[Crazy*Carl]

No.

 

[stv014]

Quote:

For a meaningful ABX comparison, volume should be adjusted in hardware only, not software, and the resulting volume difference between the samples needs to be less than ~0.25 dB (perhaps even less).

Why not software ? If the adjustment is done using 24 bit output format (which is possible with anything other than the lowest end DACs and onboard audio), it has insignificant effect on the audio quality, and it allows for very accurate matching easily.
 

 

[stv014]

Quote:

Originally Posted by kmj2587 /img/forum/go_quote.gif
 
In practical terms for me and the kind of music I tend to listen to, this translates to my classical library being 90% flac while the rest is about an even mix of flac and mp3.

Classical is not necessarily the hardest to encode to mp3. Although it does encode better (more space saved) to FLAC, than music with compressed dynamic range and a lot of noise/high frequency content (cymbals etc.).
 

 

[Jaywalk3r]

Why not software ? If the adjustment is done using 24 bit output format (which is possible with anything other than the lowest end DACs and onboard audio), it has insignificant effect on the audio quality, and it allows for very accurate matching easily.

If software is used, the output is no longer bit perfect relative to the file being played. Further, since volume need only be adjusted for one file of each comparison, only one of outputs is being degraded. Suddenly the comparison made is not the comparison desired.

 

[stv014]

Quote:

If software is used, the output is no longer bit perfect relative to the file being played.

Neither is it with hardware, regardless of whether the adjustment is digital (in which case it is the same as doing it with software), or analog (less accurate, adds noise etc.).
 
Quote:

Further, since volume need only be adjusted for one file of each comparison, only one of outputs is being degraded.

 
I doubt the quantization noise of 24-bit audio is anywhere near audible.
 

 

[Jaywalk3r]

I doubt the quantization noise of 24-bit audio is anywhere near audible.

Additional ABX test would be required to test that hypothesis before it could be validly assumed for the comparisons being discussed.

Additionally, most codec comparisons involve 16-bit files, not 24-bit. The more changes made to the files being compared, the less general and meaningful the results are.

 

[kn19h7]

Quote:

If software is used, the output is no longer bit perfect relative to the file being played. Further, since volume need only be adjusted for one file of each comparison, only one of outputs is being degraded. Suddenly the comparison made is not the comparison desired.

....isn't replaygain applied on both tracks?
 


Quote:

Additional ABX test would be required to test that hypothesis before it could be validly assumed for the comparisons being discussed.
Additionally, most codec comparisons involve 16-bit files, not 24-bit. The more changes made to the files being compared, the less general and meaningful the results are.

...I thought most software player use 32bit float internal format?
 
 
Just FYI, I did try it myself without replaygain and dsp...

 

[Jaywalk3r]

....isn't replaygain applied on both tracks?

Yes, but ReplayGain adjusts volume in 1.5 dB increments. That is too coarse an adjustment for ABX comparisons. I've read where 0.3 dB have been identified with ABX comparisons. In order for an ABX comparison for lossy codec quality to be meaningful, the volume of each file being played has to be close enough to the same that the files cannot be identified by volume. ReplayGain cannot do that, though, to be fair, that isn't what it was designed to do.

Just FYI, I did try it myself without replaygain and dsp...

"I replaygained both tracks in foobar, the difference was half a decibel or so."​

Half a decibel or so volume difference is too much to compare sound quality, because the files should be identifiable by volume alone.

 

[kn19h7]

Quote:

"I replaygained both tracks in foobar, the difference was half a decibel or so."
Half a decibel or so volume difference is too much to compare sound quality, because the files should be identifiable by volume alone.

That's not me == ...my post at #78
 

 

[Jaywalk3r]

"I replaygained both tracks in foobar, the difference was half a decibel or so."


Half a decibel or so volume difference is too much to compare sound quality, because the files should be identifiable by volume alone.

That's not me == ...my post at #78

I see. Sorry, the link didn't go to the post, just the page.

However, the same problem exists if ReplayGain (or similar software) was used for volume adjustment between tracks.

In order for an ABX comparison of sound quality between lossy and lossless codecs to offer meaningful information, all other means of differentiation must be eliminated.

 

[kn19h7]

...I've said I didn't use any replaygain or dsp, the process was just encode and compare

 

[Jaywalk3r]

...I've said I didn't use any replaygain or dsp, the process was just encode and compare

What was the volume difference? Did you measure it?

 

[kn19h7]

0.00 or parhaps a very insignificant value.. replaygain scan on both tracks return same values

 

[Head Injury]

Quote:

 

Yes, but ReplayGain adjusts volume in 1.5 dB increments.

Could you clarify this? My ReplayGain in foobar2000 adjusts to 1/100th of a decibel, though I suppose it would be hard for me to verify the accuracy.

 

[Jaywalk3r]

0.00 or parhaps a very insignificant value.. replaygain scan on both tracks return same values

Interesting. I don't know enough about ReplayGain's measurements to trust it. With what precision does it return values? (More importantly, but much more difficult to find, with what precision is it capable of measuring?) I'd be much more inclined to manually measure the volume levels if I wanted to do a serious comparison.

I don't mean to be overly critical. It is the nature of scientific experimentation to make every effort to ensure that only the thing being tested can influence the results.