by calling it "slope distortion" Where was he when Dr. Otala was first finding this stuff out? Anyone who is good at math can 'snow' the reader that many factors don't count.
Today, early critics of Dr. Otala, like Barrie Gilbert, have written modern analysis of amp circuits, basically saying the same thing that Dr. Otala set forth in 1980 or earlier. AND, the NEWEST analysis just varifies understanding of the musical listening experience going back to the 1930's. Amplifier designers ignore this understanding at their own peril.
We found it ironic that in the '80's when Dr Cherry attempted to take over TIM by calling it "slope distortion" Where was he when Dr. Otala was first finding this stuff out? Anyone who is good at math can 'snow' the reader that many factors don't count.
Today, early critics of Dr. Otala, like Barrie Gilbert, have written modern analysis of amp circuits, basically saying the same thing that Dr. Otala set forth in 1980 or earlier. AND, the NEWEST analysis just varifies understanding of the musical listening experience going back to the 1930's. Amplifier designers ignore this understanding at their own peril.
A little story about Dr. Cherry. In the late '60's, Cherry and Hooper wrote a textbook on amplifier design. I bought and used the book for years. It had about 1000 pages. NOWHERE, in the book, was any mention of slew rate, 'slope distortion' or rate of change distortion effects.
We found it ironic that in the '80's when Dr Cherry attempted to take over TIM by calling it "slope distortion" Where was he when Dr. Otala was first finding this stuff out? Anyone who is good at math can 'snow' the reader that many factors don't count.
Today, early critics of Dr. Otala, like Barrie Gilbert, have written modern analysis of amp circuits, basically saying the same thing that Dr. Otala set forth in 1980 or earlier. AND, the NEWEST analysis just varifies understanding of the musical listening experience going back to the 1930's.
Designers. Amplifier designers ignore this understanding at their own peril.
We found it ironic that in the '80's when Dr Cherry attempted to take over TIM by calling it "slope distortion" Where was he when Dr. Otala was first finding this stuff out? Anyone who is good at math can 'snow' the reader that many factors don't count.
Today, early critics of Dr. Otala, like Barrie Gilbert, have written modern analysis of amp circuits, basically saying the same thing that Dr. Otala set forth in 1980 or earlier. AND, the NEWEST analysis just varifies understanding of the musical listening experience going back to the 1930's.
Designers. Amplifier designers ignore this understanding at their own peril.
It is not that difficult for an amp's designer to built amp "X" and amp "Y" having the same overall THD number, but differing in the following:
X - has mostly 2nd and 3rd harmonics in the spectrum with very fast decay of higher order ones,
Y - has the harmonics from 2nd to some 15th more or less the same in amplitude, mostly odd order. This can be easily done by class AB with improper bias current + appropriate feedback factor.
Now listen to both circuits. The Y sounds grainy, also like sawtooth or synthesizer - this is the effect of the high order harmonics.
X - has mostly 2nd and 3rd harmonics in the spectrum with very fast decay of higher order ones,
Y - has the harmonics from 2nd to some 15th more or less the same in amplitude, mostly odd order. This can be easily done by class AB with improper bias current + appropriate feedback factor.
Now listen to both circuits. The Y sounds grainy, also like sawtooth or synthesizer - this is the effect of the high order harmonics.
john curl said:
Actually the Academics, Otala in particular, did Audio something of a disservice by indulging in willful obscurantism in this regard...
See:
Comparison of Nonlinear Distortion Measurement Methods
By Richard Cabot
at
http://audioprecision.com/
TIM as an acronym is intellectual sophism of the worst sort.....It alludes to the simple matter of slew overload in linear circuitry...Period. All the tedious algebra indulged in by its inventor and his acolytes was an unfortunate diversion from more important matters....
Indeed one poor fellow, (Peter Garde)....presented a paper at AES based on Otala, that was wrong from start to finish..
A little off-topic philosophy - I promise this is all
If one's only agenda is determining the truth of a matter, then the issue of who is right is irrrelevant. The only thing that matters is what is right. Human nature seems to make this ideal difficult to achieve in practice. I'm not there yet, but I like to think I at least try.
