Re: Incoherent is an understatement.......
AAHHHH the intrepid Homo....
http://www.sjostromaudio.com/jocko_homo.html
...need i say more?
This forum would not be half as much fun without our beloved Jock...
Jocko Homo said:
AAHHHH the intrepid Homo....
http://www.sjostromaudio.com/jocko_homo.html
...need i say more?
This forum would not be half as much fun without our beloved Jock...
Excellent Bedtime Reading....
http://signal.ece.utexas.edu/seminars/dsp_seminars/01fall/AudioMyths.pdf
http://signal.ece.utexas.edu/seminars/dsp_seminars/01fall/AudioMyths.pdf
john curl said:
It's true that time delay ,phase shift ,etc can be obvious in a null test,because they are departures from the principle output equal the input .
But as in some of my own designs ,the null test only give thermal noise (even with music as source)...and even that ,with much amplification of the residual...can this amp be improved?
tools for reasoning about distortion
"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 – with 119 references it is a truly thorough and relatively recent review article of the history of distortion measurement back to the dawn of audio reproduction combined with an excellent explanation of the complexities of the general distortion representation and measurement problem with a historical review of and pointers towards distortion audibility modeling
while no complete answer to the questions surrounding distortion perception are near, the ability to reason about and discuss the subject would be advanced by a higher level of sophistication about distortion generation and measurement, you don’t need to master Volterra math to get a lot out of Czerwinski (good thing too, I’m not exactly conversant in multidimensional convolution)
Any engineer here serious about understanding distortion should figure out how to get this article
and at the risk of sending this off into yet another reminiscence about JAES editorial policy and personality conflicts I can’t avoid referencing Cherry on this subject:
Cherry, “Estimates of Nonlinear Distortion in Feedback Amplifiers” JAES V48#4 2000 is another jewel that clarifies and simplifies analysis of distortion mechanisms and feedback without the obscuring Volterra math although it is consequently lacking in explaining Intermodulation distortion
"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 – with 119 references it is a truly thorough and relatively recent review article of the history of distortion measurement back to the dawn of audio reproduction combined with an excellent explanation of the complexities of the general distortion representation and measurement problem with a historical review of and pointers towards distortion audibility modeling
while no complete answer to the questions surrounding distortion perception are near, the ability to reason about and discuss the subject would be advanced by a higher level of sophistication about distortion generation and measurement, you don’t need to master Volterra math to get a lot out of Czerwinski (good thing too, I’m not exactly conversant in multidimensional convolution)
Any engineer here serious about understanding distortion should figure out how to get this article
and at the risk of sending this off into yet another reminiscence about JAES editorial policy and personality conflicts I can’t avoid referencing Cherry on this subject:
Cherry, “Estimates of Nonlinear Distortion in Feedback Amplifiers” JAES V48#4 2000 is another jewel that clarifies and simplifies analysis of distortion mechanisms and feedback without the obscuring Volterra math although it is consequently lacking in explaining Intermodulation distortion
john curl said:
I'm not sure what you mean by that. Certainly it's true that a harmonic distortion measurement won't tell us anything about the signal-level-dependent phase relationship between input and output at the fundamental frequency. But the end result of the AM-to-PM conversion is harmonic distortion, mostly third order. That's a direct result of going through the math of Gilbert's paper. He assumes an undistorted output signal, then works backward to the input to compute the distorted, phase shifted signal that must have been the input. In a way, it's like Cherry's "anti-distortion" concept. Of course, given a certain level of harmonic distortion, it's impossible to say whether it was generated by AM-to-AM, AM-to-PM, or some combination of the two. Maybe that's what you were saying? In that sense, it's not so much a case of "can't be measured" as "can't be cleanly separated".
Jan asked me something related to this earlier in the thread, and I realized I gave him a pretty crummy answer. Thinking about it some more, the AM-to-PM as defined by Gilbert must be calculated with a closed-loop system. By definition, it's the distortion left over after the feedback has done its thing. To characterize it as something the feedback cannot fix doesn't really make much sense because by its very definition, it's what's left over after the feedback loop is closed. What's left over after feedback is by definition not removed by the feedback. Certainly the closed-loop distortion in Gilbert's examples is less than the open-loop distortion. However, the feedback has "mutated" the pure AM-to-AM nonlinearity into an AM-to-PM nonlinearity in the process of reducing the distortion of the open-loop amplifier.
