File-based audio playback - compressed vs. uncompressed
Hi,
Following on from the debate in another thread, I thought that it would be best to start a new thread on what is obviously an area in dispute.
As more and more of us move to file-based playback solutions, I think that it would be useful to explore the anxiety about storage of audio files using FLAC containers (as opposed to WAV/AIFF files). For the record, all of my CD rips exist as FLAC files, encoded directly via dbPowerAmp. I have done some a/b testing between WAVs and FLACs without being able to hear a difference, and have compared WAV and FLAC streaming to the SB Touch without really being able to notice a consistent difference.
My take on this issue is this:
1) Some people claim that FLAC files sound different to the source WAV file
2) Some futher claims that converting from WAV to FLAC and back to WAV again affects the sound, despite the files having identical checksums (i.e. every bit in them is the same). I cannot see how this could be possible.
3) If there are audible differences, they are implementation issues with the playback software / audio driver stack / hardware, not the file format or data. If this is so we need to identify and fix the implementation issue not work around the problem (if it really exists).
As choosing a storage format for your audio library is going to be pretty fraught if you believe that subtle degradation is going on at the data level, I think that this should be exploded once and for all and then we can all relax and get on with enjoying the music again.:)
My view is that while the file is residing on a computer, i.e. not being played out, it is simply data, and can be packed or unpacked without changing at all. There is nothing special about audio data on a hard disk compared to any other type of data, and were this not so I wouldn't be able to be posting this now as data corruption would prevent my computer and the internet from working.
Your uncompressed audio files are, in fact, already compressed by the hard drive controller before being stored as magnetic domains on the disk surface by run length limiting, and we seem to accept that there are no issues arising from that process.
I read an interesting post by David Snyder on computeraudiophile.com, which I have reproduced below. I think it sums up the position well.
Hello,
At the heart of this debate over how the handling of digital music files (adding/altering metadata, transfer through email, and conversion among lossless formats, etc.) may have a residual impact on their sound lies each person's definitions for "data" and "music", and if there is a distinction, at what point the transformation occurs.
If one believes that a WAV, AIFF or FLAC file fundamentally is music, then audiophile notions of how that music should handled can come into play. Music is intricate, fragile, and must be handled carefully to preserve as much detail from the original recording as possible. Transfer over wireless networks should be avoided. Before downloading from an on-line music store, be sure to use a high quality Ethernet cable of no more than 5m in length, directly connected between your PC and DSL router or cable modem for shortest signal path. Intel network cards are better at preserving mocrodynamics, but Broadcom NICs tend to deliver digital downloads with a larger soundstage. Save your downloads directly to a firewire or internal SATA hard drive (avoid USB and SSD because of higher noise levels). Downloads done late at night tend to have better sonic purity than downloads in the middle of the day due to less traffic on the network. Make sure that both your computer and your router are connected to a good power conditioner during the download and any subsequent transfer. Do not convert your downloads to other data formats...there's no such thing as lossless conversion or especially lossless compression! Do not attempt to add or correct the metadata since doing so can have a subtle but irreversible impact to the sound quality of your digital music files.
Contrary to the view above, I personally believe that a WAV, AIFF, or FLAC digital music file is just data. It's not music or sound any more than a spreadsheet, expense report, or last year's tax return. All data files are nothing more more than sequences of bits: 1s and 0s. The contents of these data files only have meaning within the context of application software that can read and interpret them. While the physical properties of storage media and signal transfer can have an impact on the integrity of data, modern computer systems and data networks have mature and highly reliable ways of detecting errors and correcting them with bit-perfect accuracy within multiple technology or protocol layers below the application. In the rare event that an error can not be handled transparently, there are facilities to notify the application and user of the failure so that corruption does not occur unnoticed. While it's possible that low level corruption can escape detection, this is so rare that most computer audio systems will not be affected by it during their years of useful service.
Since a digital music file is no different from any other data file on your computer, the same handling rules apply. You can make a copy of a copy of a copy of a copy of...a copy of a file one hundred times and the one hundredth copy of the file will be bit-for-bit identical to the original. There's no generation loss. It doesn't matter if you transfer/store files using wireless or wired networks, USB or firewire storage, SATA or SAS disks, CIFS or NFS network shares, Mac or PC hardware. The mechanisms for protecting data file integrity work the same regardless of the file extension, type or contents. If you find this not to be the case with your particular computer system, you should find out why and have it repaired.
So, at what point does data become music? In a computer audio playback system, at what point in the playback chain does audiophile handling of the music come into play and actually matter? These are important questions. I believe that this transformation occurs when time is applied to the amplitude values in our data files. It's time, after all that makes sound (and therefore, music) possible.
All of the bits in a music data file are frozen in time; they exist in whatever state they are in all at once and unchanging (unless something goes horribly wrong with the hardware). When those bits are interpreted as amplitude values and those values are associated with a uniform series of clock pulses in real time, at that instant the data begins its journey through our playback systems. It's at that instant (and not before) that handling of these delicate signals can have an impact on the sound that eventually reaches our ears.
While the signal is still in its digital form, jitter, or tiny variations in the timing of the clock pulses can have a significant impact on sound. Computer audiophiles use high end USB to S/PDIF converters with ultra high quality clocks, DACs with buffering and re-sampling, audiophile USB and S/PDIF cables, and all sorts of other techniques to minimize jitter before the amplitude and timing information reach the DAC chips. Once the signal leaves the DAC, at least the same level of care is taken to preserve the music in its analog electrical form until it reaches our loudspeakers. At that point, the signal becomes acoustic energy and so acoustic treatment and design become important.
At all steps in the signal path past the point at which time is applied to to the data, audio engineering approaches to signal transmission and handling come into play. Prior to that point, computer science and best practices that apply to data integrity are all that one must concern themselves with in the world of computer audio.
I went over a lot of details here which I hopefully got right. I'm confident that any mistakes will be corrected by folks on this forum in short order and for everyone's benefit, including mine! My purpose in all of this (including the tedious exercise, which I hope you took part in as well) has been to hopefully dispel the notion that modern computer systems are incapable of reliably handing music data files without introducing minute levels of corruption or degradation. If it were true (or commonly accepted as true), this notion would be damaging to the growth of the computer audio industry. Audiophiles who do not currently purchase high resolution digital files or USB DACs may not be willing to even give them a try if they believe that computers can't be trusted to handle these files without loosing bits here and there. Thankfully, computers seem handle data just fine, so let's give the all clear get back to listening and enjoying music.
-- David
Let the debate begin.
Regards,
Alex