If one's only agenda is determining the truth of a matter, then the issue of who is right is irrrelevant. The only thing that matters is what is right. Human nature seems to make this ideal difficult to achieve in practice. I'm not there yet, but I like to think I at least try.
No squirrel in the other tree either!
Well, after looking in Wambacq and Sansen's book on Distortion Analysis, I couldn't even find "phase distortion" in the index. Not mentioned in the sections on differential amplifier analysis or Op. amps either. Sergio Franco's book does mention it though. Anyway, I went back and re-read the Walt Jung article on Op-Amp Audio from Electronic Design. The phase distortion they are talking about arises in the gain portion of the amplifier at high frequency, mainly in the last stages. The input diff. pair is just moving the open loop corner frequency around due to the changing gm of the diff. transistors, hence changing phase error at high frequencies versus signal level. So, since the phase lag is not arising in the diff. amp. transistors themselves to any appreciable degree, there will be no difference between the inverting feedback amplifier configuration and the non-inverting amplifier configuration. High gain bandwidth will fix either one apparently. Case closed I guess.
I still think it might be interesting though to look at an actual amplifier with the phase distortion analyser scheme mentioned earlier. Suppose the emitter connection point of the diff. amp. input pair does have some capacitance to ground due to wiring stray capacitance, or use of some inappropriately large transistor for the current source. Then the changing gm of the input transistors due to signal level could produce some varying phase effect due to varying input impedance. Ie, now its the R in the RC circuit thats changing instead of the C. Emitter degeneration resistors would be useful here to minimise this effect. Probably not a problem in an IC amp stage however due to the low distributed capacitances.
Don
Well, after looking in Wambacq and Sansen's book on Distortion Analysis, I couldn't even find "phase distortion" in the index. Not mentioned in the sections on differential amplifier analysis or Op. amps either. Sergio Franco's book does mention it though. Anyway, I went back and re-read the Walt Jung article on Op-Amp Audio from Electronic Design. The phase distortion they are talking about arises in the gain portion of the amplifier at high frequency, mainly in the last stages. The input diff. pair is just moving the open loop corner frequency around due to the changing gm of the diff. transistors, hence changing phase error at high frequencies versus signal level. So, since the phase lag is not arising in the diff. amp. transistors themselves to any appreciable degree, there will be no difference between the inverting feedback amplifier configuration and the non-inverting amplifier configuration. High gain bandwidth will fix either one apparently. Case closed I guess.
I still think it might be interesting though to look at an actual amplifier with the phase distortion analyser scheme mentioned earlier. Suppose the emitter connection point of the diff. amp. input pair does have some capacitance to ground due to wiring stray capacitance, or use of some inappropriately large transistor for the current source. Then the changing gm of the input transistors due to signal level could produce some varying phase effect due to varying input impedance. Ie, now its the R in the RC circuit thats changing instead of the C. Emitter degeneration resistors would be useful here to minimise this effect. Probably not a problem in an IC amp stage however due to the low distributed capacitances.
Don
http://home.comcast.net/~walt-jung/wsb/html/view.cgi-showresources.html-TopRes-Walt-27s-20PDFs.html
Look at the paper about input stages.
This one: http://home.comcast.net/~walt-jung/wsb/PDFs/WTnT_Op_Amp_Audio_4.pdf
Look at the paper about input stages.
This one: http://home.comcast.net/~walt-jung/wsb/PDFs/WTnT_Op_Amp_Audio_4.pdf
john curl said:
Yes, and, even more amazing, in 1863 or thereabouts Wilhelm Helmholz (of the helmholz resonator fame) talked about the different 'tone' perceptions depending on the amount and relative level of harmonics. I just read parts of his 1863 book, which was (excellently) translated in English at the turn of the century (19th, that is), titled "On the perception of tone". Amazon.com had a 1954 reprint available some time ago.
