Mean Thoughts on Measured and Perceived Performance.

The audio signal is fundamentally a wave representing changes in Amplitude and Time.

The amplitude domain.

The scant, convenient industry standard definitions and protocols of distortion measurement substantially fail to correspond with the subjectively perceived sound.

In reality everything happens in time.
Distortion analyses disregarding the time dimension and transfer function linearity suffer from severe restrictions.
Pure sine waves are vastly far away from the demanding, massive complexity of music signals, frequency based bench tests (where linearity and time are mathematical constants), steady-state conditions, in absence of dynamic variables and reactive load offer misleading hints of performance in the field.

Harmonic distortion (THD) is an added content of integer multiples to the fundamental tones, caused by amplifying device nonlinearity.

Having its biologically founded attention elsewhere, the human ear is largely insensitive to harmonic distortion, levels up to 10% are not audible, depending on its spectrum, sound pressure level and phase relationships. (The hearing evolved primarily for self-defense having an immense dynamic range, speech emerging later on, enjoyment of music is an evolutionary by-product).
The cochlea generates considerable intrinsic harmonic distortion in a specific pattern and entirely compresses harmonics within that pattern. Consequently, only external harmonics being outside are dissonant and hearable and only amplifiers producing a conforming distortion spectrum will have natural tonality. Meaningful measurements should give detailed account in this regard.
(For instance the self-generated second harmonic at 90dBA is 10%).
High order harmonics are increasingly more detectable and discordant. These are also accompanied by disagreeable intermodulation distortion products.
As not all harmonics are equally objectionable and as their inherent relative structure is of crucial importance, should not be reported as a group referenced to the fundamental, rather to each other aiming a strong weight for the more revealing high order harmonics.

Exhibiting absolute values of THD is exquisitely irrelevant and completely worthless.

Any attempt to reduce THD imposes merciless punishment as limitations and disturbances in dynamic signal handling.

Since THD is easily measurable and low values are easily (and cheaply) achievable, manufacturers have suppressed it to vanishingly small levels by huge amount of global negative feedback (GNF) and used it for marketing purposes as a proof of excellent sound quality. Those amplifiers, necessarily, must sound horrible. In different approaches endeavouring towards results of subjective preference, some of us may find, that rather high values of THD indicate good sound quality.
While the commercial presentation of THD may be explicable, there's something pathetic about DIY-people having no Sales Department widely considering simulated THD figures as a universal measure of goodness. They are definitely not, just meaningless numbers merely making sense in marketing literature to fool buyers.

Frequency and phase distortion are result of nonuniform amplification caused by coupling circuits.

Intermodulation distortion (IMD): not harmonically related signals created from different frequency input signals.
The SMPTE standard does not weight IMD products.

The time domain.

All amplifiers develop some forward-propagation delay.
The current in bipolar transistors depends on the movement of both electrons and holes, a condition constituting time delay. FETs are unipolar, their operation requires only one kind of charge carrier and are much faster (but have low Gm near cut-off meaning unlinear transconductance).
The time delay based distortions (displaced zero crossing points on the X-axis) are most bothersome, like the odd order crossover distortion and the wide variety of distortions provoked by GNF (originally intended to cancel distortions).

Transient intermodulation distortion (TIM) is caused by network anomalies, chiefly by GNF, but also by high frequency performance and power supply insufficiencies, when amplifier components cannot handle the rate of change of the signal.

Slew rate induced distortions (SID) originate in missing bandwidth representing high dynamic impedance properties at these frequencies. In conjuction with GNF, the slew rate induced phase errors are the direct cause of TIM and dynamic intermodulation distortion (DIM).

The slightest amount ot GNF in a multiplicative manner converts benign harmonics into a multitude of high order harmonic and intermodulation distortion products bringing about a serious deviation from the aural harmonic pattern and raises the signal modulating noise floor.

