GRollins said:
Grey, we all run out of arguments once in a while. But it is pretty juvenile to resort to personal attacks the way you do in this post. Please, can we keep it factual and mature?
Thank you for your cooperation and understanding,
Jan Didden
GRollins said:
I think my contribution to that thread is a little more positive and technical. Which of those two attributes prevent you from contributing I am not sure. I have starded some simulations, done one special proof, and am working on a more general proof to be posted later.
janneman said:
You rather conveniently ignored the point, which is that if you're claiming that your stereo sounds better than live music, you need to seek better circumstances under which to hear live music. Our hall here isn't anything particularly special in the sonics department, but it sounds quite a bit better than any stereo I've ever heard.
Of perhaps you were suggesting that your stereo sounds better than reality because it adds euphonic distortions? Surely not.
janneman said:
You don't say whether they use a PA or not (if so, you're listening to a big, poor quality stereo, which rather negates the sound as a reference for live music), and chances are the acoustics of the room aren't all that great. At least I've never heard of a bar with an open mic night being held in a decent venue, sonically.
fizzard said:
And therein lies the problem. Well, two related problems, actually. The first is that you are assuming that the math will cover all aspects of reality. It doesn't. Two, and more devastating because it's subtle: You will set out to 'prove' exactly and only what you have already made up your mind is the truth. You have already made up your mind that feedback is perfect--or at least close to it--and are not going to listen to anyone (in this thread or the other) who tries to tell you differently, regardless of whether math is used, or simulations, or analogies, or simple logic. You're not going to get anywhere until you release your preconceived notions.
Or, as I've said before in other contexts: Too much theory. Not enough reality. Reality beats theory, seven days a week.
Until you quit arguing, educate your ears, and go listen, you will never understand that our theoretical understanding doesn't cover everything at this point. I'd like to think that we will eventually develop our theoretical framework sufficiently to explain what we can hear, but it's not there yet.
Oh, by the way, there's a third problem. Simulations only impress people who believe that the simulations model reality with a high degree of accuracy. Although I'm fully confident that the simulation programs will eventually become useful, at this point they are haphazard, at best, no matter how loudly any given person claims that the program they're using is perfect...it's the other guy's program that's at fault. Yeah. Right.
Scott Wurcer has said that the simulation software they use at Analog Devices is light years beyond the stuff available to the public. I'm not in a position to judge as I don't have access to their code, but publicly available simulation software isn't all that great at the present time.
Grey
You claim that an
(X) audible
( ) measurable
( ) hypothetical
improvement in sound quality can be attained by:
( ) upsampling
( ) non-oversampling
( ) increasing word size
( ) vibration dampening
( ) bi-wiring
( ) replacing the external power supply
( ) using a different lossless format
( ) decompressing on the server
( ) removing bits of metal from skull
( ) using ethernet instead of wireless
( ) inverting phase
( ) reversing “polarity” of resistors
( ) ultra fast recovery rectifiers
( ) installing bigger connectors
( ) installing Black Gate caps
( ) installing ByBee filters
( ) installing hospital-grade AC jacks
( ) defragmenting the hard disk
( ) running older firmware
( ) using exotic materials in cabinet
( ) bronze heatsinks
( ) violin lacquer
( ) $500 power cords
( ) a universally applicable omnidirectional tweeter
(X) reducing or eliminating negative feedback
Your idea will not work. Specifically, it fails to account for:
(X) the placebo effect
(X) your ears honestly aren't that good
(X) your idea has already been thoroughly disproved
( ) modern DACs upsample anyway
( ) those products are pure snake oil
( ) lossless formats, by definition, are lossless
( ) those measurements are bogus
( ) sound travels much slower than you think
( ) electric signals travel much faster than you think
( ) that's not how binary arithmetic works
( ) that's not how TCP/IP works
( ) the Nyquist theorem
( ) the can't polish a turd theorem
( ) bits are bits
You will try to defend you idea by:
(X) claiming that your ears are “trained”
(X) claiming immunity to psychological/physiological factors that affect everyone else
