The Need for Objective Metrics in Audio Reproduction

[March Audio]

The items I linked to are actually metal oxide varistors. They are devices that are normally high resistance but once the voltage exceeds a certain level they go low resistance effectively shorting the spike out.

Your picture is a bit too fuzzy to figure out what's going on but those do look like caps.

 

[Thomas savage]

yea its all painted black so what you see in the picture is what it looks like. no detail missing. the varistors works like a thermistor in that its resistant until a certain temp\voltage then its not?

they looked like the soft start thermistors in my amps and my balanced tx. thanks for explaining the difference.

 

[March Audio]

A thermistor will vary it's resistance proportional to temp but a varistor remains high resistance until a threshold voltage is reached.

https://www.google.com.au/search?q=...=chrome-mobile&ie=UTF-8#imgrc=eGSrLvCXhrqkZM:

http://www.electronics-tutorials.ws/resistor/varistor.html

 

[DonH56]

Spectral plots over power (from low to clipping) would be nice to see, as would multi-tone IMD or NPR plots, and impulse and/or step response, using a greater variety of loads. It would be nice to see group delay plots and such as well. It has been a while since I have worked as a tech and taken those sorts of measurements on audio equipment, but it was (and probably is) flat-out scary how some amplifiers react to reactive loads, and even scarier what some amplifiers (and preamps) did to the signal as they clipped. At least we (hopefully) no longer have opamps in there that invert when overdriven, that was fun...

 

[h.g.]

Once you are down to 0.01% versus 0.005%, it doesn't matter that THD is not psychoacoustically correct. Both values are too low for audibility.

Consider distortion that takes the form of a blip or click that is audible and occurs now and again. Something like coil rubbing perhaps. Varying the number of blips per unit of time will enable you to have an arbitrarily low THD value with audible distortion. THD is a reasonable measure for some kinds of distortion but not others.

 

[AJ Soundfield]

What type of measurements would help?

Soundwaves. The thing our ears>brain need to process sound.

Who performs those?

Anyone with a microphone.
The issue is correlation with perception, which is fairly well established...and preference, which is ongoing. All the measurements will tell is some degree of statistical trend. They may not tell you what you will prefer, especially sighted, which for the better or worse, is part of the end user total subjective experience.

cheers

AJ

 

[TBone]

What metrics should these be? What should an audiophile look in those measurements? What type of measurements would help?

Measuring equipment is but one thing, measuring what the equipment plays yet another ....

example: Album, Hotel California, Eagles; song: Hotel California ... (foobar DR ratings)
DR15: 1976 orig LP.
DR15: 1984 orig CD.
DR15: 1987 Japan CD.
DR13: 1992 DCC CD.
DR12: 2013 HD Tracks 24.96/192.
DR10: 2001 DVD Audio.
DR9: 2011 SACD, Japan.
DR9: 2014 SHM CD.
DR9: 2006 remaster CD.

You may have the best, most open sounding system/equipment in the world, argue all the positive/negative virtues of LP, PCM or DSDxBillion until the cows come home, but your system can only replay the inherited quality within its source. Feed any system the most compressed version of Hotel Cal., and guess what, you'll hear the most compressed version of Hotel Cal.

Personally, while I understand the importance, I think audiophiles place too much importance attempting to correlate measurements w/equipment choices/formats.

 

[amirm]

Consider distortion that takes the form of a blip or click that is audible and occurs now and again. Something like coil rubbing perhaps. Varying the number of blips per unit of time will enable you to have an arbitrarily low THD value with audible distortion. THD is a reasonable measure for some kinds of distortion but not others.

Right. Another example of that is data dependent distortion which is caused by lossy compression. A transient is much harder to compress than harmonic content. In those extreme cases we do throw out all measurements and rely on bias-controlled listening tests.

The key here then is to analyze the system and see if it is susceptible to such distortions/dynamic behavior. That then determines whether the measurements are appropriate or not.