Gilbert's example is also probably the most extreme case of input stage nonlinearity that could be thought of: a non-degenerated bipolar diff amp. If we're talking power amps, simple emitter degeneration would go a long way toward fixing this. And how about this opposite extreme? Suppose you had a heavily degenerated input stage, with a very poor VAS having a bad AM-to-PM problem. That kind of turns things around, doesn't it? In Gilbert's analysis, there was zero AM-to-PM in the open-loop amplifier. If the open-loop distortion were dominated by AM-to-PM, you don't think the feedback would make it even worse, do you? I doubt it. In that case, I would think it would reduce the AM-to-PM when the loop is closed. But of course I have no proof of that.
Here's one thing I'd like to try. I have available to me (through my employer) a harmonic balance simulator, usually used by RF/microwave people. This type of simulator can be swept in amplitude and frequency, and plots of AM-to-PM can be generated. It "knows" the amplitude and phase of each frequency component, so this kind of info can be computed and plotted. I'll see if I can get hold of a copy and set it up with some SPICE models.
It just so happens that the UA709, 5534, AD797, and the AD829 for example, have the same, UNDEGENERATED, dual differential transistor input stage. Why? Because resistors add NOISE to the input.
To me, it is like this: WHEN you have significant open loop distortion generated by the input stage, even if everything else is perfect and does not contribute, (which is impossible) then the open loop bandwidth will be modulated with signal level and frequency. This will cause a dynamic phase shift of the audio output which is outside the direct control of global negative feedback. A dynamic phase shift would not be easily seen with a simple harmonic distortion measurement, at least in my estimation. Therefore, in principle, you could have a low distortion design (AM distortion that is) that actually has a significant amount of dynamic phase shift, when real audio signals are put through it. This MAY be why IC op amps (with a few exceptions) don't sound as good as most discrete designs.
But what the heck, why not use the cheapest op amp that makes you believe that you have achieved audio perfection? ;-)
To me, it is like this: WHEN you have significant open loop distortion generated by the input stage, even if everything else is perfect and does not contribute, (which is impossible) then the open loop bandwidth will be modulated with signal level and frequency. This will cause a dynamic phase shift of the audio output which is outside the direct control of global negative feedback. A dynamic phase shift would not be easily seen with a simple harmonic distortion measurement, at least in my estimation. Therefore, in principle, you could have a low distortion design (AM distortion that is) that actually has a significant amount of dynamic phase shift, when real audio signals are put through it. This MAY be why IC op amps (with a few exceptions) don't sound as good as most discrete designs.
But what the heck, why not use the cheapest op amp that makes you believe that you have achieved audio perfection? ;-)
Insofar as this is a complete argument rather than a question,
I'm not sure there is a sensible answer. "It may," is one answer, but the more I learn about masking phenomenae and sound reproduction, the less certain I am of what SPLs of higher order nonlinearities lying outside the masking area, may be detected by a listener.
If the question is ideological, and you are asking me if I believe in amelioration brought with modern progress, then as a modern person, I will say, "Of course, perhaps, er, well, sure, I expect so...etc."
If you are asking me to agree with your argument right now, I have to say, despite its formal rigour, there are material considerations: there folk on this thread with a great deal more knowledge and experience than I have, who are discussing the matter right now and I'd like to see just where this dicussion leads..... I always enjoy discussion that looks at premises.
But your argument is flawless, as an argument.
I'm not sure there is a sensible answer. "It may," is one answer, but the more I learn about masking phenomenae and sound reproduction, the less certain I am of what SPLs of higher order nonlinearities lying outside the masking area, may be detected by a listener.
If the question is ideological, and you are asking me if I believe in amelioration brought with modern progress, then as a modern person, I will say, "Of course, perhaps, er, well, sure, I expect so...etc."