(X) name-calling
(X) criticizing spelling/grammar
Your subsequent arguments will probably appeal in desperation to such esoterica as:
( ) jitter
(X) EMI
( ) thermal noise
(X) quantum mechanical effects
(X) resonance
(X) existentialism
(X) nihilism
(X) communism
(X) cosmic rays
And you will then change the subject to:
(X) theories are not the same as facts
(X) measurements don't tell everything
(X) not everyone is subject to the placebo effect
(X) blind testing is dumb
(X) you can't prove what I can't hear
(X) science isn't everything
Rather than engage in this tired discussion, I suggest exploring the following factors which are more likely to improve sound quality in your situation:
(X) room acoustics
(X) source material
(X) type of speakers
(X) speaker placement
( ) crossover points
( ) equalization
( ) Q-tips
(X) psychoanalysis
(X) trepanation
Checklist courtesy of Sean Adams, designer of Slim Devices' Squeezebox and Transporter. Checklist was slightly modified. 🙂
(X) audible
( ) measurable
( ) hypothetical
improvement in sound quality can be attained by:
( ) upsampling
( ) non-oversampling
( ) increasing word size
( ) vibration dampening
( ) bi-wiring
( ) replacing the external power supply
( ) using a different lossless format
( ) decompressing on the server
( ) removing bits of metal from skull
( ) using ethernet instead of wireless
( ) inverting phase
( ) reversing “polarity” of resistors
( ) ultra fast recovery rectifiers
( ) installing bigger connectors
( ) installing Black Gate caps
( ) installing ByBee filters
( ) installing hospital-grade AC jacks
( ) defragmenting the hard disk
( ) running older firmware
( ) using exotic materials in cabinet
( ) bronze heatsinks
( ) violin lacquer
( ) $500 power cords
( ) a universally applicable omnidirectional tweeter
(X) reducing or eliminating negative feedback
Your idea will not work. Specifically, it fails to account for:
(X) the placebo effect
(X) your ears honestly aren't that good
(X) your idea has already been thoroughly disproved
( ) modern DACs upsample anyway
( ) those products are pure snake oil
( ) lossless formats, by definition, are lossless
( ) those measurements are bogus
( ) sound travels much slower than you think
( ) electric signals travel much faster than you think
( ) that's not how binary arithmetic works
( ) that's not how TCP/IP works
( ) the Nyquist theorem
( ) the can't polish a turd theorem
( ) bits are bits
You will try to defend you idea by:
(X) claiming that your ears are “trained”
(X) claiming immunity to psychological/physiological factors that affect everyone else
(X) name-calling
(X) criticizing spelling/grammar
Your subsequent arguments will probably appeal in desperation to such esoterica as:
( ) jitter
(X) EMI
( ) thermal noise
(X) quantum mechanical effects
(X) resonance
(X) existentialism
(X) nihilism
(X) communism
(X) cosmic rays
And you will then change the subject to:
(X) theories are not the same as facts
(X) measurements don't tell everything
(X) not everyone is subject to the placebo effect
(X) blind testing is dumb
(X) you can't prove what I can't hear
(X) science isn't everything
Rather than engage in this tired discussion, I suggest exploring the following factors which are more likely to improve sound quality in your situation:
(X) room acoustics
(X) source material
(X) type of speakers
(X) speaker placement
( ) crossover points
( ) equalization
( ) Q-tips
(X) psychoanalysis
(X) trepanation
Checklist courtesy of Sean Adams, designer of Slim Devices' Squeezebox and Transporter. Checklist was slightly modified. 🙂
GRollins said:
Math predicts reality remarkably well in the world of electronics and feedback control.
GRollins said:
You do not know of which you speak, sir. In math it is not possible to prove true that which is false.
It is virtually impossible to do the exact mathematical expression of a complex transfer function. You just don't have the time or energy.
All one really has to do is to know the general math and then BUILD the circuit AND then listen to it. It is faster, cheaper, and more accurate.
For example, do you think that the people who designed the Quad 405 didn't know or care about math? Wrong! They knew plenty of math, yet made a 'questionable' product. Would you like to explain why? Or do you know what they should have done differently? Or maybe you own a Quad 405 or its equivalent and think that it is just about perfect.
For the record, I took a control theory course from Dr. Dorf (look him up) in 1968, the same time I developed the complementary differential input stage. You are debating people who have been exposed to math, not technicians.
All one really has to do is to know the general math and then BUILD the circuit AND then listen to it. It is faster, cheaper, and more accurate.
