Sound Quality Vs. Measurements

[DF96]

Originally Posted by a.wayne

Err, what is whitebox, blackbox testing ....... ?

Originally Posted by DF96

Black box testing just compares outputs with expected outputs, but does not enquire about what is going on inside the box. White box testing deliberately sets out to explore and break the internal mechanisms of the box.

THD is a black box test. "Can I make the input LTP clip" would be a white box test.

[KSTR]

Originally Posted by SY

Which is completely predicted by conventional engineering. So, what's your point?

I'm pretty sure that Thorsten was *not* referring to any differences in SPL mag and phase response, rather he relates to the differences *once* SPL mag/phase has been made equal to the same target, no matter what complex impedance profiles the amp outputs had.

This also implies that first order small signal impulse responses of transducer voltage, and current resp, are the same (as they must be because of identical SPL impulse in a causal system), yet large signal behaviour and the "fine print" can and will be different (measured HD/IMD shows part of that). How the "microphonic" velocity signal generated by the transducer will be handled is what makes the difference, which is not always better with higher impedance drive. Velocity-controlled operation (instead of force-controlled) still has it's benefits and applications. Similar to amplifier circuits, when you choose to apply feedback (in the transducer) then use lots of it, and never use it to "fix" a mediocre transfer function -- Baxandall strikes back.

BTW it was a post of Thorsten over a decade or so in some forum that first made me aware of this interface mechanism and rethink the common view. I owe him something...

An electrodynamic transducer itself is a feedback system under voltage drive, degenerative feedback that is. When you change the feedback factor (as a scalar, or even frequency dependent), all else equal, there are more or less subtle changes in sound, depending on how nonlinearities/noise and reflections get "recycled" by the feedback, if any.

- Klaus

[Wavebourn]

Originally Posted by DF96

THD is a black box test. "Can I make the input LTP clip" would be a white box test.

Or, like "What causes raise of harmonics with increased signal level: first tube, phase splitter, or output stage". Or, "How phase will be modulated by the stage on 20 Khz, let's check applying 300 KHz for better visibility". Or, "How crossover distortions will be modulated by signal envelope", and so on.
Or, "How beta of output transistors depend on current". Lots of measurements, depending on topology. In order to find optimal topology, parts, regimes.

[BFNY]

Originally Posted by DF96

I remember many years ago, when I was a computer programmer, learning about black box and white box testing. Black box testing just compares outputs with expected outputs, but does not enquire about what is going on inside the box. White box testing deliberately sets out to explore and break the internal mechanisms of the box. We need both for amplifiers. So far, we have mainly been talking about black box testing.

Probably the best example of white box testing one can do for most "commercial" amps with NFB is open loop testing.

It is always most interesting to test open loop distortion, and also gain. Then also gain /phase margins which gives a big measure of inherent stability.
I have read many well heeled amp designers speak of getting the design "right" with no feedback, i.e. get to the lowest distortion, best response, etc, and then closing the loop with moderate feedback to lower output impedance, etc. This makes sense, to me at least.

[SY]

Originally Posted by KSTR

I'm pretty sure that Thorsten was *not* referring to any differences in SPL mag and phase response, rather he relates to the differences *once* SPL mag/phase has been made equal to the same target, no matter what complex impedance profiles the amp outputs had.

Hard to read minds, I can only go by what's written. In any case, the engineering of current drive systems is well-established, with several JAES papers and at least one book working out the principles according to conventional engineering. And experiment indeed accords with theory. The differences beyond SPL mag/phase are not exactly gross, but certainly measurable and understandable.

[Wavebourn]

Originally Posted by BFNY

... and then closing the loop with moderate feedback

Why moderate?

[BFNY]

Originally Posted by Wavebourn

Why moderate?

I believe the stated reason is large amounts of NFB are not needed as much if the original open loop design is inherently low in distortion. I think many of the Pass DIY amps like the F5 use this sort of feedback philosophy.

Versus giving it 100dB of open loop gain, and then feeding back 70-80dB or more to "fix everything". Not really saying one is right or wrong, but it's interesting to look at it in a white box fashion.

For instance, most op-amps have over 100dB of open loop gain. Versus a 6SN7 class A tube stage with no FB and perhaps 20dB gain open loop. Two very different topologies.

[DF96]

Two different topologies, and both seem to work when properly engineered and used.

The right amount of feedback is not necessarily 'moderate' or 'modest' or 'massive' or 'none' but the right amount for that circuit in that application. Getting open loop behaviour correct before closing the loop is obviously right, but 'correct' does not necessarily mean 'OK as an audio amp without feedback' although this may indeed be the situation with some circuits.

[CopperTop]

Originally Posted by SY

Don't you ever get tired of saying that? For the perhaps 10,000th time, there is no-one who has ever claimed that.

Quad have always had a very straightforward engineering view of their products, and insisted that all amplifiers of adequate quality sounded the same when used within their capabilities, and that speaker cable had no sound at all, unless ludicrously long and thin wire is used.

To verify or disprove that amplifiers of adequate quality sound the same, Quad commissioned James Moir to organise and conduct listening tests comparing Quad II, Quad 303 and Quad 405 amplifiers. Statistical analysis of the expert listening panel's scores showed that "the decisions of the panel were no better than might be expected from sheer chance".

Quad Electroacoustics - Wikipedia, the free encyclopedia

[SY]

Again, that's a qualified statement (and the way it's worded, rather tautological). It is NOT "all amps sound the same."

This seems to be a difficult concept.