[limpidglitch]

  Surely 30% in the ABX shows that he preferred 44.1 to 192? In other words 70% of the time he chose CD quality.

 
Changing the criterion from one-tailed to two-tailed post hoc is not a done thing. Assuming he even had set a criterion at all.
We go on about recommending people willy-nilly to perform these ABx tests, but setting up a protocol and interpreting the results is anything but a triviality.

 

[RRod]

   
Changing the criterion from one-tailed to two-tailed post hoc is not a done thing. Assuming he even had set a criterion at all.
We go on about recommending people willy-nilly to perform these ABx tests, but setting up a protocol and interpreting the results is anything but a triviality.

 
Hence why my new mantra for ABX is "do it for yourself and let the results tell you what they may, but don't expect me to believe them for a second." If people want help on proper setup and analysis, that's good, but I'm done with 8/10 "see, hi-res matters" folks.

 

[kraken2109]

 
I wouldn't make that claim - he was correct in spotting the proper format 30% of the time, so less than he should have if he just randomly guessed. However since he did only one set this sample is a bit weak.

I know it isn't statistically significant, I just found it quite funny.

 

[Koolpep]

I know it isn't statistically significant, I just found it quite funny.

Oh yes it certainly has some nice rig to it. that it seems likely that he preferred the lesser format objectively and still didn't learn the lesson.

 

[jarrett]

What's the latest on the thread topic? 

 

[bigshot]

Oh! Happy to help with an answer to that... People can't hear a difference between lossy and lossless above the point of audible transparency and that depends on the bitrate and codec used.

 

[gevorg]

FYI: Foobar's ABX tool converts the A and B tracks to WAV first and stores it as temp files in your OS, so you can't use foo_abx to compare lossless compression formats. Still great for sample rate, bit rate, lossy vs lossless, mastering quality comparisons.

 

[RRod]

Why can't it compare lossless formats? The WAV is what gets sent to the DAC anyway, be it from FLAC, ALAC, whatever. In the case of lossless formats, all the test would show is that your program decodes the lossless format correctly.

 

[gevorg]

[CONTENTEMBED=/t/738552/testing-the-claim-i-can-hear-differences-between-lossless-formats/690#post_11370834 layout=inline]Why can't it compare lossless formats? The WAV is what gets sent to the DAC anyway, be it from FLAC, ALAC, whatever. In the case of lossless formats, all the test would show is that your program decodes the lossless format correctly.[/CONTENTEMBED]

Its not exactly the same since you're hearing WAV files being played, not the original FLAC or APE being decompressed/played on the fly. The original compressed files do not even get any involvement in ABX, the tool is using them as a source to create WAV files for A and B. Note that I don't claim that there is a difference between lossless compression formats, its just foo_abx is not a proper test if you happen to like this kind of torture.

 

[RRod]

Its not exactly the same since you're hearing WAV files being played, not the original FLAC or APE being decompressed/played on the fly. The original compressed files do not even get any involvement in ABX, the tool is using them as a source to create WAV files for A and B. Note that I don't claim that there is a difference between lossless compression formats, its just foo_abx is not a proper test if you happen to like this kind of torture.

 
The DAC chip in your soundcard / DAC unit doesn't understand FLAC; it only speaks PCM or maybe DSD. On-the-fly decompression should simply be restoring the original wav/pcm samples as it goes along, and so an ABX of the WAVs made from the two different lossless codecs should be equivalent to what you'd get on-the-fly. They in fact should be giving the same exact samples at the same exact time; if they aren't, something is quite wrong.

 

[gevorg]

Its not exactly the same since you're hearing WAV files being played, not the original FLAC or APE being decompressed/played on the fly. The original compressed files do not even get any involvement in ABX, the tool is using them as a source to create WAV files for A and B. Note that I don't claim that there is a difference between lossless compression formats, its just foo_abx is not a proper test if you happen to like this kind of torture.

The DAC chip in your soundcard / DAC unit doesn't understand FLAC; it only speaks PCM or maybe DSD. On-the-fly decompression should simply be restoring the original wav/pcm samples as it goes along, and so an ABX of the WAVs made from the two different lossless codecs should be equivalent to what you'd get on-the-fly. They in fact should be giving the same exact samples at the same exact time; if they aren't, something is quite wrong.

Of course it doesn't, the same goes for 64kbps MP3, which foo_abx will convert to WAV too. Are you going to say that playing low bitrate mp3 is the same as lossless? The premise that lossless formats sound different lies in the possible difference of decompression during playback (I can't think of anything else that theoretically can make a difference here, and whatever the cause, its probably a bug/unoptimized code in playback software, or just a really old/slow computer). Therefore, in order to do a proper ABX to test this premise, you have to compare playback directly from lossless format when on-the-fly lossless-to-PCM conversion occurs, rather than uncompressed WAV that foo_abx prepares beforehand.

 

[OddE]

Of course it doesn't, the same goes for 64kbps MP3, which foo_abx will convert to WAV too. Are you going to say that playing low bitrate mp3 is the same as lossless? The premise that lossless formats sound different lies in the possible difference of decompression during playback (I can't think of anything else that theoretically can make a difference here, and whatever the cause, its probably a bug/unoptimized code in playback software, or just a really old/slow computer). Therefore, in order to do a proper ABX to test this premise, you have to compare playback directly from lossless format when on-the-fly lossless-to-PCM conversion occurs, rather than uncompressed WAV that foo_abx prepares beforehand.

 
-To a large extent, I agree with you. However (This is the Internet, after all!) I'll engage in some nit-picking: If the decompression during playback introduces audible differences, it is a poor implementation, and what you are then testing is the implementation, rather than the format itself. (Now, I agree that to the end user, the difference doesn't matter one iota!)
 
Any competently designed decompression algorithm would read ahead, anyway - it will not try to decode, say, FLAC on the fly with no buffer to fall back on should the computer decide to do some other task hogging loads of CPU cycles during playback; it would decompress a significant chunk of the file being played back before playback even starts, then keep reading ahead as the track is played.
 
(Heck, 44.1/16 is pretty close to 10MB/min - with today's computers, it would make sense to decompress the entire track before starting playback, storing the PCM in RAM. (Or on disk, for that matter - where the OS will do the same thing; read ahead, load the data into memory in chunks for later processing.

 

[gevorg]

^^ Exactly, the ABX here is for implementation of the playback software/hardware. To check the lossless format itself, no ABX is necessary, just uncompress it, remove headers/metadata and hash the audio samples. Pure math.

 

[RRod]

^^ Exactly, the ABX here is for implementation of the playback software/hardware. To check the lossless format itself, no ABX is necessary, just uncompress it, remove headers/metadata and hash the audio samples. Pure math.

 
You can do the same sample comparison by just piping the output of the software to a file and then doing the diff after aligning the two dumps. Bottom line is that end-user audiophiles shouldn't be the ones testing lossless formats, I guess ^_^

 

[Head Injury]

For as ABX test of lossless decoding, you can always loop the output of both formats back into the line in, then record and ABX the files you get. Any changes to the output that decoding makes should be captured in the resulting files. Not that there should be any.