[miceblue]

Hi all,

People around the "hi-fi" realm often state that lossy formats converted to other lossy formats really degrades the sound quality. I think it is common practice these days to covert lossless files to lossy, but do your really lose much from a lossy-lossy conversion?



I am interested to see if anyone can hear the quality loss from such a conversion.

https://dl.dropboxusercontent.com/u/2816447/ABX.zip



This isn't too official and it's pretty basic. Inside the zip folder are 3 files (plus there might be some hidden indexing files that Mac OS X might create): 1 file is the original 24/96 FLAC file, 1 is a 320 kbps VBR AAC file, and 1 is a V0 LAME MP3 file.

One of the two lossy files was converted from the other, can you hear which one is which?

If you're interested in participating, please put your answers in a spoiler tag.



Happy listening!

Spoiler: Note to self

[contentembed]The MP3 file was created from the FLAC file, and the AAC file was created from the MP3 file. After numerous ABX tests, I cannot hear the difference between the two lossy files.[/contentembed]

 

[ab initio]

Thanks for posting the tracks for testing!
 

Spoiler: Warning: Spoiler!

After a while with foobar's ABX, I cannot with any statistical significance distinguish between flac and either of the encoded files. Therefore, no, I cannot determine which file has been encoded twice by listening.

 
 
Cheers

 

[allets]

Usually these kind of tests are best achieved by dropping all files back to a true lossless format like .wav / .aiff (flac to wav, aac mp3 to wav etc. ) I don't think you'll get accurate results with your method... especially since placebo exists 

 

 

[miceblue]

Usually these kind of tests are best achieved by dropping all files back to a true lossless format like .wav / .aiff (flac to wav, aac mp3 to wav etc. ) I don't think you'll get accurate results with your method... especially since placebo exists   

While that's usually true, I don't think it matters for this particular test. You have no idea which lossy file is which and knowing the format doesn't really help.

But while we're on that topic, how does one add "fluff data" to a file to make all 3 files the same size? At least I think that's what happens. I think I've seen tests where all 3 files are WAV files and they're all the same size despite being encoded differently.

 

[kn19h7]

While that's usually true, I don't think it matters for this particular test. You have no idea which lossy file is which and knowing the format doesn't really help.

But while we're on that topic, how does one add "fluff data" to a file to make all 3 files the same size? At least I think that's what happens. I think I've seen tests where all 3 files are WAV files and they're all the same size despite being encoded differently.

WAV is uncompressed, which means:
length (in seconds) x sampling rate (in hertz) x bit depth (in bits) = filesize (in bits)
 
So if all the three factors are same, the filesizes will stay constant.

 

[Dark_wizzie]

Or maybe, have something that is lossy from lossy from lossy, for us hearing challenged folks to try to abx. 

 

[bigshot]

I took a track and bounced it from CD to AAC256 to AIFF to AAC256 over and over again ten times and it still sounded pretty doggone good.
 
http://vintageip.com/xfers/aac10gens.m4a
 
Someone once mentioned to me that you lose more going back and forth from AAC to MP3 because the codecs affect different things. That might be the case, but I still think the loss would be pretty small.

 

[francopro]

Hi all,

People around the "hi-fi" realm often state that lossy formats converted to other lossy formats really degrades the sound quality. I think it is common practice these days to covert lossless files to lossy, but do your really lose much from a lossy-lossy conversion?



I am interested to see if anyone can hear the quality loss from such a conversion.

https://dl.dropboxusercontent.com/u/2816447/ABX.zip



This isn't too official and it's pretty basic. Inside the zip folder are 3 files (plus there might be some hidden indexing files that Mac OS X might create): 1 file is the original 24/96 FLAC file, 1 is a 320 kbps VBR AAC file, and 1 is a V0 LAME MP3 file.

One of the two lossy files was converted from the other, can you hear which one is which?

If you're interested in participating, please put your answers in a spoiler tag.



Happy listening!