For example, do you think that the people who designed the Quad 405 didn't know or care about math? Wrong! They knew plenty of math, yet made a 'questionable' product. Would you like to explain why? Or do you know what they should have done differently? Or maybe you own a Quad 405 or its equivalent and think that it is just about perfect.
For the record, I took a control theory course from Dr. Dorf (look him up) in 1968, the same time I developed the complementary differential input stage. You are debating people who have been exposed to math, not technicians.
john curl said:
I suspect some of the top manufacturer's are doing the virtually impossible then. A suspiciously non-round valued resister here, and a capacitor that leaves you scratching your head(at least me) there. It's not practical to try hundreds of different values. If something doesn't sound right it's most likely measurable.
If there's an inadequacy in the reproduction chain that prevents recorded material from sounding like the original it should not be left to the amplifier designer to correct it. The amplifier should always be designed for low noise, low distortion, perfect phase, i.e: the infamous "straight wire with gain." Maybe surgical precision doesn't sound good to some people? I feel that amplifiers shouldn't be musical instruments expected to contribute a "voice" of their own to the sound.
john curl said:
That makes you the guy to ask then: On the differential input to an amplifier (a Bryston) they have collector resistors on the input transistors (which they need obviously), but the collectors of the feedback side are connected directly to the power rails. I'm sure there's a simple explanation?
fizzard said:
Once again, you demonstrate that you don't even understand the question, much less the answer.
Obviously, we're not attempting to prove a falsehood to be true.
Duh!
Take a moment and look up the literal definition of the word sophomore.
Consider the blind men and the elephant. Each man could conceivably "prove" that he had a (insert item of your choice here) when what he was probing was a leg, a trunk, a tail, etc. But have any of them actually 'proved' that what they're testing isn't an elephant?
Or more to the point, did they actually, really, truly succeed in 'proving' that they were examining a snake, a tree, etc?
No.
They 'proved' what they set out to prove.
They think they have proven their point, and will argue their proof loudly, but the reality is that they are feeling an elephant. Each, within his grasp, is more-or-less right to conclude that he's correct when he describes the elephant's trunk as a snake, for instance. Who can prove him 'wrong' within his limitations? No one. What he perceives is completely consistent with the definition of a snake. You can't prove him wrong to his satisfaction.
Nor can you give him sight to let him see the rest of the elephant.
andy_c,
You're indulging in the same sort of personal attack that janneman accused me of earlier. We shall see if he condemns your post in the same manner.
Somehow I doubt he will.
Quelle surprise...
I posit no supernatural explanations and never have. To state or imply otherwise is: a) demonstrably wrong, b) pretty dumb.
Would you say that physicists know everything there is to know about physics? No? How about physicians...do they know everything about medicine? No? Biology? Economics? Geology? Chemistry? Psychology? Astronomy?...
No?
Then how in the world can you be so arrogant as to assume that everything is known about electronics?
Will someone call the JAES and inform them that their services will no longer be required? There's nothing left to publish, because there's nothing left to discover.
Oh? There is more? Really?
You can't have it both ways, you know.
Check your ego at the door, please. Reality doesn't match your preconceived notion that current electronic theory covers everything. Particularly in light of the strong historical tendency for 'impossible' things to be proven right in audio (e.g. audibility of passive components, absolute phase, etc.).
Don't get me wrong. I would dearly love solid explanations for some of the perceived effects in the realm of stereo, but somehow I don't think we'll be able to talk President Bush into signing a bill giving us fifty million dollars to explore sound reproduction. Lacking a significant influx of research money, the honorable thing to do is to admit that you have something that feels like it might be a snake...but perhaps it really is an elephant's trunk after all.
In actuality, continuing to insist that you have 'proved' it to be a snake only shows that you have not noticed the rest of the elephant...nothing more.
Again: Too much theory. Not enough reality. Reality beats theory, seven days a week.
Though it doesn't stop armchair theoreticians from arguing ad nauseum that they have proved beyond the shadow of a doubt that they have a snake.
I think we might have an elephant. You think it's a snake. Perhaps in time someone will demonstrate rigorously which of us is right. In the meantime, my designs continue to improve, sonically. On the other hand, those who insist that everything that can be known is already known have already achieved distortions at or below the 'proven' limit of human perception for distortion. The only way forward is to admit that either you're doing it simply as an exercise in mental masturbation, since it can be 'proved' that you can't hear distortion below a certain limit, or to admit that it sounds better when you do X or Y even though theory can't explain why, in which case you've admitted that you are hearing things that can't be proven.