If you are asking me to agree with your argument right now, I have to say, despite its formal rigour, there are material considerations: there folk on this thread with a great deal more knowledge and experience than I have, who are discussing the matter right now and I'd like to see just where this dicussion leads..... I always enjoy discussion that looks at premises.
But your argument is flawless, as an argument.
mikeks said:
http://www.gedlee.com/distortion_perception.htm
For those who cannot open the PP presentation on the above page I've created an HTML copy here:
http://members.iinet.net.au/~jmai/distortion_perception_files/frame.html
Mikeks, I'll confine myself to style so as not to get the level of toooo perrrrsonal. Things look to me like this:
J Curl does 'grumpy avuncular' quite well and stubbornly follows his direction, not mine or yours. It's his nickle, his life, he's entitled
You do 'social Methodism' quite well. It's your 10P, your life, you're entitled..
BUT
There is no official way of behaving here except for prohibitions against direct personal attacks and against counselling folk to do dangerous or illegal things.
However, there are certain unofficial minimal levels of politeness and gentillesse. You manage to negatively surpass them in your 'Methodist' enthusiasm. You escalate, even.
There is also here an unofficial class system found throughout the modern world: you are what you do; by your works you shall be known; etc. In this world here, Curl is high class and there's no point in getting your 'social Methodist' knickers in a twist over it just because he won't do things your way.
There is a practical matter with respect to rudeness.
In another forum I sometimes participate in, the participant with the most outstanding record of accomplishment and I must say useful contribution to discussion was told by a another member to take his accomplishments, academic attainment, etc, roll it, and shove it up his ****. That valuable member has not posted there since.
John Curl Grumpy?
Jeesh, you haven't seen him grumpy. You would know when he gets grumpy. He appears to be in a very good and giving mood on this board. The guy has circuit sense so use this knowledge to your benefit as you build better gear as chances are he has tried about everything that has a chance of working musically.....
Jeesh, you haven't seen him grumpy. You would know when he gets grumpy. He appears to be in a very good and giving mood on this board. The guy has circuit sense so use this knowledge to your benefit as you build better gear as chances are he has tried about everything that has a chance of working musically.....
Right, what do we know.........
What does power cord resistance have to do with anything? Didn't this guy ever hear of RF?
Oh.......wait.......our amps can not be properly designed if any Rf affects it.
OTOH........show me any electronic gizmo that isn't subject to some sort of degradation from RF. OK.....so maybe it takes a few more watts on some than others.
Jocko
What does power cord resistance have to do with anything? Didn't this guy ever hear of RF?
Oh.......wait.......our amps can not be properly designed if any Rf affects it.
OTOH........show me any electronic gizmo that isn't subject to some sort of degradation from RF. OK.....so maybe it takes a few more watts on some than others.
Jocko
more reason to look up the Czerwinski article
http://www.gedlee.com/distortion_perception.htm is good but the Czerwinski article gives more background and context on distortion properties and Volterra representation that suggests questioning the assumption by Geddes that a swept frequency single sine wave distortion measurement is sufficient to characterize a audio amplifier
I’ve updated my previous “Hidden Distortion” sim to show a simple circuit that has 1000x more distortion with a range of 2 tone excitations than can be seen with a swept frequency single sine wave distortion measurement
First build a distortion model circuit with hi and lo pass filtered frequency branches feeding into a multiplier
play with LtSpice Modulate source to get a 20 –20 KHz sweep and show the peak distortion output is –74dB – comfortably bellows mike’s 1000 ppm value
Then uncover the “hidden distortion” with a 2 tone 160 Hz + 16 KHz sine wave excitation with the same peak input amplitude (actually any 2 tones one <160 Hz, the other > 16 KHz will show this large distortion)
The behavioral source is simply multiplying the signal from the lo and hi pass filter branches, in a real amplifier most nonlinearities have only one input “port” but the usual diff pair input stage is virtually identical to a 2 port Gilbert cell multiplier, and thermal modulation (Lavardin in the audiophile world) can be modeled by adding a RC thermal model to a Temperature “port” of each transistor model (but not easily in spice transistor models)