Spoiler: Note to self

[contentembed]The MP3 file was created from the FLAC file, and the AAC file was created from the MP3 file. After numerous ABX tests, I cannot hear the difference between the two lossy files.[/contentembed]

I always found that reconverted files sound worst than original ripped files even with high end hardware and softwares, if you want another format it's always better to rip the original source again (cd,dvd-a,vynil etc) into desired format

 

[brhfl]

Just for fun, I have my home computer currently encoding a file to 320CBR MP3, then 320CBR M4A, then starting over, on a loop. I started this before I went to sleep last night, but was able to listen to the 100th iteration (200 conversions total) before actually conking out. It was noticeably degraded, bit of an old cell phone sound to it, but it sounded startlingly good for having gone through lossy conversion 200 times. Unfortunately, I 'upgraded' to Mavericks last night, and I guess it messed with my power settings - computer was asleep when I woke up this morning, at 450 iterations (900 conversions). I changed the sleep time, hopefully it stuck. Will post an update when I get home, around 8 ET. Sample was the first ~2.5 minutes of the first movement of Orff's Catulli Carmina ('Eis Aiona!'), Eugen Jochum's 1970 stereo recording.

 

[castleofargh]

I tried the same concept for .jpg files after wondering how bad it would really be to slightly post process something already saved to jpg. in max quality jpg(but not the jpg2000 lossless stuff) it took me almost 10 opening and re-saving, to start getting some real noticeable degradations. and it would still have been fine enough for consumer use.
I guess it might be about the same thing for re-saving the same compression codec, we're being told not to do it for technical reasons and we kind of exaggerate the real impact in our mind.
now going from one lossy to another is probably more of a random guess about the differences in algorithms.
 

Spoiler: Warning: Spoiler! my results will blow your mind!!!

or not ^_^.
 
only tried the 2 lossy files. I got wrong the 6first attempts making me almost confident that I could actually tell them apart(being wrong is kind of my strong attribute, kind of how I do things) then I got right the next 4.. booohooohoooo. 20 tries led me to a good old 11/9.
I didn't put my life on the line or spent too much time on it, but I did the first 10 switching in the entire file, and le next 10 with just a short loop I thought was meaningfull. so that's enough for me not to care in real life listening.
but I can talk about all the differences I'm sure I heard in great details ^_^ . the wonders of imagination.

 

[headdict]

  I tried the same concept for .jpg files after wondering how bad it would really be to slightly post process something already saved to jpg. in max quality jpg(but not the jpg2000 lossless stuff) it took me almost 10 opening and re-saving, to start getting some real noticeable degradations. and it would still have been fine enough for consumer use.

I think the fun starts when you want to edit the compressed file, e.g. change contrast on a jpg or enhance dynamics on an mp3. This is not what lossy compression is meant for. Consumers should be fine with moderate compression, even if it's applied repeatedly.

 

[bigshot]

JPGs and MP3s have a lot more latitude than you seem to think. You can apply all sorts of filtration to them with no problems. 

 

[brhfl]

  JPGs and MP3s have a lot more latitude than you seem to think. You can apply all sorts of filtration to them with no problems. 

Except the people who tend to care about JPEG vs. lossless are photographers, which would be akin to using MP3 in the studio. For the end product… no big deal… but during production stages, that latitude can be crucial. Not that lossy formats can't be designed around those constraints, but these are really end product formats. (Counterpoint: we did use ATRAC all the time for audio recording in film school… ah well…)
 
I'm going to see if my experiment makes it up to 5,000 before my room becomes unbearably hot (I like round numbers, ok?). 4,000 sounds… quite bad, but still better than I'm sure a lot of us would expect.
 
Just for reference, I'm doing all the conversion in XLD, maxed out on quality, with afconvert handling an M4A-ALAC conversion in the middle (XLD doesn't take M4A as input).

 

[bigshot]

It depends on your style of shooting. If you are a "grab and go- fix it later" kind of photographer, you need a lot of latitude. But if you are the type of photographer that tweaks every setting to get proper exposures in camera, jpeg is fine. There are professional photographers who shoot jpeg, and there are audiophiles who listen to lossy. Usually, it's the duffers who worry endlessly about file sizes.
 
When I did my test, I figured at absolute worst, I might be forced to do four or five transcodes in my lifetime. So I doubled that and did 10 and it sounded good. I'm not worrying about it. Film and analogue audio were MUCH worse about generation loss than lossy image and audio formats with a decent size.

 

[castleofargh]

 obviously the lossy files were made with the purpose of keeping what's heard and discard what won't be heard to save space. if the purpose of post processing(mastering) is to bring back something that isn't heard enough, to the front of the scene. then there is gonna be a problem with lossy format ^_^.
 
anyway that's a false problem, it doesn't make much sense to use mp3 in a studio for plenty of reasons. the question here from the OP was to determine the loss of going from lossy to lossy.