In power amplifiers GNF introduces nasty distortions during the delay period and the more persistent recovery period exposing the amplifier for sharply increased susceptibility to overloading due to maintained open-loop gain, violent overshoots due to reduced bandwidth and slew-limiting, abrupt clipping due to constricted dynamic margins, furthermore phase intermodulation distortion (PIM). Clipped waveforms give rise to voltage peaks in the input stage (a difference between the input and the clipped output voltage), which progressively overdrive the amplification stages (increasing the clipping). Under paralyzing cut-off conditions no signal is fed back, ugly waveshapes appear...

GNF does not prevent errors, it can only reduce errors once they have occurred and are present on the output being restrained by the time delay. Having lost validity, the delayed (thereby false) correction signals make exacerbating damages. GNF is an inaccurate and harmful mechanism.

Closed-loops induce instability, compensating for the phase shift costs precious bandwidth and sound quality.
The pronounced sensitivity to RF interferences is another source of significant sonic degradation.

The resulting attributes: a hard, cold, grainy sound, compressed dynamic range, loss of detail and transparency and decreased accuracy in spatial perception.

For clean, unharmed propagation and processing, the signal needs a straight, unmixed and undivided route in space and time, which is the single open-loop. (Another example of deteriorating multiple signal route is paralleled devices). GNF comprises signals heavily distorted by the output stage making open-loop linearity, gain and bandwidth to change with the constant varying load. It is a highly polluted signal processing.

The open-loop characteristics decisively determine the subjective quality, grandiose closed-loop measurement data do not.

The idea of relying the open-loop linearity of audio amplifiers on GNF is deeply deplorable.

There have been speculations about the proper amount of GNB concerning slew rate, rise-time and bandwidth. Obviously, the GNF generated effects are in no way avoidable, each and every dB will inevitably lead to lost sonic properties called musicality.
Musicality implies a collection of subjective values, which can not be mathematically or technically quantified being far beyond the ability of all methodologies.
The ear recognizes the tiniest elements of acoustical signal with astonishing sensitivity, interpreted by the brain`s tremendously refined information processing. Assuming measurement techniques to ever be approaching such a degree of analytical capability would be extremely naive.
The only reliable and truthful test method is, and always will be, careful listening.
The best sounding amplifiers do not have good measured performance. The chances of obtaining optimal subjective results based on measurements are not quite zero, but close by.
 
Hi,

while I usually keep myself out of the usual flame wars, your post looks scientific, but isn't really. I reads more like what manufacturers write in their brochures - the manufacturers that abandoned THD specs for their 'true' Hifi-customers.

Well, I don't want to argue, just point out a couple of things that really don't fit the picture:

Pure sine waves are vastly far away from the demanding, massive complexity of music signals

Fourier-theorem proves that every signal can be decomposed into a sum of harmonic waves. The Fourier-theorem is a scientific fact and contradicts your statement.

the human ear is largely insensitive to harmonic distortion, levels up to 10% are not audible

I would love to know which study did prove that. If you refer to PMAs posting in the Blowtorch-thread, then this is again untrue, as PMA refered to 10kHz, where already the first harmonic is at 20kHz - a frequency most mature people cannot hear anyway - or only strongly reduced in amplitude.

Exhibiting absolute values of THD is exquisitely irrelevant and completely worthless.

Well, nobody here will say THD is the holy grail.

However, if you don't care wether an amplifier is able to reproduce a single frequency without too many distortion products, feel free to think so. I personally don't care too much about THD specs, but I want to have them at a low number, since this is an easily avoidable error in reproduction.

Any attempt to reduce THD imposes merciless punishment as limitations and disturbances in dynamic signal handling.

You do not give any reasoning why this should be so.

So what you basically say is that a great sounding amp must produce lots of THD? Distortion is distortion and an error in reproduction - so you say, it excels in sound reproduction, but makes large errors at the same time? Come on!

manufacturers have suppressed it to vanishingly small levels by huge amount of global negative feedback (GNF) and used it for marketing purposes as a proof of excellent sound quality.

So you throw all manufacturerers and hobbyists into the same pot?

There are a couple of nice amps with low THD and without global NFB ready to build here on DIYAudio. Maybe you want to care to look at them?

Having lost validity, the delayed (thereby false) correction signals make exacerbating damages. GNF is an inaccurate and harmful mechanism.