So which is it?
(Bob Cordell backed himself into this very same logical conundrum a month or two ago in another thread when he advanced the idea that feedback might be lowering other distortions besides the accepted THD and IM. What distortions might those be, Bob? Unrecognized, unproven ones? Er...uh... Now, granted, Bob doesn't portray himself to be quite as specs-uber-alles as some around here, but it is a cornerstone of the "everything is known" viewpoint that there is a scientifically proven limit to the human perception of distortion. Ergo, anything below that is wasted effort unless you intend to say that the perception limit, itself proven scientifically, is bunk, or there are distortion mechanisms that haven't been elucidated yet, which directly undermines the conviction that we already know everything necessary. There are no other alternatives. Clearly the position is untenable. Speaking for myself, when I got to this point I eventually [after far too long a time] admitted that the entire everything-is-known viewpoint was flawed and started over.)
I have some ideas as to why X or Y sounds better, but if you don't like them because they aren't 'proven' then that's okay. I don't even insist that I'm right on the why, only the observed results, which are observable by anyone who cares to actually listen rather than argue theory. (Galileo: Still, they move. [Though he knew not why.]) They're just working hypotheses and will either stand or fall as empirical results dictate. I'm disposed towards practical solutions, not endless yik-yak about the number of angels that can dance on the head of a pin. I try things and if they sound better I adopt them. If they don't, I drop them. Incidentally, that's a fairly decent counterproof to the oft stated 'you heard what you expected to hear' nonsense. Some of my best (or so I thought at the time) ideas didn't amount to squat, sonically. Had I heard what I 'expected' to hear, they would have been in the "seven veils removed" category...okay, maybe five veils for some of them...sadly, it didn't work out that way. But that's inconvenient for your personal set of convictions so feel free to disregard it.
I didn't even want to end up in my current position. I got here by noticing that if you reach high enough, what at first seems like a tree truck flares out and seems to be attached to something bigger and somehow, no matter how I reach, there aren't any leaves. Something's up, here. I may or may not have an elephant, but it sure isn't a tree.
Grey
I like the word, 'sophomore' . It tells it all!
Many here, for some reason, don't seem to realize that we have held similar beliefs when were too were 'sophomores' and even further up the education trail.
In 1966, I was among the first to use a sophisticated circuit program (ECAP) to prove electronic circuits and we had to use a mainframe IBM computer in those days. I did the modeling from real measurements. Even then, I found the potential pitfalls of relying on a computer simulation, exclusively.
Many here, for some reason, don't seem to realize that we have held similar beliefs when were too were 'sophomores' and even further up the education trail.
In 1966, I was among the first to use a sophisticated circuit program (ECAP) to prove electronic circuits and we had to use a mainframe IBM computer in those days. I did the modeling from real measurements. Even then, I found the potential pitfalls of relying on a computer simulation, exclusively.
Alternative Measurements to THD/IMD
I certainly agree that measuring only an amplifier's distortion into a static load under steady state conditions is far removed from driving an actual speaker with real program material that is of a highly dynamic nature.
A steady state analysis is valid only if the circuit under test is guaranteed to have no "history". In other words, the output signal at every point in time is dependent only on the time shifted input signal at a corresponding point in time. This time shift represents the input - output delay. The presence of a feedback loop invalidates this assumption, because the input to the amplifier now sees two signals: the input and the fed back output signal which has a delay associated with it. Such a system will display different transient and steady state behavior. What would be desirable is a transient test that yields a figure of merit in the time domain just as THD and IM tests do in the frequency domain.
In studying signal integrity at multi-GHz rates, where the system's history may include several dozens of symbol times, we have resorted to a pulse distortion analysis technique. This technique compares an input pulse to the same pulse seen at the output. By appropriately time shifting the two pulses, and accounting for a circuit's gain vs. frequency response, it is possible to determine the amount of distortion in the time domain, as opposed to the frequency, domain.
One thing more to consider: A speaker presents a complex load that can both sink and source current. Since a speaker cone has dynamic mass and inertia, it will backdrive into the amplifier, further complicating measurement efforts. It would be beneficial to quantify a more realistic test load than the simple R or R||C usually used for amp testing.
