Quote:
Does anyone know of software that is available to evaluate 2 WAV files by first optimally matching the amplitude level and sync and then produce a differential WAV file?

This can be done very easily in Adobe Audition running on a PC. Align the two files at the sample level by having both files open in the multi-track window. Invert the polarity of one file, then you can produce a difference file by playing them back simultaneously.

This is how I produced the difference files that I played back in my seminars at the 2008 RMAF, if anyone here caught them.

John Atkinson
Editor, Stereophile

Have you seen this? I've not tried it myself:
http://www.libinst.com/Audio%20DiffMaker.htm

Quote:
To evaluate the effect of MP3 conversion on an audio file, I would like to find software will produce a WAV file that is the difference between the best match between two sample WAV files.

If File1.WAV is the original, and is converted to File2.MP3 and then reconverted to File3.WAV, the difference between File1.WAV and File3.WAV (= file4.WAV) should reveal the sound lost (or added) in compression, provided that one can match the sync and levels appropriately.

A spectral analysis of the resultant differential file4.WAV could then be used to display the artifacts of compression as a function of time and frequency for various compression parameters. This could be done for other lossy compression algorithms as well.

Appropriate matching of amplitude level and sync is fairly critical to the evaluation and not straightforward. It may be necessary to add a calibration segment to do this effectively.

Does anyone know of software that is available to evaluate 2 WAV files by first optimally matching the amplitude level and sync and then produce a differential WAV file?

File differencing programs are readily available, have long been available for use, and are generally agreed upon by knowlegable people to be fairly useless.

The genrerally-agreed upon software for evaluating MP3 coders is known as a "double blind comparator". The downside - using them requires trained listeners who listen very intentionally and carefully. The two most popular double lind comparison software products are freeware, one called FooBar, and the other called WinABX. Happy Googling!

andy_c tries DiffMaker here, note the comparision of 24/96 to 44.1:
http://www.diyaudio.com/forums/showthread.php?postid=1851714#post1851714

Nice work Andy!

I use Diffmaker all the time!!

Bill Waslo is a genius!(the creator)

Quote:

Quote:
Does anyone know of software that is available to evaluate 2 WAV files by first optimally matching the amplitude level and sync and then produce a differential WAV file?

This can be done very easily in Adobe Audition running on a PC. Align the two files at the sample level by having both files open in the multi-track window. Invert the polarity of one file, then you can produce a difference file by playing them back simultaneously.

This is how I produced the difference files that I played back in my seminars at the 2008 RMAF, if anyone here caught them.

John Atkinson
Editor, Stereophile

the problem is that none of the DAW programs can accurately align to within less than one whole sample. being aligned to within one sample is not accurate enough.

Quote:
I use Diffmaker all the time!!

Bill Waslo is a genius!(the creator)

Yes, I've been using Bill's LAUD measurement system for years. He has also published many excellent articles.

Quote:
andy_c tries DiffMaker here, note the comparision of 24/96 to 44.1:
http://www.diyaudio.com/forums/showthread.php?postid=1851714#post1851714

Nice work Andy!

Thanks Pete - but credit belongs to Bill Waslo for an awesome piece of software.

When I did the 24/96 -> 16/44.1 -> 24/96 sample rate and bit depth conversions, I had to scale the data to prevent clipping. Audio DiffMaker found the needed scale factor for comparison extremely accurately, giving what it calls a "correlated null" in excess of 130 dB.

Quote:
the problem is that none of the DAW programs can accurately align to within less than one whole sample. being aligned to within one sample is not accurate enough.

Okay, I have to ask. Were you the original poster of this thread?

Quote:

Quote:
the problem is that none of the DAW programs can accurately align to within less than one whole sample. being aligned to within one sample is not accurate enough.

Okay, I have to ask. Were you the original poster of this thread?

nope, not me, though that thread on "booby" is classic.

why do you ask?

Quote:
To evaluate the effect of MP3 conversion on an audio file, I would like to find software will produce a WAV file that is the difference between the best match between two sample WAV files.

If File1.WAV is the original, and is converted to File2.MP3 and then reconverted to File3.WAV, the difference between File1.WAV and File3.WAV (= file4.WAV) should reveal the sound lost (or added) in compression, provided that one can match the sync and levels appropriately.

A spectral analysis of the resultant differential file4.WAV could then be used to display the artifacts of compression as a function of time and frequency for various compression parameters. This could be done for other lossy compression algorithms as well.

Whta you want is an align/invert/ mixpaste process, and as others have noted, Audition can do it, as can AudioDiffmaker.
Another possibly useful tool is Synchotron, which is designed to 'synch' two similar files.

However, offering difference files to evaluate the *sound* of an mp3 vs source, is not a valid procedure. It will tell you little or nothing about the audibility of the difference, since the whole point of perceptual encoding is to reduce file size by removing perceptually masked or redundant data -- components that research has shows are typically 'poorly heard' in the context of complex signals like music.

IN other words, you can't necessarily predict the sound of an mp3, from its spectral view.

Quote:
nope, not me, though that thread on "booby" is classic.

why do you ask?

Oh, I just noticed some similarities between you and the OP. He has a studio, uses DiffMaker, and posted about the Kunchur paper. So I thought it might be you. I wasn't sure whether or not you knew about Bill's comments on that paper.

Bill's comments correspond to some of the concerns that have been raised here - namely that sub-sample delays can be realized using Fourier transform techniques. It all works, even though it may seem counterintuitive. This is another way of looking at the "delay resolution" issue - finding the optimum delay, which may not be an integer multiple of the sample period, so the difference between two waveforms can be minimized.

Hey Andy,
I was in lala land... yes, the first poster is me. my mind was elsewhere, I thought you were asking if I was bwaslo(that is the post you linked to)

yes, I am El' Dante.

anything with Johnny Cash as an avatar..probably me.

I've forwarded the thread on to Dr. Kunchur, hopefully he will address Mr. Waslo's concerns. I respect them both a great deal.

Quote:

Quote:
nope, not me, though that thread on "booby" is classic.

why do you ask?

Oh, I just noticed some similarities between you and the OP. He has a studio, uses DiffMaker, and posted about the Kunchur paper. So I thought it might be you. I wasn't sure whether or not you knew about Bill's comments on that paper.

Bill's comments correspond to some of the concerns that have been raised here - namely that sub-sample delays can be realized using Fourier transform techniques. It all works, even though it may seem counterintuitive. This is another way of looking at the "delay resolution" issue - finding the optimum delay, which may not be an integer multiple of the sample period, so the difference between two waveforms can be minimized.