[VNandor]

I remember this website as it has already come up already at a different thread:

Well have an open mind test it for yourselves:

A Fool For You by Carmen Gomes Inc in FLAC 24/352
A Fool For You by Carmen Gomes Inc in WAV 24/352

From the publisher:
The song is called 'A Fool For You' and is performed by Carmen Gomes Inc.
It’s taken from the album 'Carmen Sings The Blues'.

http://soundliaison.com/studio-showcase-series/276-carmen-gomes-sings-the-blues

All the different formats have the same source file, DXD 352kHz (Studio Master).
We used the XiSRC app for the conversion to DSD and Steinberg Wavelab for the conversion to the other PCM formats and FLAC.

Multi-format Album is free: https://www.soundliaison.com/index.php/6-compare-formats

The people that uploaded the files unfortunately don't know how to properly convert wavs to flacs because the PCM data contained in them is different. I strongly recommend avoiding this website and doing the conversion yourself if you want to do a proper listening test about wavs and flacs and not the incorrectly converted files uploaded here.

 

[CAJames]

I don't think the issue is you get different bits from the different (lossless) file formats. I think the issue is whether the amount of processing needed to play the different formats affects SQ. JMO.

 

[VNandor]

An audio DAC can only receive PCM or sometimes DSD data, so it does not do any extra amount of processing to the signal. The hardware that runs the playback software does process the FLAC in order to extract the (bit perfect, if the conversion is done properly) PCM data. Some players store a "sizeable" chunk of this PCM data in the memory but typically not the full length, since it would be insane to take up gigabytes of RAM whenever an album gets loaded into the player. Some players use much smaller chunks, essentially just the buffer size to store the PCM data.

FLAC was released in 2001. I think we had Pentium 4 processors around that time. Rockbox came out like a year or two later and it gives support for playing back FLAC files to portable players. The portable players had enough capacity to play them back. In 2023, I doubt that a computer or a portable DAP would struggle with the converting FLAC to PCM. Of course the SQ may be less than perfect but I think that the above strongly suggests it is unrelated to FLAC or any other lossless compression being used.

 

[CAJames]

...FLAC was released in 2001. I think we had Pentium 4 processors around that time. Rockbox came out like a year or two later and it gives support for playing back FLAC files to portable players. The portable players had enough capacity to play them back. In 2023, I doubt that a computer or a portable DAP would struggle with the converting FLAC to PCM. Of course the SQ may be less than perfect but I think that the above strongly suggests it is unrelated to FLAC or any other lossless compression being used.

What I know for sure is my laptop was running at only a couple percent CPU when I was playing files in Foobar. But turning off a bunch of unneeded services, while not changing CPU use in a meaningful way did improve the SQ noticeably. I don't think processors "struggle" playing files, but I think small changes in overall load can affect SQ. Similar to digital cables. JMO/FWIW/YMMV and all that.

 

[VNandor]

Is this effect enough to prefer standard sample rates over hi-res (for your friend at least)? Playing back a 192kHz file is defintely going to put a higher load on the CPU than playing back a 48kHz file. There are people using HQplayer to upsample their music real-time, which something that actually puts a real strain even on today's CPUs but they still say it increases the sound quality.

 

[bmichels]

Very interesting feedback here, and what about DSD ? Should the DSD versions of a particular track sound better than the PCM version (with Modern/powerful DAP) ?

 

[CAJames]

For me (in general) hi res PCM sounds better than 16/44 and DSD sounds better still. But for me FLAC and WAV sound the same so go figure.

My friend only listens to 16/44 AFAIK.

 

[mrjayviper]

AIFF and WAV both encode the audio as PCM. The difference between them is how they encode metadata. FLAC (and ALAC) compresses and encodes data without any loss. Once they are decoded to PCM, the resulting PCM is going to be the same as the one contained in WAV (or AIFF). This can be verified by a "null test", by creating a FLAC file from the WAV, decoding the FLAC to get the PCM data, and comparing the stored sample points to the one contained in the WAV by subtracting them from each other. The result is going to be 0 if both PCM data coming from the WAV and FLAC are the same.

md5 of the wav vs flac-converted-to-wav is sufficient test?

 

[McCheese]

What I know for sure is my laptop was running at only a couple percent CPU when I was playing files in Foobar. But turning off a bunch of unneeded services, while not changing CPU use in a meaningful way did improve the SQ noticeably. I don't think processors "struggle" playing files, but I think small changes in overall load can affect SQ. Similar to digital cables. JMO/FWIW/YMMV and all that.

There's no reason to believe that system load on your CPU or SoC which decodes the FLAC file into a PCM stream has anything to do with sound quality out of the DAC because they're completely separate chips and that DAC chip is getting fed the same bits regardless of whether it's a FLAC, WAV, AIFF, or ALAC file. It doesn't know the difference.

Exception of course being instances where the system cannot keep up with the realtime demand of feeding the stream to the DAC, but that shows up as artifacts that are very obvious to the ear, e.g. the music stops and buzzes for a half second before continuing.

 

[gregorio]

Thanks, but IF storage space is not an issue, then what sound best: FLAC or AIFF ?

Assuming the FLAC is made from the same aiff (or wav) file then it must sound the same.

Download it (or the whole album from Sound Liaison) and test it yourselves. If no difference then fine. If there's a difference, then fine too.

If there’s a difference that’s not fine at all, because it means either there’s something seriously broken with your system or one of the files has been corrupted/changed.

G

 

[Designsfx]

Go uncompressed WAV all the way. In the FLAC to WAV comparison some cannot detect a difference but to me WAV has always sounded better. Given the cheap cost of storage the concerns over saving disk space are not what they once were.

 

[gregorio]

Go uncompressed WAV all the way. In the FLAC to WAV comparison some cannot detect a difference but to me WAV has always sounded better.

Some cannot detect a difference because there is no difference (unless your system is seriously broken). FLAC is WAV (or whatever else you encode with it), so a “FLAC to WAV comparison” is actually a WAV to WAV comparison and therefore you’re claiming that WAV has always sounded better than WAV. So users can go either way, because they’re identical.

G

 

[VNandor]

md5 of the wav vs flac-converted-to-wav is sufficient test?

No, the MD5 doesn't test the PCM part of the file, it tests the file as a whole. You could have the same PCM data in two differently named wave files and they would still have a different hash (although the different hash would not come just from the different file name itself). Doing a null test is almost as easy as checking the MD5 hash and it actually tests the PCM audio data, not metadata, encoding and such.

 

[McCheese]

It's true that WAV files can contain metadata other than the PCM stream, but just a small nitpick I want to make is that changing the name of a file doesn't change its hash value.

 

[Designsfx]

Some cannot detect a difference because there is no difference (unless your system is seriously broken).

My point was not to dispute the “bit perfectness” of FLAC nor to start an endless drill down on codecs but to rather state my observations that FLAC can be perceived differently (given certain circumstances/equipment) when played back in real time. I don’t think the equipment has to be “seriously broken” as much as not being designed for great sound reproduction.