| T O P I C R E V I E W |
| Paul Shinn(Lossless Compression Guy) |
Posted - 12/30/2003 : 09:33:16 AM
I was wondering when will we get Lossless audio formats?
flac.sourceforge.net
www.monkeysaudio.com
both of these are Open Source Software!!!!!!!
the R*O Ka*MA 20GB uses FLAC.....
what about us?
|
| 20 L A T E S T R E P L I E S (Newest First) |
| chad(at)gambit.net |
Posted - 11/18/2005 : 10:27:36 AM
It's worse than just skips.
It will jump back and forth inside the song by 2-20 second jumps. And it's in the source file, not a playback issue with the Neuros.
I've tried purging and re-installing dbPowerAmp, and that doesn't fix it, but OggDrop works fine on the same system. |
| Don |
Posted - 11/15/2005 : 06:07:57 AM
quote:
Originally posted by chad(at)gambit.net
dbPowerAMP was inserting odd audio artifacts into the OggVorbis files I was making off of FLAC to put on the neuros (Skipping, hiccoughing, pops, snaps, the whole 9 yards.)
I've never heard of any such problem with dbpoweramp on audio message boards like hydrogenaudio. Are the skips there playing on the PC, or just on the Neuros?
My best guess is that you have dbpoweramp and oggdrop set up to use different ogg encoders. It has been reported here that the aoTuV encoder plays better on Neuros (higher Q value before skips appear) than the basic one from vorbis.com. The issue is some resource constraint on the Neuros rather than one or the other encoder being defective.
Edit: I have dbpoweramp set up to use the aoTuV encoder. I haven't been hearing any skips, but I don't generally don't encode for the Neuros at over about Q4.5.
-Don |
| chad(at)gambit.net |
Posted - 11/07/2005 : 11:22:33 AM
Okay, while I realize that Josh is working hard on getting actual FLAC support in the firmware, an idea came up.
Why don't people implement a call to OggDrop into the sync managers?
dbPowerAMP was inserting odd audio artifacts into the OggVorbis files I was making off of FLAC to put on the neuros (Skipping, hiccoughing, pops, snaps, the whole 9 yards.) so I started using OggDrop today to do the conversion and it seems to run fairly well, and oggdropXPd even has FLAClib built into it for the input files....
|
| Entwined |
Posted - 04/04/2005 : 11:39:24 PM
yes, i meant Wavpack in lossless mode.
Neuros Wishlist: Gapless Playback + Wavpack Support + Replaygain = Bliss. |
| Don |
Posted - 04/04/2005 : 11:49:56 AM
quote:
Originally posted by Entwined
FLAC in lossless mode seems to be on par with WAVPACK.
What the heck does that mean? Isn't FLAC *ALWAYS* lossless?
-Don |
| Supacon |
Posted - 03/31/2005 : 3:50:33 PM
quote:
Originally posted by Entwined
WAVPACK -F: 2:10/24.67x | 367MB/68.1% | 0:50/64.14x
Still falls behind in filesize and decoding time...
Well way to go Josh, in that case :)
I notice that the encode time for WavPack -f is just slightly faster than FLAC -5 though. |
| Entwined |
Posted - 03/31/2005 : 3:45:01 PM
Encode File Size (WVUNPACK -M -V)
---------------------------------------------
WAVPACK -F: 2:10/24.67x | 367MB/68.1% | 0:50/64.14x
Still falls behind in filesize and decoding time...
Neuros Wishlist: Gapless Playback + Wavpack Support + Replaygain = Bliss. |
| Supacon |
Posted - 03/31/2005 : 12:01:56 PM
That's what I'd expect. FLAC is definitely more optimized for decoding. I'm noticing that you haven't been testing WavPack's -f mode, however. That could possibly be faster than FLAC. |
| Entwined |
Posted - 03/31/2005 : 01:29:24 AM
as suggested, i ran the files using "FLAC -t" and then, for fun, ran "WVUNPACK -m -v", -m checking the md5 checksum and -v verifying the file.
decoding results:
FLAC Q5: 0:47/68.23x
FLAC Q8: 0:41/78.22x
WAVPACK: 0:59/54.36x
WAVPACK -H: 1:17/41.65x
of course, i do not know if using those Wavpack arguments do the exact same test as the FLAC one, but FLAC is the surefire winner there...
Neuros Wishlist: Gapless Playback + Wavpack Support + Replaygain = Bliss. |
| Entwined |
Posted - 03/31/2005 : 01:17:59 AM
wouldn't
wvunpack -m -v
do the same thing as
flac -t
Neuros Wishlist: Gapless Playback + Wavpack Support + Replaygain = Bliss. |
| Supacon |
Posted - 03/31/2005 : 01:15:33 AM
Hmm... maybe I should bug David to add something like that to 4.2. Is it a complicated thing to add in, Josh? I'm not sure exactly how it works on a technical level. This kind of stuff is probably better suited to a thread at HydrogenAudio.org though. |
| Josh Coalson |
Posted - 03/31/2005 : 12:18:27 AM
quote:
Originally posted by Entwined
hmmm, well, im running the official binaries, so i dont know about any 3dnow or SSE optimizations, but if diskwriter indeed takes most of the time writing the files, it might have fast times because im running on a RAID 0 array.
probably not. take the example
prog A: runs in 10s
prog B: runs in 8s
=> prog B is 20% slower
subtract 6 seconds i/o time that will not be in a player
prog A: 4s
prog B: 2s
prog B is now twice as slow.
you will need a more accurate test like I described to get good a performance estimate.
