Quibbles and Bits

Taking FLAC

You hear and read a lot of strange stuff in the world of high-end audio.  It can be hard to sift through what is real and what is rubbish.  Perhaps the gold standard for audio weirdness is still Shun Mook’s (in)famous ‘Mpingo Discs’. [Personally, I would point to the creams and foils sold for many years by Peter and May Belt—Ed.]  Made from a particular (and, naturally, rare) type of Ebony, and treated with a ‘proprietary process’, these wooden discs are placed on top of audio components such as preamplifiers, where, via a process described somewhat vaguely as ‘Sympathetic Resonance’, they are said to improve the sound of said component.  The little wooden discs can also be mounted groups of multiple discs (expensive, at $50 per disk) on little wooden racks, and disported about your listening room, where they claim to endow benefits that appear to border on the miraculous.

Mpingo Discs have been around for over 20 years now, and by all accounts people are still buying them.  There are countless reviews – some of them by credible names in the audio reviewing business – which appear to buy into the Mpingo Disc magic.  I myself have actually heard demonstrations of Mpingo Discs, but could not detect anything of any great substance.

There is a story about Albert Einstein paying a visit to another great physicist of the day, Niels Bohr.  Einstein was surprised to find a good luck charm – a Gypsy horseshoe – nailed to the wall above Bohr’s desk. “Surely, Professor Bohr,” he asked, “you don’t believe in such superstitious nonsense?”.

“Of course not,” replied Bohr, “but the Gypsy who sold it to me assured me it would bring me luck, whether I believed in it or not!”.  Thus it was that in the spirit of Bohr I kept an Ebony disc – which I picked up off the floor of a carpenter’s shop – on one of my loudspeakers for many years.  I even engraved a crude arrow on the surface so that, if needed, I could rotate it to a precise orientation.  I don’t believe it ever made any audible contribution, but I had a lot of fun with it trying to con gullible audiophiles.

The thing to bear in mind, though, when discussing the Mpingo Discs of this world is that resonance control in audio equipment is a real thing, and that – in principle, at least – anything placed on the surface of a preamplifier (or whatever) has the potential to interact with the inherent mechanical resonances of the structure.  Which, if nothing else, holds open the door to a discussion.

There is another discussion that has been going on for some time on the fringes of the digital audiosphere, which holds that compressed file formats such as FLAC and Apple Lossless (ALAC) don’t sound as good as uncompressed file formats such as WAV and AIFF.  Then there are related arguments, such as the one that says files containing metadata sound worse that files containing no metadata.  Finally there are utterly unrelated arguments such as the one that says the original files ripped directly from a CD sound better than copies of those files.  I know some big names in our industry who hold fast to those beliefs, and despite my best efforts I cannot convince them otherwise.

Many of these arguments founder on the rocks of a flawed interpretation of logic.  This is the notion that because you can set up an experiment in which a FLAC file can be heard to sound different from the exact same audio data stored in a WAV file, that this proves the claim.  But it doesn’t.  The raw audio data stored in both FLAC and WAV files is bit-for-bit identical – a simply provable fact.  Therefore it is the playback of one file type versus the other which is different.  It’s like when Bill Leebens and I both get out of my car.  Bill is wearing sandals and I’m wearing sneakers.  Bill’s feet get wet, but mine don’t.  This proves that sneakers are waterproof, and sandals are not.  Except that it proves no such thing … because there is a puddle on Bill’s side of the car, but not on mine (what, you think I’m gonna park with the puddle on my side?).

When reading a compressed file, before the music can be played the file has to be de-compressed.  With an uncompressed file it doesn’t.  It follows that there is more CPU activity involved in reading a compressed file than an uncompressed file.  Most of you will know that I work for a small company, BitPerfect Sound, developing audio playback software that runs on Macs.  Our mission is to optimize the sound quality of computer audio.  To that end, we have developed numerous correlations between things you can strictly observe and measure inside the computer, and things that subjectively sound better.  As a broad generalization, CPU activity is one of those things.  If you can reduce the CPU activity, then you will generally improve the sound quality.  Additionally, HDs are electronically noisy devices, and there is good evidence that computer-based audio systems tend to sound poorer when their HDs are active.  The same applies for SSDs as well.

Most playback software plays the audio stream directly from the file, so while it is playing music it is both reading the Disk, and extracting the audio content.  These activities will be slightly different if the file is a FLAC file or a WAV file.  Because a FLAC file is typically half the size of a WAV file, there will be half as much disk activity involved in reading it.  But because the FLAC file needs to be decompressed, there will be massively more CPU activity involved in extracting the audio content.  If nothing else, these mechanisms provide a basis for arguing that the playback of the two different file types can sound different.

To resolve this potential conundrum, therefore, we need to eliminate all playback differences from the picture.  Then, if there are any fundamental differences in sound between identical audio data stored in FLAC and WAV file formats, these can unambiguously emerge.  To a large extent, this is what BitPerfect, in the company of a select few other no-compromise audiophile playback Apps such as Audirvana, is able to do.  BitPerfect does not stream the audio data direct from the disk.  It pre-reads the file, and loads the audio data into RAM, where it sits in its native form as a raw PCM bitstream ready to be transmitted directly to the DAC.  Any processing that may be called for – such as sample rate conversion – is also done in advance.  Once the file has been read, decoded, pre-processed, and the raw data loaded into RAM … at that point we have identical raw audio data located in the exact same memory location – ready for playback – regardless of whether the file it came from was WAV, FLAC, AIFF, Apple Lossless, or whatever.  From that point forward, the original file format can have no impact whatsoever on how playback proceeds.

To me, given the above scenario, there should be no detectable differences in the sound of the music, and indeed, I am utterly unable to perceive any.  With BitPerfect, the process of reading the file, decoding it, pre-processing it and loading it into RAM typically takes between 2 to 5 seconds.  During those brief seconds, I would concede that the potential exists for an audible difference to be present.  But it is beyond my capabilities to detect subtle sonic differences in such a short time.  Believe me, I have made many repeated efforts to compare ALAC vs AIFF (the Mac equivalent of FLAC vs WAV) over the years.  I maintain a significant proportion of my music library in various file formats so that I can readily call up a comparison if and when I wish to do so – being a software developer I need to do far more critical listening than the average sensible audiophile (if indeed there is such a thing).

One final observation.  Of all those who have told me they hear differences between FLAC and WAV files, as well as some of the other strange things I mentioned in the opening paragraph, one thing they all have in common is that all of them listen using Windows PCs.  I can’t help but wonder whether that is mere coincidence.