Since it is possible to load samples in to Saol as MPEG audio files
for saving bandwidth, I examined if MPEG layer 3 audio is able to store
samples properly. The answer seems to be *negative*!
Would somebody please verify my observations?
My samples are at
http://www.funet.fi/~kouhia/alignwav.bin --- original audio
http://www.funet.fi/~kouhia/alignmpg.bin --- encoded audio
http://www.funet.fi/~kouhia/aligndec.bin --- decoded audio
("bin" endings are because Netscape was unable to load those files to disk.)
I used the MPEG audio example encoder (distribution 10, Feb 1997, from
Hannover site) to encode the original audio. Even the encoder is quite
old, it is the latest freely available.
For decoding I used all mpg123, splay and the MPEG audio example decoders.
mpg123 and splay produced indentical results but the example decoder
additionally produced garbage at the end of audio. The decoded audio
available above is the one produced with mpg123.
Problem: all decoders shifted the audio forward and dropped the end
(about 700 stereo samples). Such approximate handling sure is not suitable
for handling instrument samples.
Currently I have no idea where the problem is. From the transcript of
the encoder, I clearly understood that the encoder loaded all given samples
(because it padded the incomplete and last frame with zeros). Thus, I would
like to know if align.mpg still has all samples even the decoders won't
output them.
I'm in believ that all freely available decoders are based on the example
implementation or on the Fraunhofer example decoder. Could somebody verify
if the commercial Fraunhofer decoder (and encoder) is working properly?
I hope this misalign problem will be removed as soon as possible.
Yours,
Juhana
This archive was generated by hypermail 2b29 : Wed May 10 2000 - 12:14:10 EDT