Torsdag, 23 januar 2003, skrev Miroslav Lichvar <lichvarm@phoenix. inf.upol.cz>:>If input is 16 bit, side channel will be 17 bit (16bit - 16bit is >17bit number). And warmup samples will be (17 - wasted_bits) bit.Voila, this was the source of all my frustration, sync problems and who knows what. My decoder now works correctly for files encoded with the default settings. Are there any decoder parameters I can set to get an encoded file using as many "features" as possible, so that I can stress my decoder and hopefully find all bugs? Tor ==================================================================EASY and FREE access to your email anywhere: http://Mailreader.com/ ===================================================================
On Thu, Jan 23, 2003 at 04:44:06PM +0100, Tor-Einar Jarnbjo wrote:> Torsdag, 23 januar 2003, skrev Miroslav Lichvar <lichvarm@phoenix. > inf.upol.cz>: > > >If input is 16 bit, side channel will be 17 bit (16bit - 16bit is > >17bit number). And warmup samples will be (17 - wasted_bits) bit. > > Voila, this was the source of all my frustration, sync problems and > who knows what. > > My decoder now works correctly for files encoded with the default > settings. Are there any decoder parameters I can set to get an encoded > file using as many "features" as possible, so that I can stress my > decoder and hopefully find all bugs?Have a look at test/test_streams.sh script from flac package. If your decoder has a command line interface and can output to raw data, modify the script in way that your decoder will be used for decoding. I know, overkill, but then you can be pretty sure that it is decoding well. At least try to decode files encoded with -1, ..., -8 options. -- Miroslav Lichvar
Tor-Einar Jarnbjo
2004-Sep-10 16:45 UTC
AW: AW: AW: [Flac-dev] Incomplete format description?
> -----Ursprungliche Nachricht----- > Von: Josh Coalson > > Miroslav's suggestion is the best right now. I probably should > hand-craft a stream that exercises a decoder as much as possible.I've already tested my decoder with the options -0 to -8 and it works, but the streams doesn't contain any verbatim or constant subframes. I guess this depends heavily on the input file. Unpredictable noise will be coded as verbatim frames and periods of complete silence as constant frames? I'll see if I get to run the test-script mentioned, but I am not sure if my Linux installation is working very well. I've made the current version of the decoder available here: http://user.cs.tu-berlin.de/~bjote/Ogg2Wav.jar It has no JMF-bindings yet, only support for Ogg-framed FLAC-files and at least my code for writing the WAV-file only supports 16-bit files, but if someone would like to test it, feel free to do so. The decoder is started like this (Java 1.4 is required): java [-Danalyze=true] -jar Ogg2Wav.jar in.ogg out.wav The analyze-option will produce similar output as flac's -a option for fixed and lpc subframes. Tor
--- Tor-Einar Jarnbjo <Tor-Einar_Jarnbjo@grosch-link.de> wrote:> Torsdag, 23 januar 2003, skrev Miroslav Lichvar <lichvarm@phoenix. > inf.upol.cz>: > > >If input is 16 bit, side channel will be 17 bit (16bit - 16bit is > >17bit number). And warmup samples will be (17 - wasted_bits) bit. > > Voila, this was the source of all my frustration, sync problems and > who knows what.Cool, it looks like Miroslav covered everything... but it's good you asked because I can use this thread to make the format docs clearer.> My decoder now works correctly for files encoded with the default > settings. Are there any decoder parameters I can set to get an > encoded > file using as many "features" as possible, so that I can stress my > decoder and hopefully find all bugs?Miroslav's suggestion is the best right now. I probably should hand-craft a stream that exercises a decoder as much as possible. Josh __________________________________________________ Do you Yahoo!? Yahoo! Mail Plus - Powerful. Affordable. Sign up now. http://mailplus.yahoo.com