 

[h.g.]

In those extreme cases we do throw out all measurements and rely on bias-controlled listening tests.

Listening tests are measurement if you ask the listeners what they heard. Measurement is fundamental to science and not optional. If you wish to quantify the audibility of, say, rubbing voice coils you would use an appropriate measure and not an inappropriate one like THD. I haven't looked but one might do something like windowing the clicks in order to separate periods of distortion from periods of no distortion. Perform something like THD on the windowed click and match the audibility of a single click and the audibility of continuous clicks using the click rate. It seems a fairly common requirement and so I would expect there to be a fair amount in the literature if one looks.

 

[amirm]

There is considerable amount of research into audibility of various signals (see the bible from Fastl and Zwicker: http://www.amazon.com/gp/product/3540231595?keywords=fastl zwicker&qid=1458333100&ref_=sr_1_1&sr=8-1). The research though is representative of general population. There is little interest in knowing what the people with the highest acuity can tell.

As to using humans for measurements, that is what we attempt to do in formal listening tests but no matter what, the results are subjective. They are rarely repeatable to the level we do with measurements. This is why even double blind, controlled listening tests are still subjective data, not objective like measurements are.

 

[h.g.]

There is little interest in knowing what the people with the highest acuity can tell.

Would that be golden ears, youngsters or some other group?

As to using humans for measurements, that is what we attempt to do in formal listening tests but no matter what, the results are subjective. They are rarely repeatable to the level we do with measurements. This is why even double blind, controlled listening tests are still subjective data, not objective like measurements are.

When an audiologist is checking someone's ears by playing quiet tones and asking if they can or cannot hear the tone is that subjective and not measurement?

 

[amirm]

Would that be golden ears, youngsters or some other group?

Any and all at the tails of the bell shape. Their population is too small to be an interesting category to research for general psychoacoustics.

When an audiologist is checking someone's ears by playing quiet tones and asking if they can or cannot hear the tone is that subjective and not measurement?

Subjective. Having been tested that way, I can tell you as the levels get borderline, I start to second guess myself in answering if I hear or don't hear the stimulus. So just testing me alone a second time may give different results.

If you could make a true measurement that does not rely on my judgement, then yes, that would be a measurement . Blood pressure or heart rate are such examples.

 

[Ron Party]

Measuring equipment is but one thing, measuring what the equipment plays yet another ....

example: Album, Hotel California, Eagles; song: Hotel California ... (foobar DR ratings)
DR15: 1976 orig LP.
DR15: 1984 orig CD.
DR15: 1987 Japan CD.
DR13: 1992 DCC CD.
DR12: 2013 HD Tracks 24.96/192.
DR10: 2001 DVD Audio.
DR9: 2011 SACD, Japan.
DR9: 2014 SHM CD.
DR9: 2006 remaster CD.

You may have the best, most open sounding system/equipment in the world, argue all the positive/negative virtues of LP, PCM or DSDxBillion until the cows come home, but your system can only replay the inherited quality within its source. Feed any system the most compressed version of Hotel Cal., and guess what, you'll hear the most compressed version of Hotel Cal.

Personally, while I understand the importance, I think audiophiles place too much importance attempting to correlate measurements w/equipment choices/formats.

This is exactly what we (I) need/want: a database for albums not based on what some audiophile claims but based on objective evidence. Great post TBone.

 

[TBone]

This is exactly what we (I) need/want: a database for albums not based on what some audiophile claims but based on objective evidence. Great post TBone.

thx Ron ... I have my own more accurate db, but the following online"db" (below) may be of interest, but be careful ... many of the posted results are either measured wrong or are questionable and the underlying data often incomplete (example: the 1987 japan cd I listed above w/DR15 rating could be inflated b/c many of these early japan4japan cds contained pre-emphasis)

And as a vinyl ripper; the ability to reconcile these values - is of the utmost importance.

http://dr.loudness-war.info/

 

[Mivera]

thx Ron ... I have my own more accurate db, but the following online"db" (below) may be of interest, but be careful ... many of the posted results are either measured wrong or are questionable and the underlying data often incomplete (example: the 1987 japan cd I listed above w/DR15 rating could be inflated b/c many of these early japan4japan cds contained pre-emphasis)

And as a vinyl ripper; the ability to reconcile these values - is of the utmost importance.

http://dr.loudness-war.info/

Nice profile pic Tbone! What a legendary collaboration!

