What objectivists and subjectivists can learn from each other

[rbbert]

...The logic flaw IMO is using ears to judge component accuracy given there is no ear calibration or consistancy....or, in other words, how many components have audiophiles gone through, through the years, and still can't get a "live" performance in their room or a good enough illusion to keep what they got...

Tom

Is this failure unique to audiophiles? Do you have an audio system which produces music indistinguishable from live??

 

[rbbert]

Music is emotions, and you simply cannot measure it.

But "sound" is measurable, and it is my contention that those measurements are both incomplete and imperfectly understood, and IMO the two "objectivists" here are failing miserably at their attempts to demonstrate the opposite, and that despite my desire for them to be at least mostly right.

 

[NorthStar]

First, what are we measuring exactly?

And how does it truly affect (effect) the way we feel?

 

[fas42]

The logic flaw IMO is using ears to judge component accuracy given there is no ear calibration or consistancy....or, in other words, how many components have audiophiles gone through, through the years, and still can't get a "live" oerformance in their room or a good enough illusion to keep what they got...

Tom

Tom, the most important measurement tool IS the ears, because that is what has to be "fooled". Tremendous precision in some area may be completeless useless for our hearing mechanism to "get" it, so relying on specifications to find the right qualities is a dead end. At least at the moment ...

A revolving door of components is another version of "audio hell"; what only a very few realise is that it is not the components that are the answer, but how well they are tuned. Of course, in other areas of aiming for high performance, the notion that you can just buy a winning combination would be laughed out of the room -- just consider motor sports, in its myriad forms ...

Frank

 

[fas42]

Music is emotions, and you simply cannot measure it.

Disagree, Bob. You can measure it, or at least I can: the speakers, the components producing the music completely disappear; there is nothing in what you hear that says, "This is not quite right ...". It IS an achievable goal, just something, unfortunately, not available by simply handing money across a counter ...

Frank

 

[jkeny]

I was talking about a total null, to zero volts. That's what I'd expect when nulling a Wave file against a lossless copy of itself, which was the context:

Please show such a test & the results - I presume you have done many to make these claims?

http://www.whatsbestforum.com/showt...rom-each-other&p=101452&viewfull=1#post101452

Not really. Music is nothing more than a bunch of sine waves of varying amplitude, frequency, duration, and phase.

Ah the old canard "music is just a series of sine waves" so we can just use a sine wave in our tests/measurements - it's the same as music. Wrong! You really should stop trotting out worm out mantras as evidence. If you don't know how testing a music signal might differ from testing a sine wave then you are really so far away from what real music reproduction is about that you couldn't possibly be serious!

The residual from a nulling distortion analyzer proves that there's nothing more to audio fidelity than is already known. But yes, modern nulling of Wave files lets us prove that two files are the same, even when people are convinced they sound different. If you believe otherwise, please explain how that's possible.

--Ethan

So let me get this straight - you are saying that the nulling test from 50+ years ago is as good as it gets?

 

[NorthStar]

Disagree, Bob. You can measure it (emotions), or at least I can: the speakers, the components producing the music completely disappear; there is nothing in what you hear that says, "This is not quite right ...". It IS an achievable goal, just something, unfortunately, not available by simply handing money across a counter ...

Frank

Frank, I'm real glad that you can! :b

 

[rbbert]

I know you are kidding.

Tom

Why would I be kidding? I'm sorry I neglected to include you with Ethan and Tom in my earlier post's references, because like them you seem to be implying that if only microphones were better, or speakers were better, or ( "2 channel/2 speaker audio can not replicate live, only attempt an illusion. Any audio signal/air movement is fully measurable and nobody's ears are as high resolution as test gear. What happens at the ear/brain interface is a preference, an opinion. Imaging occurs at the ear/brain interface. Only unsighted listening tests remove expectation bias and reveal your true preferences for the illusion.") perhaps if we had enough speakers, the reproduction (or illusion) would be "perfect". This last despite the fact that it is rarely if ever spatial cues which most immediately distinguish live music from reproduced music.

BTW, how does one measure a microphone, or a speaker? Does anyone else see a tautology here?

 

[fas42]

Krank, I'm real glad that you can! :b

Out of interest, Bob, how close have you ever got to the sort of the thing I talk about; either with your own system or listening to others?

Frank

 

[andy_c]

If anybody is interested in trying out free audio differencing/nulling software, there is the freeware Audio DiffMaker by Bill Waslo. It takes two nominally identical WAV files, determines the scale factor and time shift of the "test" file relative to the "reference" in order to get the deepest null in the difference. It then forms a third file consisting of said difference, which you can listen to. I've used it in the following way.

