Why Do WAV And FLAC Files sound different?

A very interesting and comprehensive article Why Do WAV And FLAC Files Sound Different?.
There’s seemed to point that meta data may degrade sound but without it there’s no way to manage our music.

Ugh… that article was depressing to read. This quote:

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.

…a simple programmatic comparison between the PCM data in the each conversion is either the same or it isn’t - you don’t need listening tests. If the data doesn’t match, after transcoding, then it’s what’s known as a software defect, which I can’t imagine is the case with a modern converter like XLD.

Most folks really don’t understand what LOSSLESS COMPRESSION really means.

Having said all that, I’ve personally measure CPU utilization when unpacking/playing different container formats for the same music files. Apple Lossless is particularly difficult to unpack, as metadata is allowed to exist almost anywhere, and on a low-powered machine (e.g. a dedicated playback appliance running a low-power CPU), I’ve seen ALAC consume an order of magnitude more CPU than a corresponding WAV container of the same music data. The real question is whether or not that translates into anything audible… YMMV.

I wish audiophiles without a CSci background would stop writing reviews and articles about computer audio and go back to worrying about cables and interconnects.


[quote=“Krutsch, post:2, topic:12487”]
I wish audiophiles without a CSci background would stop writing reviews and articles about computer audio and go back to worrying about cables and interconnects.
[/quote]Agreed, it’s very easy to prove that a wav taken to FLAC and back to wav are identical, no matter how many iterations and different lossless transcodes one goes through. I wonder whether they find their documents read differently after having been zipped and subsequently unzipped, perhaps the fonts won’t be quite as crisp. People that write crap like that article should be denied oxygen, the gene pool could definitely do with some chlorine.


Someone who’s opinion on audio gear I respect once gave me an analogy for FLAC vs. WAV: if you take a piece of paper and crumple it up and then un-crumple it and wipe it flat on a table, you still have all the paper, but is it really the same as before?

I really didn’t know what to say, so I just returned a blank stare. I am surprised how many people struggle with this concept and, of course, drive themselves crazy worrying about file formats and metadata.

I use the above anecdote as a reminder to myself of the power of expectation bias, whenever I start wondering if I need some tweak or another.

I’ve no doubt there are tweaks that make differences at the margin and I guess those inclined to experiment should. As a friend put it, “if someone perceives a difference, then it’s real to them…leave them be” and I’m happy to until they start proclaiming these perceived differences as universal truths that cannot be measured because pick one of 1) science doesn’t know what to measure, 2) measurement equipment isn’t yet up to the task or 3) DBT is flawed.

In the case of lossless files being taken through iterations of transcoding it’s a simple matter of proving they’re bit for bit identical using by simply calculating a hash of the audio contents using any one of many hash algorithms. But, of course, pundits would find a reason to argue the results invalid and having no bearing in audio.

I suspect Kenneth’s comment on CPU utilisation when unpacking is at the heart of any perceived differences here. As I understand it, noise and timing issues arising from the computer source can be overcome by using an Ethernet output to a low footprint network device which then feeds a USB input to the DAC. If anyone was inclined to do an A/B/X test between formats then I’d be interested to see how the results varied between architectures.

I certainly do not agree with the notion that conversion between FLAC and WAV degrades the quality of the file. I have done this myself and CRC’ed the results where, of course, lossless files remain identical even after conversion into ALAC and AIFF.
But, nevertheless, i also have a preference for the sound of AIFF/WAV/PCM files over ALAC/FLAC. I have tried this on several occasions and always prefered the uncompressed versions. Of course, this is debatable if you are so inclined… :wink:
And it is not about the metadata! The RIFF chunks containing this is placed at a distinct place within the WAV/FLAC/AIFF container and does not mix with the music.

My compromise solution is to store my music in uncompressed FLAC, which does sound basically as good as AIFF to my ears and in my system.
It is also compatible with all of my playback gear, soft or hardware.