Helmholz was of course very limited in his intrumentation. The way he generated these harmonics in itself would be reason enough to give him the nobel-price, I mean with speed-controlled DC motors driving AC generators driving coils energizing tuning forks, with the tuning fork damping determining the harmonics level!!
He coulnd't do much testing beyond the 5th or 6th harmonic, but his findings have been fully confirmed. I don't have the numbers in my head (I have the book at work), but IIRC he concluded that 5th harmonics up to 0.01 % of the carrier clearly influenced the tonal perception.
On another note, this varying Gm etc is IMHO the root of the discrepancy between measurements and listening experiences. Basically, I think we can agree that THD is a good measure for linearity *under those particular conditions*. The superposition theorem says that if you add more signals to the one you did the measurement with, they would all have the same THD. However, the superposition theorem assumes that the black box we are measuring is invariant, but now we know that that is not true: if we add more signals, like a high level lf signal to the mid-freq signal, or if we add a common mode component, the black box (the amp) reacts differently.
BUT that also would mean that if we had an amp that WAS invariant, we should find good correcpondance between measurements and listening tests. Not necessarily low non-linearity, but good correspondence between meter and perception. Once we have that, we can finally write amplifier specs that predict performance (or so i hope).
Maybe we should look in that direction to bring the meter readers and golden ears together.
Jan Didden
PMA said:
0.003%(30ppm) 3rd harmonic in an amp. delivering say 100 Watts(40Vpeak), is equivalent to 1.2mV...
Somehow i don't think you'll hear a 1.2mV odd harmonic in the presence of 40V clean signal across your transducer....
Indeed, your loudspeaker system is exceedingly unlikely to audibly reproduce 1.2mV of anything.....
I would like to relate a true story regarding Richard Cabot.
Decades ago, I was at an AES meeting featuring Richard Cabot. He, even then, attacked Otala's work on distortion. He claimed that he did NOT get the same measured results as Otala. However, I got up and asked him WHY he did not use the SAME circuit topologies as Otala, if he expected to get the same results. Don't get me wrong, Richard Cabot is a smart guy. When we meet, on occasion, we are friendly, BUT he had it in for Otala from the get-go. The same goes for Bob Cordell. This does NOT free Otala from any responsibility. Once, after one of his AES lectures, I wanted to get one of Paul Klipsch's "BULL****" buttons that he handed out at AES conferences and pin it on Matti ! Why? Because Matti is a political guy. He did NOT always make it easy to understand his work. In this case, he deliberately avoided relating slew-rate with TIM. This annoyed me as much as the other engineers, but I let it go, because he still contributed so much to understanding audio design.
As far as the name 'TIM' was concerned, it was just as good as 'SID' or 'slope distortion'. It had to be called SOMETHING! And he was first.
Decades ago, I was at an AES meeting featuring Richard Cabot. He, even then, attacked Otala's work on distortion. He claimed that he did NOT get the same measured results as Otala. However, I got up and asked him WHY he did not use the SAME circuit topologies as Otala, if he expected to get the same results. Don't get me wrong, Richard Cabot is a smart guy. When we meet, on occasion, we are friendly, BUT he had it in for Otala from the get-go. The same goes for Bob Cordell. This does NOT free Otala from any responsibility. Once, after one of his AES lectures, I wanted to get one of Paul Klipsch's "BULL****" buttons that he handed out at AES conferences and pin it on Matti ! Why? Because Matti is a political guy. He did NOT always make it easy to understand his work. In this case, he deliberately avoided relating slew-rate with TIM. This annoyed me as much as the other engineers, but I let it go, because he still contributed so much to understanding audio design.
As far as the name 'TIM' was concerned, it was just as good as 'SID' or 'slope distortion'. It had to be called SOMETHING! And he was first.
andy_c said:
Linear or invariant, Andy? Suppose we have an amp that say distorts 1%, but is invariant, I mean it distorts *every* signal 1%, then we would not have the problems with phase modulation with signal level etc. The measurements at one frequency would predict the results at any freq and at any combinations of signals, and in that case it would be able to predict how it sounds based on single-signal measurements alone. That's what I was getting at. What do you think?
Jan Didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.