Finally, it would be interesting to see if perceived sound quality correlates more closely with transient distortion than it does with THD/IMD measurements.
I certainly agree that measuring only an amplifier's distortion into a static load under steady state conditions is far removed from driving an actual speaker with real program material that is of a highly dynamic nature.
A steady state analysis is valid only if the circuit under test is guaranteed to have no "history". In other words, the output signal at every point in time is dependent only on the time shifted input signal at a corresponding point in time. This time shift represents the input - output delay. The presence of a feedback loop invalidates this assumption, because the input to the amplifier now sees two signals: the input and the fed back output signal which has a delay associated with it. Such a system will display different transient and steady state behavior. What would be desirable is a transient test that yields a figure of merit in the time domain just as THD and IM tests do in the frequency domain.
In studying signal integrity at multi-GHz rates, where the system's history may include several dozens of symbol times, we have resorted to a pulse distortion analysis technique. This technique compares an input pulse to the same pulse seen at the output. By appropriately time shifting the two pulses, and accounting for a circuit's gain vs. frequency response, it is possible to determine the amount of distortion in the time domain, as opposed to the frequency, domain.
One thing more to consider: A speaker presents a complex load that can both sink and source current. Since a speaker cone has dynamic mass and inertia, it will backdrive into the amplifier, further complicating measurement efforts. It would be beneficial to quantify a more realistic test load than the simple R or R||C usually used for amp testing.
Finally, it would be interesting to see if perceived sound quality correlates more closely with transient distortion than it does with THD/IMD measurements.
Re: Alternative Measurements to THD/IMD
- Klaus
A strong point you make here. I have successfully used controlled current sources (not with audio, but ATE test-pin drivers) modulated with noise or some input derived signal to create any dynamic voltage/current relationsships. Then, after factoring out the linear part of the distortion (Zout-induced) the distortion seen then is way different (larger) than what one gets with any static load, be it complex or real, even with transient signals. Also, it is a good test for worst-case SOA/power conditions (besides the quasi-short-to-rail test, where many amp's safety circuits fail). It might be exaggerated to assume a speaker to have totally uncorrelated current (WRT magnitude and polarity) but an amp designed to handle that will probably handle real speakers with ease.analog_guy said:
- Klaus
How about an RLC network? It's more realistic than the simple R || C solution, and more economical from a DIY perspective than a complex impedance implemented with active electronics.
It seems to me that any designer of audio equipment who trusts their models (theories & simulations) more than their own ears when judging how realistic something sounds may need more than their ears checking. 😎
A test such as what KSTR suggests primarily tests damping factor across bandwidth.
Ironically (but appropriately) something that feedback improves drastically.
Ironically (but appropriately) something that feedback improves drastically.
Actually come to think of it LC is a little big at 40Hz, so it's not that practical.
Why not just use an actual loudspeaker? You could connect a high common mode input difference amplifier to each of the emitter resistors, and sum the outputs to get the output current of the amp.
KSTR's idea could be programmed to test just about anything. Stability into awkward impedances would be the most significant.
Why not just use an actual loudspeaker? You could connect a high common mode input difference amplifier to each of the emitter resistors, and sum the outputs to get the output current of the amp.
KSTR's idea could be programmed to test just about anything. Stability into awkward impedances would be the most significant.
- it really isn't 1970 still
uh, guys, how about checking the calender?
