Article explaining why wav and flac sound different

Could it be that  the dbpoweramp flac encoder and the J River decoder are not fully compatible? I wonder why this was not eliminated by just using J River as both the encoder/decoder? I have played with J River wav/flac encoding and found no SQ deterioration worth writing about.  I have downloaded both wav & flac files (Qobuz allows this) and found no significant SQ difference again using  J River for playback.

However I have noted an audible loss in fidelity converting from Monkey Audio (APE) files to FLAC using  J River (rev 21).

Eloise posted:

I turned off at...

Part three reported that a FLAC file sounded inferior to the WAV file from which it was made, and we found to our surprise that when these FLAC files were reconverted, the resulting WAV file did not recover the full sound quality of the original. We repeated these conversion steps five times and observed a hyperbolic decline in WAV sound quality, the greatest loss occurring in the first two or three conversions.

Sorry but (excepting the possibility of corruption and possibly fragmentation) it is fundermentally impossible for a computer to "know" if a file has been converted into a flac file and back again.  

Ha ha ha ha ha ha ha ha ha ha ha ha ha ha! No really, they're serious? Actually serious? They believe the sound quality deteriorates through conversion from WAV to FLAC and back again? Oh good grief.

If so . . . if they really (breaks off to collapse into giggles and slap the chair arm a few times) believe that the computer 'knows' that this has occurred, even though the files are identical, then they should come round to mine because I've got a really nice bridge in London to sell them. Hardly used or anything.

One problem, I believe, with audiophilia is that many of its tribe were brought up in an analogue era where the flow of electricity from a mains switch through the whole system to the speakers was fragile and easily disturbed. Subjectivity was inevitable. Now, in the digital era, there is a fundamental break in that flow: the DAC. Everything upstream of the DAC is in the digital domain, and you can copy it on to a 1990s floppy disk, post it to yourself, zip it, unzip it, copy it back on to your XBox and then stick it on the NAS, as long as the file integrity is maintained, it makes no difference whatsoever. None, nada, zero. Only downstream of the DAC, in the analogue domain, do those old rules apply. 

Since there is a mathematically defined lossless compression / decompression between FLAC and LPCM, and WAVE files contain LPCM data, then, if there is any degradation in the file data, that shows error in the FLAC <> WAVE encode or decode processes.

There is some potential for the real time stripping of metadata to cause timing shifts (i.e. increase jitter) if the files are being interpreted directly (e.g. by a computer sending the data to a DAC, particularly if it has limited buffering capacity).  This would indicate that the file processing system (e.g. the computer) isn't adequately specified for real time processing (N.B. neither PCs nor MACs are any where near optimal for real time processing, but this is primarily due to their OS designs).

Using UPnP this shouldn't be a problem as the metadata have been stripped out before the stream packets are put on the network,

Solid Air posted:

Eloise posted:

Sorry but (excepting the possibility of corruption and possibly fragmentation) it is fundermentally impossible for a computer to "know" if a file has been converted into a flac file and back again.  

Ha ha ha ha ha ha ha ha ha ha ha ha ha ha! No really, they're serious? Actually serious? They believe the sound quality deteriorates through conversion from WAV to FLAC and back again? Oh good grief.

If so . . . if they really (breaks off to collapse into giggles and slap the chair arm a few times) believe that the computer 'knows' that this has occurred, even though the files are identical, then they should come round to mine because I've got a really nice bridge in London to sell them. Hardly used or anything.

Well, could happen if the conversions aren't perfect - each pass degrading it slightly further: nothing to do with the computer 'knowing' if it's the first pass or not. So this finding points to imperfect conversion. Whether that is inevitable, or that some converters are better than others is another matter. If inevitable, then wav will always be better than flac provided it hasn't been flac'd first. If the converter, we need to know which are good and which not.

Solid Air posted:

...

One problem, I believe, with audiophilia is that many of its tribe were brought up in an analogue era where the flow of electricity from a mains switch through the whole system to the speakers was fragile and easily disturbed. Subjectivity was inevitable. Now, in the digital era, there is a fundamental break in that flow: the DAC. Everything upstream of the DAC is in the digital domain, and you can copy it on to a 1990s floppy disk, post it to yourself, zip it, unzip it, copy it back on to your XBox and then stick it on the NAS, as long as the file integrity is maintained, it makes no difference whatsoever. None, nada, zero. Only downstream of the DAC, in the analogue domain, do those old rules apply. 

Not quite:  Anything in the real-time part of the decoding chain can add timing errors (e.g. jitter) or RFI to the system, and those things can disturb the DAC and the other analogue components downstream.

In respect of the non real-time parts of the system:  Yes exactly.

@Innocent Bystander 

No, it can't. By their nature, the conversions are perfect every time. You can perform the conversion a million times, and the final copy will be identical to the first. This is the point - it's not subjective, it's just a process.  

Imagine this: imagine I convert long strings of numbers into letters, with 1=A, 2=B and so on. Except every time there's a string of zeros I put in the quantity of zeros instead. That would mean that 1745000000012 would become AGDE7AB. As you can see, I've compressed the number considerably, but it's lossless conversion - no matter how many times I convert it to and from, it will always end up the same. There's no subjectivity, it's a simple mathematical process. (Obviously, it doesn't actually work that way, I'm just illustrating a point.)

When you download a copy of Windows or OSX, many of the files are compressed, and they're uncompressed by the computer as they're installed. If the process were imperfect, every copy would be slightly different and your software wouldn't work. Come to that, neither would your bank transfers or, um, the internet. Or your television. This message wouldn't work. It's all converted, one way or another, often many times.