- Josh Coalson - FLAC developer - |
| Entwined |
Posted - 03/30/2005 : 11:30:06 PM
hmmm, well, im running the official binaries, so i dont know about any 3dnow or SSE optimizations, but if diskwriter indeed takes most of the time writing the files, it might have fast times because im running on a RAID 0 array.
Neuros Wishlist: Gapless Playback + Wavpack Support + Replaygain = Bliss. |
| Supacon |
Posted - 03/30/2005 : 11:02:57 PM
I believe that there are some speed advantages when decoding, using the latest beta. WavPack 4.2 will be released very soon, however. There are a handful of new features, including built-in ape tagging abilities, and some advanced options.
The results you achieved especially seem good considering that you are using 4.1. I'm wondering if you happened to have a rare case where for whatever reason, WavPack performs exceptionally better and maybe FLAC isn't at the top of its game.
I'm going to wait a couple of weeks, and hopefully I'll have a nice, comprehensive test of my three favorite lossless encoders - FLAC, WavPack, and APE (which seems to be fading fast, but still has good compression). |
| Entwined |
Posted - 03/30/2005 : 10:57:47 PM
I used Wavpack 4.1... are there any proported advantages in using the beta?
Neuros Wishlist: Gapless Playback + Wavpack Support + Replaygain = Bliss. |
| Supacon |
Posted - 03/30/2005 : 7:34:07 PM
FLAC Q5 | 2:26/21.97x | 361MB/67% | 1:17/41.65x |
FLAC Q8 | 12:32/4.26x | 359MB/66.6% | 1:07/47.89x |
WAVPACK | 2:08/25.05x | 359MB/66.6% | 1:10/45.81x |
WAVPACK -h| 2:49/18.98x | 352MB/65.3% | 1:43/31.14x |
Wow. Those times for Wavpack are really impressive! I didn't think that it would be even close to the decode speed of FLAC, especially not at the same compression ratio. Of course, this is only one album. A better test would involve a few types of music.
Entwined, what version of WavPack are you using for this? 4.1, or one of the 4.2 betas? |
| Josh Coalson |
Posted - 03/30/2005 : 4:50:06 PM
diskwriter timings may not be indicative because so much of the time is spent writing to disk.
flac.exe has a test mode which will be a little more accurate, but it also does MD5 checks which are not used in a player. and depending on where you got it, it might have 3dnow optimizations. and it's built for x86 which is very different that the neuros cpu.
not sure if wvunpack.exe has a test mode.
- Josh Coalson - FLAC developer - |
| Entwined |
Posted - 03/30/2005 : 4:40:27 PM
I ran a quick test to see how fast and efficient the modes were, to hopefully clear up some misconceptions:
Computer: AMD Athlon 2000+ (old, I know)
Software: Foobar2000 - Diskwriter
Album: Radiohead, OK Computer, (53:27, 539MB)
Results:
Codec Encoding Time File Size Decoding Time
------------------------------------------------------------------
FLAC Q5 | 2:26/21.97x | 361MB/67% | 1:17/41.65x |
FLAC Q8 | 12:32/4.26x | 359MB/66.6% | 1:07/47.89x |
WAVPACK | 2:08/25.05x | 359MB/66.6% | 1:10/45.81x |
WAVPACK -h| 2:49/18.98x | 352MB/65.3% | 1:43/31.14x |
-------------------------------------------------------------------
This is what i concluded as the best total efficiency to the least:
Wavpack Normal > Flac Q5 > Wavpack High > Flac Q8
Interestingly enough, it looks like Flac Q8 would be fastest to decode.
FLAC in lossless mode seems to be on par with WAVPACK.
Id get ecstatic to see any of these codecs be implemented in the Neuros.
Neuros Wishlist: Gapless Playback + Wavpack Support + Replaygain = Bliss. |
| Supacon |
Posted - 03/30/2005 : 1:38:21 PM
You've got a good point, Josh. FLAC is going to be easier to support. Bryant doesn't yet know if portables will have a problem with WavPack -h mode, because it does take some amound of processing power, and it is possible that some portables might not yet be sufficient to handle it. If that is the case, one would have to re-encode all their WavPack files again, in a faster decode mode (i.e. wavpack -fx4).
I still think that FLAC is the best choice for a first lossless codec, is it's "safer" from the perspective of being easy to decode. Nonetheless, it would still be nice to see what WavPack can do. |
| Josh Coalson |
Posted - 03/30/2005 : 1:33:42 PM
this flexibility in decode complexity can be a disadvantage outside the PC. if some device "supports" wavpack but cannot handle all the modes because some require too many mips, and this is not clear to users and the decoder, there are going to be customer support problems.
this is why with FLAC I put almost all the complexitity and variability in the encoder.
Josh
- Josh Coalson - FLAC developer - |
