This is kind of a split from the Aac Vs Aps (http://www.hydrogenaudio.org/forums/index.php?act=ST&f=13&t=3946) thread.
JohnV has been kind enough to host for me several problem samples from the Smashing Pumpkins album, Mellon Collie and the Infinite Sadness.
All samples have annoying artifacts when encoded with AACEnc v2.15 using the -normal switch; the vocals are distorted and pan side-to-side.
Pumpkins_Bullet (http://static.hydrogenaudio.org/extra/Psytel/Pumpkins_Bullet.flac): Taken from about 3:05 into the song "Bullet With Butterfly Wings"
Pumpkins_Galapogos1 (http://static.hydrogenaudio.org/extra/Psytel/Pumpkins_Galapogos1.flac): Taken from about 0:45 into the song "Galapogos"
Pumpkins_Galapogos2 (http://static.hydrogenaudio.org/extra/Psytel/Pumpkins_Galapogos2.flac): Taken from about 1:10 into the song "Galapogos"
Pumpkins_Ode (http://static.hydrogenaudio.org/extra/Psytel/Pumpkins_Ode.flac): Taken from about 0:06 into the song "An Ode To No One"
Using the -safems switch improves the samples considerably, but problems can still be heard. -extreme handles the clips better, but I think I may still be able to ABX them.
Edit: Only -mst 0 (completely disabling mid/side stereo) fixes the samples, as of AACEnc v2.15.
JohnV has been kind enough to host for me several problem samples from the Smashing Pumpkins album, Mellon Collie and the Infinite Sadness.
These are hosted at HA's server, so it's pretty permanent place.
Yes, I am aware on this artifact (posted long time ago) - right now I am preocupied with something on higher priority, but I will try eliminating it until official next release.
Artifact could appear on tracks with "pan-potted" human voice and very similar masking thresholds on L and R channels - it might be toggling of M/S and L/R too often for a particular sfb, or very relaxed M/S switching criteria.
Yes, I am aware on this artifact (posted long time ago) - right now I am preocupied with something on higher priority, but I will try eliminating it until official next release.
Artifact could appear on tracks with "pan-potted" human voice and very similar masking thresholds on L and R channels - it might be toggling of M/S and L/R too often for a particular sfb, or very relaxed M/S switching criteria.
Yes, I've been harassing you about this artifact for over half a year
I'm very interested in your "higher priority" work, although from what I've heard discussed publicly, it's top-secret or something. So don't let me distract you from that.
In the meantime, I will use -mst 0 to avoid these artifacts.
I'm very interested in your "higher priority" work, although from what I've heard discussed publicly, it's top-secret or something. So don't let me distract you from that.
MPEG4 video and audio support in Nero.
It was secret some months ago, but nowadays it's public.
Aaaw... that sounds boring compared to AAC+SBR if you ask me Although it would benefit the AAC format I should think... soon I'm going to be making some digital videos (i'm religiously recording my faviourite TV series each week!)... so things to help combining divx 5 video with mp4/aac audio, and making it playable will be handy
We need to clone Ivan!
Oh, yes, "I think" they are developing AAC+ at Nero too. (Hope this news were spreadable. If they weren't, now they are. )
Gotta love those Nero people.. They know on what to invest their money and work .
Just tested these samples with the new Nero plugin (v5.5.10.0), downloaded 12/18/02 or so. They still fail. I had some trouble getting the files to decode properly, so I couldn't do an ABX, but it's really not necessary with these samples.
Bullet still fails on Normal (although Nero output isn't as bad as Psytel, it's still pretty ghastly). Ode still fails on Extreme. Those are the only two samples/encodings I tested so far, so even Audiophile might not be transparent.
Settings used were those Ivan recommended in the 5.5.10.0 news announcement:
Ok, rules for highest quality encoding:
1. Go to "preferences / plug-in look up / defConvertor" and set resampling to SINC , quality 100 for other purposes (combo box that allows selection between "playback" and "other purposes" - this will enable highest quality resampling available in NERO
2. Switch to "high" encoding quality for MPEG-4 AAC plug-in Default is "medium" which is also good quality, and 30-40% faster - but with small degradations
Other options: Variable Bitrate (either Normal :: High or Extreme :: High), LC (Low Complexity) profile, and Export ISO-13818-7 AAC Track.
Decoding by the latest RareWares FAAD/MP4 plugins.
At least I found the reason for the problem - when L and R channels are almost the same, side channel is getting less bits than necessary, and in some frequency areas it just erases the entire signal - after reconstruction signal could have "thrill" in stereo image over short period of time.
This usually does not happen too often, but in some cases it could.
Side note: good point, I'm wondering about Lame in a similar case...
At least I found the reason for the problem - when L and R channels are almost the same, side channel is getting less bits than necessary, and in some frequency areas it just erases the entire signal - after reconstruction signal could have "thrill" in stereo image over short period of time.
This usually does not happen too often, but in some cases it could.
I'm glad you found the cause!
Can you estimate when this problem will be fixed?
UPDATE(I didn't want to bump the thread, but anyone browsing the archives should see this)
For at least a couple months now, the AAC encoder in Nero has had no problem with Bullet, as far as I can tell, so the error is fixed.