 

[Sal1950]

thx Ron ... I have my own more accurate db, but the following online"db" (below) may be of interest, but be careful ... many of the posted results are either measured wrong or are questionable and the underlying data often incomplete (example: the 1987 japan cd I listed above w/DR15 rating could be inflated b/c many of these early japan4japan cds contained pre-emphasis)

And as a vinyl ripper; the ability to reconcile these values - is of the utmost importance.

http://dr.loudness-war.info/

The online db may have some slightly erroneous numbers, but it's greatest value is in the PR. The db is easy to read and understand for the non-technical viewer and hopefully will have the effect of increasing public awareness and opposition of the current practices.
Personally I run a DR14 on every file that crosses my hands and post the numbers if not already up. I also posted the very disappointing DR7 result of the new Bonnie Raitt HDA release around a number of the audiophile websites in hopes of affecting a change sometime in the future.

 

[TBone]

The online db may have some slightly erroneous numbers, but it's greatest value is in the PR. The db is easy to read and understand for the non-technical viewer and hopefully will have the effect of increasing public awareness and opposition of the current practices.
Personally I run a DR14 on every file that crosses my hands and post the numbers if not already up. I also posted the very disappointing DR7 result of the new Bonnie Raitt HDA release around a number of the audiophile websites in hopes of affecting a change sometime in the future.

Agreed. Sal, not certain what you mean by "run a DR14 on every file..." ... may have missed something in translation, or perhaps I'm having another T Bone Burnett moment ...

 

[Blumlein 88]

Oh the Bonnie Raitt was really a shame.

I think the DR value is raising public awareness quite a bit. Yet, bothersome is the fact it has had absolutely zero effect on how mastering is done. Recent recordings, not re-issues, show very nearly universal lack of dynamic range. People who use mastering services get back slammed to the max results even while asking for "go easy on the compression please". Soon most recordings will be able to be accurately described as having both decibels of dynamic range.

 

[Blumlein 88]

Spectral plots over power (from low to clipping) would be nice to see, as would multi-tone IMD or NPR plots, and impulse and/or step response, using a greater variety of loads. It would be nice to see group delay plots and such as well. It has been a while since I have worked as a tech and taken those sorts of measurements on audio equipment, but it was (and probably is) flat-out scary how some amplifiers react to reactive loads, and even scarier what some amplifiers (and preamps) did to the signal as they clipped. At least we (hopefully) no longer have opamps in there that invert when overdriven, that was fun...

I would have thought better testing of amps with real speaker loads would have been worked out somewhat formally by now. No reason it can't be done, you simply don't see much published about it. Stereophile has a sample loudspeaker load which is helpful. More could be done. I think this is even more useful with so many class D amps around. The output filtering reacts with reactive loudspeaker loads. It is why some speakers with a class D amp sound near world class, and others sound like something must be broken.

 

[Blumlein 88]

There is considerable amount of research into audibility of various signals (see the bible from Fastl and Zwicker: http://www.amazon.com/gp/product/3540231595?keywords=fastl zwicker&qid=1458333100&ref_=sr_1_1&sr=8-1). The research though is representative of general population. There is little interest in knowing what the people with the highest acuity can tell.

As to using humans for measurements, that is what we attempt to do in formal listening tests but no matter what, the results are subjective. They are rarely repeatable to the level we do with measurements. This is why even double blind, controlled listening tests are still subjective data, not objective like measurements are.

The Brian C.J. Moore book is also a good one for such information.