1) Using the freeware DVD-A Explorer, extract a WAV file from a DVD-A with 24/96 resolution. Keep it as the reference and call it "file A".
2) Downsample a copy of "file A" to 16/44.1 using triangular PDF dither with the freeware SoX program.
3) Upsample that file again to 24/96, again using SoX. Call it "file B".
4) Using Audio DiffMaker, take the difference between "file A" and "file B". Call it "file C".
5) Listen to "file C", the difference between the original 24/96 file and the one that's gone through "round trip" resampling and resolution reduction.

There's a link to Bill's AES Convention paper about Audio DiffMaker at that link as well.

 

[mep]

If anybody is interested in trying out free audio differencing/nulling software, there is the freeware Audio DiffMaker by Bill Waslo. It takes two nominally identical WAV files, determines the scale factor and time shift of the "test" file relative to the "reference" in order to get the deepest null in the difference. It then forms a third file consisting of said difference, which you can listen to. I've used it in the following way.

1) Using the freeware DVD-A Explorer, extract a WAV file from a DVD-A with 24/96 resolution. Keep it as the reference and call it "file A".
2) Downsample a copy of "file A" to 16/44.1 using triangular PDF dither with the freeware SoX program.
3) Upsample that file again to 24/96, again using SoX. Call it "file B".


4) Using Audio DiffMaker, take the difference between "file A" and "file B". Call it "file C".
5) Listen to "file C", the difference between the original 24/96 file and the one that's gone through "round trip" resampling and resolution reduction.

There's a link to Bill's AES Convention paper about Audio DiffMaker at that link as well.

So what happens?

 

[NorthStar]

Out of interest, Bob, how close have you ever got to the sort of the thing I talk about; either with your own system or listening to others?

Frank

Impossible to answer that question Frank, and you should know that.

Everybody's different (mentally and emotionally); plus everyone's system setups, gear, loudspeakers, rooms, houses, cars, boats, watches, kitchens, wifes, children, etc., are also different.

P.S. Sorry for the typo in your name Frank, I fixed it.
I realized soon I saw your quoting me just above.

 

[JackD201]

Ethan,

For the record, please define FIDELITY as clearly as you can in your own words.

Thank you,

Jack

 

[andy_c]

So what happens?

In my case, with the particular song I chose (Uncle John's Band, from Workingman's Dead), I had to turn the volume all the way up and put my ear right next to the speaker to hear anything at all. What I heard was a slight increase in hiss, nothing that sounded like music or distortion at all.

Of course, this was a "sample of one", but I did choose the song by doing spectrum analysis of a number of songs from several DVD-A discs I had ripped, and picking the one with the most high-frequency content.

There is a bug in DiffMaker that causes an audible "blip" at the beginning of the difference file, lasting a fraction of a second. This occurs even when subtracting two identical copies of the same file. So it's necessary to keep the volume down until the "blip" has passed before turning it up to listen to the difference.

I originally tried the freeware r8brain resampler, and that one caused audible clicks in the difference file. When I switched from it to SoX, on my first downsampling attempt, SoX gave me a warning about clipped samples in the downsampled file. But SoX allows scaling in the conversion, so I experimented a bit until I arrived at a scale factor just a wee bit less than 1, just enough to avoid the clipped samples warning. When computing the difference, DiffMaker correctly figured out this scale factor, and there were no clicks in the difference file. I'm guessing the clicks in the difference file with r8brain were due to clipped samples, but I did not try to prove that.

 

[mep]

I guess I don't *get* the point to it all. Unless something goes wrong during the rip process, I think there is exhaustive proof that a copy of a digital file should and can be bit-for-bit perfect. Big deal. Next.

 

[andy_c]

Well yeah, but this had been resampled from 24/96 to 16/44.1, then back up again. So the two files were not bit-identical.

Edit: Once you do the resampling along with conversion from 24 to 16 bits, that resolution is lost forever.

 

[mep]

I guess the good news is that we don't listen to null tests, we listen to music. I feel comfortable that when I rip a CD to my hard drive that I'm getting a copy that is identical to the CD.

 

[andy_c]

What I was testing for was not the bit-perfectness of a rip from CD or DVD-A, but the potential audible effects of conversion from 24/96 to 16/44.1.

 

[mep]

And my next question would be why would you want to downsample if you didn't have to?

 

[andy_c]

Good question! My sound card at the time (E-MU 0404 USB, with S/PDIF output into an external DAC) required going into this total PITA sound card setup program to change resolution and sample rate. For some reason, it was incapable of doing so automatically. So if I did a playlist having a mixture of 16/44.1 and 24/96 files, my player program would bomb when the resolution or bit rate would change from one file to the next.

I eventually got a sound card that was capable of switching sample rates and bit depths automatically on the fly, which solved the problem.

Also, that Grateful Dead DVD-A is a complete remix (not just a remaster) done by the band's drummer Mickey Hart, so it sounds a lot different from the original CD. I wanted to be able to convert it to a form allowing it to be burned to CD.