My point was, if the converter is NOT doing it right, and producing a bit-perfect compressed version each time. (Of course, that would mean each successive wav conversion back would change and so sound different as well

Innocent Bystander posted:

My point was, if the converter is NOT doing it right, and producing a bit-perfect compressed version each time. (Of course, that would mean each successive wav conversion back would change and so sound different as well

if the converter is NOT doing right, it is a buggy converter, and should be trashed.

FLAC is lossless which means no loss.
Unlike mp3 that if you convert back to AIFF and then back to mp3, you will have further degradation EVERY time you go from AIFF to mp3.

YanC posted:

Innocent Bystander posted:

My point was, if the converter is NOT doing it right, and producing a bit-perfect compressed version each time. (Of course, that would mean each successive wav conversion back would change and so sound different as well

if the converter is NOT doing right, it is a buggy converter, and should be trashed.

FLAC is lossless which means no loss.
Unlike mp3 that if you convert back to AIFF and then back to mp3, you will have further degradation EVERY time you go from AIFF to mp3.

Again, that is my point - I was merely saying that it does not require the computer to 'know' for it to be possible for the effect described. IF that serial deterioration was true as claimed to have been observed, it points to the converter not having been converting as it should. I am not suggesting that to have been the case, just pointing it out as an explanation, vs the clearly ridiulous suggestion that the coputer is somehow deciding to mess things up.

The above seems like a heated agreement!

The article isn't talking about changes in the data causing audible differences, it's talking about changes in the metadata (and hence changes in the handling of the metadata) causing audible differences.

The metadata in FLAC and WAVE files are usually held in a different formats (Ogg Vorbis vs [usually] ID3).  Conversion between these formats may or may not be lossless, and cumulative changes in successive conversion cycles cannot be ruled out.

Innocent Bystander posted:

Again, that is my point - I was merely saying that it does not require the computer to 'know' for it to be possible for the effect described. IF that serial deterioration was true as claimed to have been observed, it points to the converter not having been converting as it should. I am not suggesting that to have been the case, just pointing it out as an explanation, vs the clearly ridiculous suggestion that the computer is somehow deciding to mess things up.

This makes sense to me. There's a cumulative negative effect on sound quality reported, and failing another explanation this seems good enough to hang one's hat on.

Let's just agree that generally speaking WAV and AIFF are better

Adam Zielinski posted:

Let's just agree that generally speaking WAV and AIFF are better

IF they do - which may well be a function of the DAC used. But it shouldn't be an issue, as it is easy for virtually anyone to compare for themselves and choose their preference, and that includes comparing transcoding on the fly. With batch converters around that will convert existing files if presently in the 'wrong' format, should transcoding not be an option or not be as good as native. (... Of course, that i assuming that the converter doesn't degrade it and that the computer doesn't know what is being converted and deliberately corrupt it!)

Innocent Bystander posted:

Adam Zielinski posted:

Let's just agree that generally speaking WAV and AIFF are better

IF they do - which may well be a function of the DAC used. 

Surely it should be a function of the streamer / computer used as by the time it gets to the DAC, everything is decoded into PCM (or DSD) and no longer in any "format".

Very true.

My comment on WAV and AIFF was merely referring to a fact that WAV and AIFF are most common uncompressed file formats currently available (except for DSD of course).

Eloise posted:

Innocent Bystander posted:

Adam Zielinski posted:

Let's just agree that generally speaking WAV and AIFF are better

IF they do - which may well be a function of the DAC used. 

Surely it should be a function of the streamer / computer used as by the time it gets to the DAC, everything is decoded into PCM (or DSD) and no longer in any "format".

Ah, deliberate mistake spotted! Replace DAC with renderer in my sentence.

I started with a file  A1 16/44.1 AIFF (bought that way) converted it to FLAC using xACT compression level 8 into file F1. Then decoded F1 into A2 AIFF and again A2 into F2 FLAC and F2 into A3 AIFF.

I compared A1 vs A2 and wre different, beeing A2 slightly smaller (hundreds of bytes). After doing a binary comparison I detected only a very small portion of the files was different (may be as the article says related to metadata).

Then I compared A2 with A3 and they were identical.

I understand that may be the differences in SQ between A1 and A2 could exist due to the referred differences but between A2 and A3, any SQ difference is not due to the generation of the file but due to something in the environment affecting the renderer (RF, computer load, mains , etc) or the DAC.

I don't eat the statement that every new generation is going to sound worst than the former. It could be for the first encoding/decoding process but not for the following.

Regards.  Erich

Your results would be interesting if you compared the audio stream of bytes only, and not the entire file.

It maybe (for instance), that the 1st file carries the audio data split into 10 containers, while the 2nd file into 1 container. That would account for a few (hundred) bytes of difference.

OR that the 1st file has some metadata (such as the name of the app that created the file), that's missing in the 2nd file.

No SQ diff possible either way. 

Guys, trust the naysayers on this one. Lossless is lossless unless there's some buggy app involved. Life's too short to debate everything.

True, lossless is lossless - for the data only, NOT for the metadata that's also contained within the file.

They are rendering not from an extracted audio stream, but directly from the files in real time.  Therefore changes in the metadata can affect the performance of the renderer.  In theory, this should not happen if the rendering chain was perfect, but in practice it can happen.

There is no rendering involved. The post above refers to conversion from FLAC to AIFF multiple times. 

The audio stream will always be the same, regardless of whether some metadata may not make it across. If not, the compressor/decompressor is buggy.

There is rendering involved...

It's very difficult to estimate sound quality just by looking at the files... you have to render them to be able to hear anything!
(They actually did listen to the files [in each 'digital generation'] using a computer / DAC / amp / speakers:  i.e. a rendering chain.)

yeah, with a paintbrush ♬♫