Sound Quality Vs. Measurements

[ThorstenL]

Hi,

Originally Posted by CopperTop

And yes, the question is precisely that: could a null at -110dB, as measured by a state of the art piece of equipment be considered 'perfect', or would there still be room to quibble?

There would be room to quibble. In some areas -110dB may be insufficient, in others it may be too stringent.

Originally Posted by CopperTop

My main concern is that I suspect there are people who would not trust an amplifier that had op amps and digital elements in it, on principle. (I'm not completely immune to those thoughts myself!). If the test equipment itself is built like that, could they accept its findings?

That depends on what we wish to find.

If I desire to reliably know the steady state THD of a DUT and the breakdown by harmonics then the AP2, with all it's op-amp's etc. is a perfectly decent choice. Of course I can accept it's findings, no matter how it works.

But, something like the above is ALL that the AP2 can give you. It can give you a quantitive answer to a specific question, regarding a specific quantity.

Originally Posted by CopperTop

And another question I'm trying to tease out, is that some people think that an audio system's job is to actually add something that makes the final sound at the speakers more like listening to a live performance than the recording itself

This, strictly speaking is not possible. You CANNOT make a recording that is "better than live" objectively, however, subjective preferences are another thing. It is possible to make something sound more "likable".

Originally Posted by CopperTop

many threads contain such suggestions e.g. the measured distortions of vinyl and tubes serendipitously enhance the sound. A true difference of zero would mean no added 'musicality'.

Only a complete and utter ejit who has no idea about how the human hearing works would make such a suggestion as anything but troll bait.

It has been said here many times by many, adding distortion does not give "vinyl sound" or "tube sound".

If it did, we all would be listening to CD's produced with the usual low-end pro-sumer garbage gear (which some mistake as "Pro" - the relation is about like that between a Hasselblad and a Sony "Megazom 30 Megapixel" point and shoot camera) and with suitable DSP Plugins applied and gush how "analog" things sound.

Ciao T

[KSTR]

Originally Posted by CopperTop

If the input and output of a hypothetical amp were compared (with any real speaker load on the amp), and found to give a difference of zero with any and all test signals, including real music, to the limit of an AP2's resolution, could we declare that amplifier to be 'perfect', for all practical intents and purposes?

I would think that can only work if you use linear models for both complex frequency response (mag and phase) and complex output impedance, to get to any possible nulling in the first place.

Otherwise there will be a conceptual difference (the linear response variation caused by actual loudspeaker current producing a I*Z drop along the output impedance) as well as an overlaid swamping nonlinear distortion, the distorted nature of that speaker current. Even with a perfect amp (say, perfect opamp, single pole model) with truly ohmic Zout of 10mR the test would fail, depending on the real-world speaker load. And a complex dummy load still provokes linear distortion along a perfect Zout that must be acutely dealt with first in order to see the real (nonlinear) differences of the amp vs its linear model.

[SY]

Originally Posted by CopperTop

If the input and output of a hypothetical amp were compared (with any real speaker load on the amp), and found to give a difference of zero with any and all test signals, including real music, to the limit of an AP2's resolution, could we declare that amplifier to be 'perfect', for all practical intents and purposes?

That's less important than the question, "At what level will human listeners be unable to distinguish differences between output and input by ear only?" At that point, it's as perfect as it needs to be. That unfortunately would present a problem for marketeers, magazine writers, and would-be gurus...

[ThorstenL]

Hi Sy,

Originally Posted by SY

That's less important than the question, "At what level will human listeners be unable to distinguish differences between output and input by ear only?"

I am shocked that I actually agree with you. Of course, tests used to establish such limits would need to be carried out to quite seriously scientific standards.

The only such tests I am aware of having been done re. audibility to sufficient standards regarded perceptual coding (MP3, AC3 etc.) and we can learn a fair bit from these tests, however as they had subtly different questions from those you and I listed independently (and phrased differently) we cannot directly extrapolate.

Originally Posted by SY

At that point, it's as perfect as it needs to be.

Again, we agree.

Originally Posted by SY

That unfortunately would present a problem for marketeers, magazine writers, and would-be gurus...

I cannot see why this would be the case. If the tests to establish audibility are done with due care, diligence and controls I see no issue. Issues only appear when unscientific nonsense is being put forward under the mantle of being scientific and actually offering evidence of the true situation, when nothing could be further from the truth.

When I moved to the UK at least two of the magazines published the results of blind preference testing of amplifiers, cd-players etc. (usually in the "Budget Class") each and every month. Did they find everything sounded the same and no differences existed?

Hardly...

Also, I did post a summary of a blind test I did a page or so back. The silence on your part regarding this has been deafening. No suggestions what we did wrong and why we reliably heard differences that could not be measured?

Ciao T

[SY]

Originally Posted by ThorstenL

Also, I did post a summary of a blind test I did a page or so back. The silence on your part regarding this has been deafening. No suggestions what we did wrong and why we reliably heard differences that could not be measured?

