Darin Morrison
2001-Jun-15 11:17 UTC
[vorbis] Difficult sample for vorbis w/ audible artifacts
I came across an interesting test sample the other day while trying to compress some of my music and trying out different encoders. You can grab the sample in question from here: http://www.animus-facticius.org drone_clip.zip is a 9 second clip of the track where the artifact is most apparent, drone_short.pac is a 1 minute clip of track. The sample is from the song "Drone" from "Autechre - Peel Session". The longer clip is compressed with LPAC which can be found here: http://www-ft.ee.tu-berlin.de/~liebchen/lpac.html Basically, what happens is when encoding this file at lower bitrates with vorbis beta 4 there is some very bad pre-echo on the sharp attacks. I have been told that pre-echo handling is still being worked on though so maybe there is not much that can be done yet. I was told something about wavelets being implemented in the future, possibly after 1.0, and that this might make a difference? The majority of the pre-echo does go away as the bitrate increases but its still not perfect. The other artifact doesn't go away no matter what the bitrate is (I tried -b 350). It sounds like low frequency noise or vibration, or maybe some sort of "squeak" on the bass pulses. I have also tested this sample with both mp3 and AAC and they both did quite poorly, the only codec that seems to be able to handle this perfectly is mp+/mpc. Vorbis is in my opinion the second best though on this sample though, by quite a bit. I was quite surprised to see how well it actually did after most others failed. Anyway, maybe one of the developers can take a look at this sample and see if there is anyway to improve vorbis to handle this better or something like that. I have heard of some other low frequency artifacts on heavy bass notes with vorbis in the past but I'm not sure if they were supposed to have been addressed yet or not, and I'm not positive this case is caused by the same thing either. There has already been some discussion about this clip on the r3mix.net forum for those interested, though it has mainly been centered around mp3. There is some mention of testing it with vorbis in there however: http://66.96.216.160/cgi-bin/YaBB.pl?board=c&action=display&num=992254752 Dibrom --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
On Fri, Jun 15, 2001 at 11:17:56AM -0700, Darin Morrison wrote:> I came across an interesting test sample the other day while trying to > compress some of my music and trying out different encoders. You can grab > the sample in question from here: > > http://www.animus-facticius.orgClip grabbed, but not yet listened to. I'll write more a little later once I get to. Monty --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
I grabbed the WAV, and encoded it with the latest Oggdrop at 350kbps. I dragged 5 instances of the WAV file, and 5 instances of the Ogg file into a Winamp playlist, and randomized it. I don't hear any artifacts in the Ogg file.. Next, I encoded the WAV file to a 128kbit VBR MP3 with RazorLame, and into a 128kbit Ogg file with Oggdrop 1.0b4. I put three instances of each into a Winamp playlist, and 3 instances of the WAV file. Hit randomize, turned up the volume, closed my eyes, hit play. I really don't hear any artifacts. Admittedly, I'm using Klipsch Promedia 4.1s.. Not reference quality stuff. I left my Sennheisers at work today.. I get the impression from your message that it's not supposed to be that subtle. If it would help, I could make my mp3 and ogg test tracks available to you.. ----- Original Message ----- From: "Darin Morrison" <dibrom@moscow.com> To: <vorbis@xiph.org> Sent: Friday, June 15, 2001 2:17 PM Subject: [vorbis] Difficult sample for vorbis w/ audible artifacts> I came across an interesting test sample the other day while trying to > compress some of my music and trying out different encoders. You can grab > the sample in question from here: > > http://www.animus-facticius.org > > drone_clip.zip is a 9 second clip of the track where the artifact is most > apparent, drone_short.pac is a 1 minute clip of track. The sample is from > the song "Drone" from "Autechre - Peel Session". The longer clip is > compressed with LPAC which can be found here: > > http://www-ft.ee.tu-berlin.de/~liebchen/lpac.html > > Basically, what happens is when encoding this file at lower bitrates with > vorbis beta 4 there is some very bad pre-echo on the sharp attacks. Ihave> been told that pre-echo handling is still being worked on though so maybe > there is not much that can be done yet. I was told something aboutwavelets> being implemented in the future, possibly after 1.0, and that this might > make a difference? The majority of the pre-echo does go away as thebitrate> increases but its still not perfect. The other artifact doesn't go awayno> matter what the bitrate is (I tried -b 350). It sounds like low frequency > noise or vibration, or maybe some sort of "squeak" on the bass pulses. > > I have also tested this sample with both mp3 and AAC and they both didquite> poorly, the only codec that seems to be able to handle this perfectly is > mp+/mpc. Vorbis is in my opinion the second best though on this sample > though, by quite a bit. I was quite surprised to see how well it actually > did after most others failed. Anyway, maybe one of the developers cantake> a look at this sample and see if there is anyway to improve vorbis tohandle> this better or something like that. I have heard of some other low > frequency artifacts on heavy bass notes with vorbis in the past but I'mnot> sure if they were supposed to have been addressed yet or not, and I'm not > positive this case is caused by the same thing either. > > There has already been some discussion about this clip on the r3mix.net > forum for those interested, though it has mainly been centered around mp3. > There is some mention of testing it with vorbis in there however: > > http://66.96.216.160/cgi-bin/YaBB.pl?board=c&action=display&num=992254752 > > Dibrom > > > > --- >8 ---- > List archives: http://www.xiph.org/archives/ > Ogg project homepage: http://www.xiph.org/ogg/ > To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' > containing only the word 'unsubscribe' in the body. No subject is needed. > Unsubscribe messages sent to the list will be ignored/filtered. >--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.