You do know the time scale of the delay you're talking about? In a 2 stage amp the delay can be of the magnitude of microseconds, thus only the MHz regime is significantly affected.

Obviously, the GNF generated effects are in no way avoidable, each and every dB will inevitably lead to lost sonic properties called musicality.

I strongly invite you to build MikeBs SymAsym, an amplifier with quite a bit of NFB - and a lot of people like its sound. Some of these people have also build low NFB designs and can compare.

While I really want to avoid a heated discussion, which probably cannot be avoided, I think that many of your statements have as little scientific ground (especially your associations about sound and global NFB) as the marketing tunes you criticize.

Maybe you give it a thought.

Have fun, Hannes

PS: if you are just searching for somebody to argue with, I'm not at your disposal.
 
the text looked promising at first but the more I read the more sceptical I become..
Hmm as a DIY'er i tend to use the IRL method, and it works surprisingly well in detecting not so good sound so the ear cant be completley worthless :) hehe
Just another thought while I'm at it and I dont know about you but I find sometimes people is beeing very restrictive in their thoughts for exampe regarding different devices used.. maybe this thougt is beeing a little offtopic here but I'll say it anyway since it's about people..
First thing one must do to create miracles is step down and realise one does'nt know everything and TRY stuff even if it makes little sense, like using irf610 as input devices ;-)..
and now experimenting with fdv301p and 302n it might not work well but the interest here seems to be low about new stuff.
I mean why would I want to try to construct an amp with the classical old matched jfet's or mat01,02, it's been done maybe a thousand times before... OK i might build one if I wanted to be "sure" it sounded good but that's not really diy is it??
SO what i wanted to say is, life evolved though errors in development? right? so open up a little , try new components and stuff :-D.. .
And dont bite my head of too hard now..
 
We can distill this down to just a few lines:

I don't really understand the Fourier transform.
I don't really understand the concept of scale.
I'm not actually familiar with the literature.


There. Isn't that simpler and easier to read? I suspect, however, that someone is just testing out the audio version of the Postmodernism Generator.
 

gtforme00

Member
2004-10-21 4:35 am
It is not my intention to enter here into the extensive debate on the conceptual foundations of quantum mechanics. Suffice it to say that anyone who has seriously studied the equations of quantum mechanics will assent to Heisenberg's measured (pardon the pun) summary of his celebrated uncertainty principle:

We can no longer speak of the behaviour of the particle independently of the process of observation. As a final consequence, the natural laws formulated mathematically in quantum theory no longer deal with the elementary particles themselves but with our knowledge of them. Nor is it any longer possible to ask whether or not these particles exist in space and time objectively ...

When we speak of the picture of nature in the exact science of our age, we do not mean a picture of nature so much as a picture of our relationships with nature. ... Science no longer confronts nature as an objective observer, but sees itself as an actor in this interplay between man and nature. The scientific method of analysing, explaining and classifying has become conscious of its limitations, which arise out of the fact that by its intervention science alters and refashions the object of investigation. In other words, method and object can no longer be separated.

Along the same lines, Niels Bohr wrote:

An independent reality in the ordinary physical sense can ... neither be ascribed to the phenomena nor to the agencies of observation.


There. I can play the game too.
 
Hi

I believe I read in this paper just yesterday, that Bob came to the conclusion that GNF does not affect PIM, but rather helps. Mr. Cordell is a respected designer and his work has been a great source of information that has helped me immensely, and others I'm sure, navigate my path in audio design. This information I will believe until someone can prove otherwise. Thanks Bob!:)

I will agree that too many stages with-in a GNF loop is a bad thing, generally, but negative feedback is certainly not a bad thing. This argument will continue to the end of time.
 
Lumba Ogir said:
Jan,
I am a little bit curious: what do you dislike most in the story?
It is controversial so I actually did not count on warm applause...


Dislike? I didn't say I dislike it. As a story, its nice. That's all. It's just an anecdote of opinions, not backed up by any expertise or knowledge. That's the definition of a story, no?

Jan Didden