I'm going to say something in all seriousness, and if you like, it's also directed to dvv. Engineers, especially those with a stake in the outcome (not of that particular test, perhaps, but of that sort of test in general) rarely have the knowledge and background to design and set up proper sensory tests; sometimes that leads to deliberate error, sometime to inadvertent error. The very best thing they can do, lacking the on-site availability of a sensory test expert, is to hire a magician. If the magician is a specialist in mentalism, so much the better. Let the magician set up the blinding, scoring, and tabulation protocols.

[dvv]

Originally Posted by SY

I'm going to say something in all seriousness, and if you like, it's also directed to dvv. Engineers, especially those with a stake in the outcome (not of that particular test, perhaps, but of that sort of test in general) rarely have the knowledge and background to design and set up proper sensory tests; sometimes that leads to deliberate error, sometime to inadvertent error. The very best thing they can do, lacking the on-site availability of a sensory test expert, is to hire a magician. If the magician is a specialist in mentalism, so much the better. Let the magician set up the blinding, scoring, and tabulation protocols.

Two things related to DVV here:

1. I get to walk away from this because I'm not an engineer, and

2. As far as I am aware, you have not stated what was wrong with what we did. I may have missed it, and if I did, some directions, please.

On a personal note, I resent your attitude towards my cousin Thorsten. Yes, cousin. Like most things in this thread, THEORETICALLY all people of German descent are related from God know how far back, and THEORETICALLY my 25% of Austrian blood (Austrians are just expatriate Germans, trust me) makes me a distant relative of Thorsten, many times removed.

Exactly the same thing applies to Wavebourn - we are also very probably related, some 13+ (and possibly as far back as 23+) centuries ago, but hey, THEORETICALLY blood is a lot thicker than water, even if many times removed.

If you are Caucassian, you should apply much more tact and care when addressing your THEORETICALLY kinsmen; the archetypical "white man" is quite rightly called Caucassian, and the Caucassus mountains are in what is Russia today. Slavs originate very directy from right that there ground zero, with a big move in the 7...10th century, when Slavs moved into the Balkans, to become Serbs, Croatians, Slovenians, Macedonians and Montenegrins later on.

When Serbs moved in here, they were pagans, and it was two Greek Orthodox Christian priests who gave us our alphabeth (one was named Cyrrilus, hence the name Cyrillic alphabeth), our faith and some of our customs. To this day, I can read Greek, even if I don't understand most of it, so THEORETICALLY, Greeks are also my cousins. Also Wavebourn's, because in his native lanugage he also uses Cyrrilic, so he too is THEORETICALLY a cousin to the Greeks.

ANYTHING is theoretically possible, even such cosmic events like finding a grain of honesty in a politician's head.

With so much theory being flung around, no wonder so few mention actual, real life issues any more.

Which brings me back full circle to my question from above: what did my friend and I do wrong in our internal wiring test?

[SY]

You didn't consult someone who knows how to set these tests up properly. The information you gave missed nearly all of the critical points in setting up a good controlled test, so the chances are that these weren't considered. The statistics you gave were inconsistent with your descriptions of the magnitude of the differences. I'm not being critical of you as an engineer- this is just not something engineers know how to do any moreso than sensory psychologists know how to set up the bias arrangements on a source follower.

Next time, hire a magician, then once a tight protocol is established, follow it to the letter. You're making an extraordinary claim and that requires particular care and transparency in how the data are obtained.

[ThorstenL]

Hi Sy,

Originally Posted by SY

I'm going to say something in all seriousness, and if you like, it's also directed to dvv. Engineers, especially those with a stake in the outcome (not of that particular test, perhaps, but of that sort of test in general) rarely have the knowledge and background to design and set up proper sensory tests; sometimes that leads to deliberate error, sometime to inadvertent error.

I was quite candid with the test conditions.

If you have any specific issue you may wish raise I will be happy to hear it and see if the concern is valid or based on misunderstanding.

Originally Posted by SY

The very best thing they can do, lacking the on-site availability of a sensory test expert, is to hire a magician.

Well, now any subjective test not set up by a "sensory test expert or a stage magician" are automatically invalid (howls of anguish and despair from the ABX Mafia)?

Or are they only invalid if YOU dislike the results, or worse, the implications of the results?

Originally Posted by SY

hire a magician. If the magician is a specialist in mentalism, so much the better.

Enquiring minds do want to know, if you had particular magician in mind (like the Amazing Randi who can magically make any differences between anything disappear, or at least as top escape artist escape having to pay up) and what specific function the Magician would full that for example =ROUND(RAND()*(3-1)+1,0) in Excel (for my case of testing three different items) cannot?

Or is the Job of the Magician to create the illusion that real differences have suddenly all disappeared?

Ciao T

[GloBug]

[SY]

A sensory testing expert is ideal but expensive- a good magician is considerably cheaper and will eliminate most of your potential control errors. I seriously doubt you could afford James Randi, but I'm certain that there are plenty of decent magicians to be hired in London.