This is an artificial example that gives rather more extreme results than could possibly be expected in practical amplifiers but it clearly illustrates the principle that single tone distortion measures are inadequate
The assertion that the GedLee metric and, by Geddes assumption, swept frequency single sine wave distortion measurement are sufficient for audio power amplifier charterization requires more empirical data – it is clearly not sufficient in principle and multitone techniques linked to perceptual modeling looks to be the way to move forward in amplifier testing as well as loudspeaker design
http://www.gedlee.com/distortion_perception.htm is good but the Czerwinski article gives more background and context on distortion properties and Volterra representation that suggests questioning the assumption by Geddes that a swept frequency single sine wave distortion measurement is sufficient to characterize a audio amplifier
I’ve updated my previous “Hidden Distortion” sim to show a simple circuit that has 1000x more distortion with a range of 2 tone excitations than can be seen with a swept frequency single sine wave distortion measurement
First build a distortion model circuit with hi and lo pass filtered frequency branches feeding into a multiplier
play with LtSpice Modulate source to get a 20 –20 KHz sweep and show the peak distortion output is –74dB – comfortably bellows mike’s 1000 ppm value
Then uncover the “hidden distortion” with a 2 tone 160 Hz + 16 KHz sine wave excitation with the same peak input amplitude (actually any 2 tones one <160 Hz, the other > 16 KHz will show this large distortion)
The behavioral source is simply multiplying the signal from the lo and hi pass filter branches, in a real amplifier most nonlinearities have only one input “port” but the usual diff pair input stage is virtually identical to a 2 port Gilbert cell multiplier, and thermal modulation (Lavardin in the audiophile world) can be modeled by adding a RC thermal model to a Temperature “port” of each transistor model (but not easily in spice transistor models)
This is an artificial example that gives rather more extreme results than could possibly be expected in practical amplifiers but it clearly illustrates the principle that single tone distortion measures are inadequate
The assertion that the GedLee metric and, by Geddes assumption, swept frequency single sine wave distortion measurement are sufficient for audio power amplifier charterization requires more empirical data – it is clearly not sufficient in principle and multitone techniques linked to perceptual modeling looks to be the way to move forward in amplifier testing as well as loudspeaker design
FrankWW said:
No disrespect intended here Frank, but this is manifest nonsense...
I have no inclination to get my proverbial knickers in a sailors knot merely because John 'won't do things your way'...
On the contrary, i am quite...quite happy in a sadistic sort of way, that John insists on doing things his 'way'....
What i take exception to, as i have pointed out elsewhere, is the implicit expectation by the worthy that novices should take his design idiosyncracies at face value without question......
THAT is nonsense!
http://www.diyaudio.com/forums/showthread.php?postid=483377#post483377
P.S:
What you call rudeness, is considered plain speaking by the enlightened....
Re: more reason to look up the Czerwinski article
Hi JCX,
A most illuminating analysis..Cheers.
However, i cannot help thinking that although an amps. non-linear mechanisms are more rigorously excercised by multi-tone stimuli, this does not in itself mean that the simple THD test will not register changes in linearity detected by the multitone method...
viz:
Can you imagine circumstances in which an increase in non-linearity detected by multi-tone stimuli cannot similarly be detected by simple THD+N....?
Moreover, do you envisage situations in which the multitone method registers an increase in non-linearity, while THD registers a spurious decrease?
This is key methinks.
Cheers.
jcx said:
Hi JCX,
A most illuminating analysis..Cheers.
However, i cannot help thinking that although an amps. non-linear mechanisms are more rigorously excercised by multi-tone stimuli, this does not in itself mean that the simple THD test will not register changes in linearity detected by the multitone method...
viz:
Can you imagine circumstances in which an increase in non-linearity detected by multi-tone stimuli cannot similarly be detected by simple THD+N....?
Moreover, do you envisage situations in which the multitone method registers an increase in non-linearity, while THD registers a spurious decrease?
This is key methinks.
Cheers.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.