Quote Originally Posted by StanleyB View Post
I think that some of the more modern DAC chipsets are far better in detecting even minute differences in the audio signal. The compression/decompression process that a FLAC file experiences is probably leaving behind artefacts that are affecting the sound. Less detailed digital to analogue conversions are incapable of noticing these minute differences. I reckon that is why so few people have been able to notice the difference.
I compared the APE audio format against the FLAC and WAV ones. No problem with APE. So it is without any shadow of a doubt an issue solely related to FLAC and no other lossless format. But only the low jitter receiver chips seem to be able to pick out the problem. The common (and higher jitter) SC8416 fails to resolve the difference, whilst the superior low jitter WM8804/5 is far better at it.
But note that these are just my test results. Other DAC designers/testers/owners who have access to DACs with differing receiver chips might be able and willing to disclose their test results in due course.
I've followed this with some interest. Firstly, assuming FLAC is working properly, there is absolutely no loss of data, and a round trip of the form:

File1 -> FLAC compression -> FLAC decompression -> File2

should show that the data in File1 is EXACTLY equal to File2.
If this is not the case, then there is a problem with the compression or decompression process.

There is nothing approximate about lossless compression/decompression.

It is the case that some lossless methods may make use of a lossy component, which in the case of audio could be mp3 or aac. Essentially they do something roughly equivalent to a round trip compression/decompression cycle, and then subtract the result from the input to give a residual. The residual is then compressed and then stored and/or transmitted.

Various techniques are used for the digital encoding, such as run length encoding and Huffman encoding, both of which are not lossy.

It is claimed that FLAC files sound different from, for example, WAVs. This could be due to problems in implementation in either FLAC or WAV playback.
If the claim is that FLAC files sound worse than equivalent WAV, then clearly WAV is being used as the reference, and preferred. It does not automatically follow that a difference in perceived sound implies one is better than the other. Let's assume for the moment that FLAC does sound inferior.

With either format, the data is converted ultimately to a data stream which is fed into a DAC. If both FLAC and WAV based systems are working properly, the data stream to the DAC should be identical.

Ways in which they may then sound different should then only be due to timing issues, or possibly corruption in the data. Corruption is unlikely under normal circumstances. If the data is buffered sufficiently just before the DAC, there should be no difference. Note though that if there is any data corruption in compressed format data, that the audible effects could be worse (though not necessarily so) compared with similar corruption in other formats, because it is possible that data corruption in compressed data could lead to error propagation effects. This kind of situation can arise in video encoding/decoding in, for example, MPEG, where if the I frames are corrupted, the errors will propagate through frames up to the next I frame.

It does not follow that such corruption in compressed audio streams will always lead to significantly worse sounding output, as it's possible that the corruption is in part of the signal which has little perceptual effect. In both audio and video encoding and transmission, it is now common to protect parts of the encoded data which are more likely to have a damaging effect if they are corrupted to a higher degree of protection than other components.