> My DAT records at 48 kHz so I need to convert this to 44.1 kHz in order t=
o
> process the sounds on my computer. To do the conversion I simply feed th=
e
> sounds from the DAT to my iMac via analogue inputs, where the sound is
> reduced to 44.1, the call is less pleasing and rather sharper, and better
> listened to at a lower volume. There isn't a lot of difference, but you =
can
> hear it.
Vicki,
I'd be curious if this is an artifact of the sample rate per se, or the
conversion process -- my money's on the latter, but my mind is open. The
software I use (Samplitude) has some very high quality sample rate
conversion tools that would allow me to covert your files in the digital
domain. (I had to do this recently in the other direction to support a
performance environment that could only accept 48Khz files...)
Would you be willing to send me a copy of the source files on DAT or
digitally transfered at 48Khz to CD-ROM? Or can you host a representative
excerpt on a website (or email it to me)?
I'd be happy to send the results of various re-sampling techniques back to
you for your comments, which I'm sure others would be interested in as
well (or I could simply host them on my own site...).
---
On the topic of ATRAC and conversion -- I've been wrestling with an
unusual issue myself: differences in the de-compression to WAV of MD
tracks in two arenas. My belief that DEcompression should always be
identical (regardless of ATRAC ENcoder version...) has been challenged.
Originally I used a Sony JB920 deck with SPDIF outs to transfer files
realtime from MD to my PC for editing. The ATRAC data (recorded on a
variety of portable MD decks) was (of course) decoded by the JB920, then
the reconstructed digital waveform was transported as SPDIF to my
soundcard (very good quality, Yamaha or recently RME).
To save time, I invested in the MD Transfer Editor system from EDL in the
UK, which lets you use the old Sony MDH-10 data drives to 'rip' ATRAC
files directly to your PC via SCSI, to be decoded in software into WAV
files (or archived in their compressed state!). I've been using the system
for a few years or so and it's saved me countless hours.
However!
I've noticed that the files generated by these two methods are not
identical. I had not thought much of it until this year when it became
clear that the files being decoded by the EDL software have distinctly
more pronounced artifacts than the old realtime transfers -- something
Jeremiah Moore on this list also mentioned (with no prompting) when I did
some ripping for him. Particularly when handling certain kinds of
challenging (low-volume, very complex) material.
I've investigated more seriously recently, and am waiting for comments
from EDL.
If you're interested, here are representative clips, which are
perfectly sample aligned; each is about 1.7 MB:
http://www.quietamerican.org/download/mdte/flags_mdte.wav
http://www.quietamerican.org/download/mdte/flags_sony.wav
http://www.quietamerican.org/download/mdte/flags_diff.wav
the diff is the result of inverting the polarity of one side and bouncing
it against the other -- if they were bit identical this should of course
be silence. Btw I'm quite confident that no digital "gain" was introduced
at any point -- both transfers should be 'straight.'
The source recording is of prayer flags flapping in Tibet: something very
complex, broadband noise-y, approaching white noise. Exactly the kind of
material in which any compression scheme is maximally taxed -- and
exactly the circumstances in which the artifacts I hear are most
pronounced. I tried to pick a clip that would reveal them ~ I'll refrain
from saying more to avoid prejudicing anyone's reaction.
I had two or three ideas about the differences I'm hearing:
(a) the Sony is doing some un-specified post-processing to handle cases
where it must generate a lot of 'filler' white noise, to insure that it's
either masked by higher frequency noise, or as non-periodic as possible;
(b) EDL is not implementing the ATRAC decode recipe correctly;
(c) there are ambiguities in the recipe when the compression is highly
taxed and I subjectively like Sony's handling better;
(d) there is some problem in my SPDIF transfer (eg jitter) that happens
to produce subjetively appealing distortion in these cases.
....
Any opinions or analysis most welcome -- one thing I *haven't* done (and
haven't the tools/expertise to do well) is do a spectral analysis of these
clips. I can hear differences -- but only characterize them crudely.
Remember -- both these signals were transfered and decoded entirely in the
digital domain... :/
Best regards,
aaron
http://www.quietamerican.org
________________________________________________________________________
________________________________________________________________________
|