Edit: Clarify by spelling better! :slight_smile:

Using Roon the “work” involved in decoding is done at the server, so if you’re using a Roon Ready endpoint or Roon Bridge enabled endpoint as transport the transport and DAC are receiving PCM, ergo there should be no differences as the CPU activity at the endpoint will be the same regardless of what container the file was decoded from on the server side.


I’ve seen measurements done on different playback software playing back the same file, in the same format, and the PC in question puts out different amounts of noise, depending on which playback software is being used. Same for WAV and Flac versions of same file. This is regardless of the issue of if the WAV and Flac convert perfectly to each other in a non-playback situation.

So I can imagine a scenario in which this affects the ultimate sound of the WAV and FLAC files being played back.

That said, I don’t hear the difference. I just don’t think the idea that the formats may sound different to some people in some setups should be dismissed out of hand.

As long as it is mathematically lossless, the correct checksum and bit to bit perfect then many may assume there were no change in sound quality…This is however, in computer term. If one put it into an audio chain then it is not just the data integrity, noise, jitter and latency due to additional processing can affect the sound quality. When I set out the goal to rip all my CD collections many years back, I decided to go to the shortest, simplest path with least the processing, WAV with meta tagged.

1 Like

You chose a wise path! I have tried the same, but unfortunately my streamer cannot handle metadata within WAV-files… The metadata gets the audio treatment and plays back as bursts of white noise when changing tracks… :frowning:

Thats one of the reasons i have chosen uncompressed FLAC as my chosen format.

I agree, I have no business telling someone what they can or can’t hear in their system. However I have little to no patience for people telling me what I should or shouldn’t hear and if I don’t agree, then I can’t hear the difference because I haven’t spent enough money.

This reminds me of the people who claim to have a “special” USB stick for moving files to their music server that “magically” aligns the bits for a “night and day” improvement in sound quality versus other transfer methods.

[quote=“mrvco, post:12, topic:12487”]
This reminds me of the people who claim to have a “special” USB stick for moving files to their music server that “magically” aligns the bits for a “night and day” improvement in sound qual
[/quote]kinda drives home the notion that audiophiles are not necessarily intelligent people

1 Like

Wow these guys have a serious misunderstanding of how lossless audio works. Really a profound lack of understanding of core digital audio terms and their meanings. I don’t care how many times they transcode the flac and wavs streams they will be exactly the same as the original.

1 Like

Credit where credit is due: I love the graphs. Nice, bright colours. What’s not to like?


It’s really funny (or sometimes really frustrating) to see what strange magical components some people see in data processing on a computer … if the data is music.
No one would question the bit perfect result of uncompressing a zip-file with excel sheets, even if the cpu is really busy when unpacking the data. But if it’s music … everything is possible.
That’s beyond my comprehension.

1 Like


Your computer waits for all the data to get there, be checked, verified, and THEN outputs the data to you. You have to wait for that zip file to unpack while it’s verified.

However, once the timing thing is accounted for (probably buffered), I agree.

People don’t generally care about latency (particularly changing latency) and noise when unpacking a spreadsheet; when the data is music and they are trying to play it, then those things become important if their effects can be heard.

Several years ago I bought the rights to the domain name utterbollocks.com with a view to starting a website with my favourite items I considered to be… well, I’m sure you can guess.

It would have been a personal list, inherently subjective and open to criticism. Just a bit of fun, really. But in the end I abandoned it once I found it too hard to whittle down the vast number of entries to a manageable size.

This would have made it though.

The Methodology section alone caused much hilarity involving contortions on the floor and my bottom becoming detached. :grinning:


I can accept as worthy of debate some of the contentions about realtime computing and RFI affecting D to A conversion (though this should still be possible to remove with adequate system design)

But the article in question is about sound quality degradation being reliably correlated to metdata concatenation using tape measures on a back wall where trained listeners can accurately tell you the height of instruments.

I really wish I’d never let my domain name go now.

1 Like