How does one down-sample a music track for ABX tests?

Exactly what the title says. I'm trying to see if I can detect the difference between 24/88 and 16/44. I know there are technical differences, but I want to see if I can hear them.

I don't want this to be an HD vs CD thread, or FLAC vs MP3 as I know those exist already.

I purchased Seal's "Best of 1991-2004" album from HDTracks.com and I have Seal's first album accurately ripped from a CD via EAC.

When I compared his "Crazy" track from both albums with Foobar's ABX component (hiding my results), I got this log file output. To me the difference between the two was incredibly obvious and I actually preferred the sound from the B sample as it sounded airier and less aggressive (apparently it was the one ripped from the CD). Perhaps my test results were because of the difference in albums in itself (maybe the "Best of 1991-2004" album had re-mastered tracks), so I was wondering if there is a way to down-sample a music track so I can get a more accurate test result.

Test was done with a JDS Labs Objective DAC, Objective 2, and a 2012 AKG K 701.

Quote:

Originally Posted by PleasantSounds 

It's quite common that the HD versions of recordings undergo a different mastering process - thus the differences are quite obvious.

 

If you are using Foobar, then the easiest way to downsample is to right-click on the track and pick 'Convert'. In the Convert Setup menu pick 'Processing' and select Resampler. If you haven't done that yet, I recommend to get the SoX resampler. Once you added it to the Active DSPs click on 'Configure selected' and choose your desired sampling frequency. 

Ah, that sounds like a really great Foobar add-on. I'll be sure to try it out when I get the chance. Thank you so much for the recommendation! :)

OK, that's what I thought....I really can't tell the difference between 24/88 and 16/44....

To me it seems that science has surpassed my ability to detect it. On a technical note, the 24/88 file should have been superior, when in fact I can't tell the difference between it and the "inferior" CD-quality file. While I could hear a very, very, very slight difference between the two while listening to the A and B samples next to each other (the A file sounded a bit more dynamic, and I happened to be right when it was in fact the HD track), I could not distinguish the difference between them and the X and Y files, oddly enough. I might do a second trial to confirm my results above.

I did a second run and although the results are better, I still failed:

Again, this was tested using a JDS Labs Objective DAC, Objective 2, and a 2012 AKG K 701.

Quote:

Originally Posted by skamp 

88.2 kHz is not a supported sampling rate with the ODAC. You'd need to resample it to 96 kHz. As for the lower rate file, you should resample the 88.2 kHz file to 44.1 kHz, and then resample that to 96 kHz, and ABX that.

Congrats for trying to run a proper ABX test, though!

Dooh, you've got an excellent point there. Thanks for pointing that out!

If you re-sample a track multiple times (to 44.1 then back up to 96), does that degrade the quality of the file? The creator of the ODAC said 24/88.2 is nearly indistinguishable from 24/44.1.

Quote:

Originally Posted by skamp 

Quote:

Originally Posted by stv014 

Also, if you do not use floating point samples for the intermediate files, some clipping may occur, especially with "loud" tracks.

That's why SoX comes with the -G / --guard parameter, which will automatically reduce volume, should clipping occur.

Quote:

Originally Posted by stv014 

Quote:

Originally Posted by miceblue 

If you re-sample a track multiple times (to 44.1 then back up to 96), does that degrade the quality of the file?

 

That depends on the quality of the resampler. If it is really good, and used with the right settings, then converting the file a few times is not a problem. Obviously, once you convert to a low sample rate, the high frequency information is lost irreversibly. Also, if you do not use floating point samples for the intermediate files, some clipping may occur, especially with "loud" tracks.

Here is a quick experiment I tried: using my resampler utility, I converted a 3 minute test track from 44100 Hz to 96000 Hz, then from 96000 Hz to 67139 Hz, and finally from 67139 Hz back to 44100 Hz. All files were in 32-bit float format. I subtracted the converted file from the original, and for most of the file the peak amplitude of the difference was not higher than -135 dBFS; only at the beginning and end was there some short artifact because of truncating the pre- and post-ringing of the lowpass filter in the intermediate files.

Ah, that's good to know. I'll be sure to try it out when I get the time. Thanks again for the recommendations!