2007 - 25 yrs => Cherry/Cambrell's papers on "Interface Modulation Distortion"
http://www.diyaudio.com/forums/showthread.php?s=&postid=553489&highlight=#post553489
http://www.diyaudio.com/forums/showthread.php?s=&postid=676793&highlight=#post676793
any sucessful amplifier must have enough current and voltage swing, lack deadzones, dynamic under bias defects, ect. without regard to the application of how much feedback or where in the loop
when the basic amp circuit can control the worst case output load the appication of high feedback can reduce distortions and lower output impedance - Cherry's papers shows that externally sourced current on the amp's output generates similar or lower distortion or IMD components compared to a forward path signal giving the same output
admittedly the Carver Challenge post constitutes anecdotal evidence only, but more than a few people have put some effort into turning sensory impressions into repeatable data and the known psychoacoustics doesn't support the idea that "just listen" is likely to be a valid approach without a lot more controls than anyone here seems to be advocating when criticizing "naive" measurements:
http://www.amazon.com/gp/product/0470869232/ref=cm_cr_pr_product_top
I don't think any engineer here has been proposing we can describe all psychoacoustic effects that might contribute to a sense of realism in audio playback
What we can do is point to Signal theory, what is actually transferred through our audio signal chains and compare the best no feedback amps to the best high feedback amps and show that the low/non feedback power amplifier outputs have much larger measurable deviations both linear and nonlinear over the range of audio/musical signals than high feedback amps with similarly well designed output stages for the applied load
Since the deviations of todays best high feedback amps seem to be below the (incomplete) estimates from psychoacoustcs of audibility it strains logic to claim that high feedback amps sound different Because they Add something that low feedback amps do not
Preferring "the sound" of low/non feedback amps in a signal chain is not the same as establishing that good high feedback amps are adding something bad
uh, guys, how about checking the calender?
2007 - 25 yrs => Cherry/Cambrell's papers on "Interface Modulation Distortion"
http://www.diyaudio.com/forums/showthread.php?s=&postid=553489&highlight=#post553489
http://www.diyaudio.com/forums/showthread.php?s=&postid=676793&highlight=#post676793
any sucessful amplifier must have enough current and voltage swing, lack deadzones, dynamic under bias defects, ect. without regard to the application of how much feedback or where in the loop
when the basic amp circuit can control the worst case output load the appication of high feedback can reduce distortions and lower output impedance - Cherry's papers shows that externally sourced current on the amp's output generates similar or lower distortion or IMD components compared to a forward path signal giving the same output
traderbam said:
admittedly the Carver Challenge post constitutes anecdotal evidence only, but more than a few people have put some effort into turning sensory impressions into repeatable data and the known psychoacoustics doesn't support the idea that "just listen" is likely to be a valid approach without a lot more controls than anyone here seems to be advocating when criticizing "naive" measurements:
http://www.amazon.com/gp/product/0470869232/ref=cm_cr_pr_product_top
I don't think any engineer here has been proposing we can describe all psychoacoustic effects that might contribute to a sense of realism in audio playback
What we can do is point to Signal theory, what is actually transferred through our audio signal chains and compare the best no feedback amps to the best high feedback amps and show that the low/non feedback power amplifier outputs have much larger measurable deviations both linear and nonlinear over the range of audio/musical signals than high feedback amps with similarly well designed output stages for the applied load
Since the deviations of todays best high feedback amps seem to be below the (incomplete) estimates from psychoacoustcs of audibility it strains logic to claim that high feedback amps sound different Because they Add something that low feedback amps do not
Preferring "the sound" of low/non feedback amps in a signal chain is not the same as establishing that good high feedback amps are adding something bad
jcx,
Well said. And thanks for the link to the book, although at 123 $ I'm hesitant to buy it...
Jan Didden
Well said. And thanks for the link to the book, although at 123 $ I'm hesitant to buy it...
Jan Didden
It seems that in one practical complete audio amplifier, it is difficult to judge whether it is the feedback that is adding something bad. Because the blocks inside one amp is not perfect yet. Perfect input differential stage, perfect VAS, perfect driver, perfect output stage. If we can assure that each block is already perfect and linear, then we can judge whether feedback is contributing bad things to the sound. If we judge feedback inside one imperfect audio amp (that has non-perfect blocks), we don't know whether the bad sound is due to feedback or the imperfection of each block wrapped in global feedback.
jcx wrote:
I really don't approve of underestimating people's ability to tell real from fake. Barring medical impairments, old age or too many Tangerine Dream concerts in the front row, my experience is that our audio perception is extremely fine.
What isn't extremely fine is the correlation between electrical or simulation measurements that are in the public domain and the realism of the reproduction.
Isn't that obvious?
I think this paragrah has gone off the rails of reasoning and ito the pit of conjecture. 😉
I really don't approve of underestimating people's ability to tell real from fake. Barring medical impairments, old age or too many Tangerine Dream concerts in the front row, my experience is that our audio perception is extremely fine.
What isn't extremely fine is the correlation between electrical or simulation measurements that are in the public domain and the realism of the reproduction.
Isn't that obvious?
- Status
- Not open for further replies.