File-Based Audio Done Right at John's

[Gazjam]

Gaz, there should be no difference in the playback of AIFF vs WAV. Both are a wrapper / container for PCM data, yes AIFF is stored little endian and WAV big endian...this makes no difference! Yes the headers are different but we are not playing the headers. There must be some other reason for your perception of better playback with AIFF.

For reference I've not heard any difference with FLAC vs WAV either when first decoded on my squeezecentre server and transmitted to the squeezebox, decoding on the touch though is a slightly different matter but in reality its all but inaudible.

Oh, forgot theres two versions of AIFF one little endian and one big endian...

I hear you Neal, on paper they should be the same.
Listening though can be a different matter

member of the "can hear the difference" camp here - very much so.

[nat8808]

The thing with upsampling to 96 from 44.1 is about the algorithm to an extent but more about how that algorithm handles the fact that 96 / 44.1 is only accurate to 21 decimal places. (2.176870748299319727891).

Using that number then will cause non-terminal decimals (ones that don't end) or others that are beyond the scope of the bit handling capability of the processor and it then down to the algorithm to work out how to handle them. Some processors even go a bit funny when reaching limits and if the algo author doesn't know the chip it's running on well enough then they won't be able to work with the specific limitations. Easy way would be to make a limited scope algo that rounds up or truncates to less accuracy than is possible etc etc.. Hence, personally, I'd trust most those by expensive hardware manufacturers like dCS or other studio gear or software that costs loads and pays the work needed to get algos right. i.e. wouldn't trust cheaper software with complicated upsampling where the algo is probably some off the shelf maths.

Of course, this is much much harder when done on the fly rather than upsampling the file and saving as a new one, in which case the software can take as long as it likes to do the calculations.

Again, whether you can hear it or not is another thing!

[NRG]

I cant question your experience, I just cant see how and it is polar opposite to mine.

I hear you Neal, on paper they should be the same.
Listening though can be a different matter though

member of the "can hear the difference" camp here - very much so.

[nat8808]

Then again, for other reasons (and you get a conversion factor of only 5 decimal places to have to deal with) Benchmark upsample to 110KHz.

Their reasons are given here:

http://www.benchmarkmedia.com/discus...-to-110kHz.pdf

[nat8808]

Gary and Neal blind test experiment! Fight!

[Gazjam]

I cant question your experience, I just cant see how and it is polar opposite to mine.

Well, that's hifi for you
always for me the ultimate measuring device is my lug holes....
(usually sanity check it with another bod or two)

[MartinT]

Wonder why, receiver chip?

That's my understanding. 24/192 sounds superb, though. In any case, I've never seen a commercial 176 recording.

[NRG]

Gary and Neal blind test experiment! Fight!

No fight here Nat! I respect Gaz experience, I'm just trying to understand.

[MartinT]

Gary - I know you've done a lot of work on the MCRU server. Have you found that upsampling files (not on the fly) has given any worthwhile improvement? I'm simply amazed at what John and I heard with SOX file upsampling.

[Gazjam]

Depends on a lot of factors Martin tbh.
Upsampling generally I found can make a difference, but its not always a preferable one.
It was an option I looked at during testing.
I found the improvement varied too much between type of music played, what Dac was used...so I wouldn't advocate it quite honestly.

There's an advantage to not doing it on the fly as it increases the workload on the system, but in my experience removing the need to uncompress the Flac file (by converting to PCM) gave a bigger improvement than upsampling in general.

For instance a 16/44 file upsampling to 24/192...gives you what?
Imo A lot of bit padding and increased file size for a subjective improvement in sound.

Be interesting for you guys to blind test a 16/44 Flac up sampled to 24/96 compared to the same Flac converted to AIFF PCM.
It's horses for courses, I don't think there's a definitive answer to this one, except to apply the usual caveats?