Sound Quality Vs. Measurements

[vacuphile]

Originally Posted by bcarso

A not-too-bad short discussion: Stochastic resonance - Wikipedia, the free encyclopedia

Yep, good summary, that's why I have to disagree with A.wayne. If anything, some noise would help to bring the faintest detail out, rather than obscure it.

There is another reason why some noise in the signal might contribute to improvements, that is on the side of the sound source. No transducer is completely devoid from hysteresis. Keeping it ever so minutely in motion by random noise will also randomize the distortions caused by hysteresis, or in other words, decouple them from the signal.

The first para is based on accepted science, the second on my conjecture, so take it with a grain of salt. At any rate, for me it must be part of the explanation why ClassD sounds better than it should.

vac

[nigel pearson]

The Japanese wording I gave was too general . This music approximates . However more installments would be used and one artist . A bit off beam from the discussion . Hope some take a look . FAL does not conform to some accepted standards ( frequency response dips , mic too close , standard test , standard problem I would say ) . When listening they seem to contradict the graphs .

Eitetsu Hayashi ? Bolero part1 ???????? - YouTube

[a.wayne]

Measure them while being driven by his amplifier and it may not look so contradictory.

Funny thing when you voice speakers to match your setup .......

[a.wayne]

Originally Posted by vacuphile

Yep, good summary, that's why I have to disagree with A.wayne. If anything, some noise would help to bring the faintest detail out, rather than obscure it.

There is another reason why some noise in the signal might contribute to improvements, that is on the side of the sound source. No transducer is completely devoid from hysteresis. Keeping it ever so minutely in motion by random noise will also randomize the distortions caused by hysteresis, or in other words, decouple them from the signal.

The first para is based on accepted science, the second on my conjecture, so take it with a grain of salt. At any rate, for me it must be part of the explanation why ClassD sounds better than it should.

vac

The noise floor is already delivered by the recorded medium, adding more via the electronics is good ...?


Err no cant buy into that ....

[nigel pearson]

I tried a standard almost blameless amp with them ( damping factor circa 90 ) . They were fantastic . FAL agent was stunned and said if only his friend would listen to transistor amps . It was mostly the same sound .

I take your point . Fostex are interesting if trying current drive . The are fine in voltage drive also .

The amp was my own variation of the Hitachi MOS FET design with some care taken over the VAS . I tried Szikai pairs and FET in a variation recently and was stunned to find I preferred the Szikai . The measurements were very slightly different . The sound was surprisingly similar . I stuck with FET's as I like the simple biasing ( 100 mA is a good place to start ) Exicon if wondering .

[nigel pearson]

I think Hypex have proven the point about class D . They use 5532 ( or 4 , it was a couple of years ago ) . We replaced some and were surprised how good the 5532 is . As they are SMD it is a difficult task . The op amp can have a separate supply . Try it . I think they do have slight bus pumping issues and are better in bridge perhaps .

Genesis in USA have a very nice version .

[dvv]

Originally Posted by DF96

I thought speaking against the NE5534/32 was the true religion. Now you have confused me. Should I use these chips everywhere, or nowhere? Or should I make engineering judgements based on facts, and ignore the 'fashion engineers' on both sides?

DF96, frankly, I don't really care what you or anyone else uses anywhere. That is a good op amp, however, it was a killer a long, long time ago, when it first appeared in the late 70ies.

Since then, over the last 30 years, I believe we've all learnt much and new, sonically superior (in my view) op amps have appeared. I don't even know what the current fashion in op amps is, nor do I care, because I do not run with the "in" crowd.

But I have followed up on some which appeared to be promising, notably those which offer good speed, but low HF distortion; I find that too many "in" op amps produce inordinate amunts of HF distortion even if they shouldn't. In the end, I do keep a stock of four op amps I sometimes use for audio, although my preferences still rung strong towards the discrete circuits.

I think I better not name them, else I might be accused of advertising, since they all come from the same manufacturer. Three I know to be results of Walt Jung's involvement (to what extent I don't know), the fourth is an older one, and I don't think he worked on that one (but I could be wrong).

So you go right ahead and use whatever you feel is worth using, but do not each and every time take offence when somebody says something you don't agree with.

[a.wayne]

Originally Posted by nigel pearson

I think Hypex have proven the point about class D . They use 5532 ( or 4 , it was a couple of years ago ) . We replaced some and were surprised how good the 5532 is . As they are SMD it is a difficult task . The op amp can have a separate supply . Try it . I think they do have slight bus pumping issues and are better in bridge perhaps .

Genesis in USA have a very nice version .

CLass-D , The camels dangles ..... ..

[DF96]

So I should not take offence when someone has a dig at UK engineers and their alleged worship of NE5532, or a dig at all PhDs and their alleged lack of practical skills, or when those who understand and believe standard loop theory are suggested to have been off the planet for so long that they have never heard of a disputed paper by Otala? Should I just ignore such nonsense?

[benb]

Originally Posted by bcarso

A not-too-bad short discussion: Stochastic resonance - Wikipedia, the free encyclopedia

I may have heard the phrase before, but wasn't aware of what it meant. With the word resonance, it sounds a little bit new age, but it's not:

Stochastic resonance is observed when noise added to a system changes the system's behaviour in some fashion. More technically, SR occurs if the signal-to-noise ratio of a nonlinear system or device increases for moderate values of noise intensity. It often occurs in bistable systems or in systems with a sensory threshold and when the input signal to the system is "sub-threshold"

That description sounds quite familiar, in dithering of A/D's (which it mentions) and bit reduction within the digital domain, and in ac bias in magnetic tape recording (which it does not mention). I'm surprised to find it to be so recent:

Stochastic resonance was discovered and proposed for the first time in 1981 to explain the periodic recurrence of ice ages.[2] Since then the same principle has been applied in a wide variety of systems. Nowadays stochastic resonance is commonly invoked when noise and nonlinearity concur to determine an increase of order in the system response.

This National Semiconductor app note on dither is dated 1992, but has references that go back to the 1960's and 1950's:
http://www.ti.com.cn/cn/lit/an/snoa232/snoa232.pdf

The nonlinearity of magnetizing a domain is both large and has a large amount of hysteresis. The bias signal fixes both of these. I wonder how well an ultrasonic noise source would work compared to the single-frequency bias signal used in tape recorders.

There are many nonlinear devices that might benefit from the addition of either a constant single-frequency "ac bias" signal or a (probably noise-shaped) "dirher signal," such as audio transformers (though they are often used, especially at mic level, BECAUSE of the distortion they generate), And while we're at it, power amplifiers as well.

Surely someone has tried this. Class D is kind of like this, but that is "linearizing the extremely non-linear" for the sake of electrical efficiency. Could "ac bias" or "dithering" improve a power amplifier that is already "almost perfectly" linear?