[flac-dev] Flac HiRes decoding problems
Klaus Schulz
kls.schlz at googlemail.com
Tue Feb 21 01:22:39 PST 2012
Hi folks.
My first post over here.
While working on a music server optimzation (which ususally goes hand in
hand with running pretty low buffers on realtime
streams)
I figured that certain HiRez flacs were causing XRUNS/hickups on my Linux
server platform.
The interesting thing is to run into XRUNS even though the stream gets
send to a Squeezebox Touch which got
a 20s buffer on its receiving end.
However. Even a potentially short break of that flac decoding stream seems
to cause a problem further downstream.
What to do!?!? I did this and that. Nothing worked.
I tested the integrity of those files. I also run them as .wav. That turned
out to be OK.
I'm running flac 1.2.1 on a 64bit Ubuntu server btw..
The interesting thing is that not all flacs behave the same. Some are
running smooth. Others skip. Which is reproduceable.
Than I found a thread discussion in the Squeezebox Server area about this
very same subject.
The discussed solution solution was: RE-ENCODING to alias 0. Higher
compression level seemed to
cause those hickups. Ok. It's not just me.
I wrote a batch script and re-encoded all those flacs to alias level 0.
However. The problem didn't go away in my (rather special) case.
If I'm running the offline decoded .wav of those 24/96 files I do not have
any problems. Hmmh.
Next.
I then looked up the flac formatting guide.
I figured that the encoding options used by the majority of applications
are just the basic
aliases 0-8.
What struck me was that the flac formatting guide recommends higher
(whatever) blocksizes in
case of >48khz HiRez files and also a residiual partition order of 2,2.
Since I do not have a lot of background about the internals of the
decoder. I can't make much sense
out of the advanced features of the formating guide.
I'd really appreciate if somebody could tell me what best options to
choose for encoding HiRez flacs.
My goal would be to achieve the most efficient realtime stream decoding.
Probably I'd have to play with
blocksize and -r.
Another comment:
I'm wondering if the flac formatting guide shouldn't be updated with
information related to HiRes
data.
If the blocksize should be higher for Hirez as described in the guide there
perhaps should be new
aliases for Hirez material.
I guess the majority out there is still using use those standard aliases
which would apply for samplerates <= 48khz only.
Thx a lot.
Cheers
Klaus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.xiph.org/pipermail/flac-dev/attachments/20120221/05e1a54e/attachment.htm
More information about the flac-dev
mailing list