From: Mark T. <mt...@su...> - 2001-01-07 20:30:55
|
> > --__--__-- > > Message: 5 > Date: Sun, 07 Jan 2001 15:17:14 +0900 > From: Naoki Shibata <n-s...@ic...> > To: lame-dev ML <lam...@li...> > Subject: Re: [Lame-dev] Re: Lame-dev digest, Vol 1 #24 - 8 msgs > > > Hi, > > Mark> the infamous "switching artificts". MS has one type of noise, > Mark> LR has another type of noise. Each type may not be so bad, > Mark> but to switch back and forth every 26ms might create its own > Mark> low frequency artifacts which sound bad. > > But, the actual noise I hear from encoded track7.wav using old lame is > high frequency one. Do you really hear low frequency switching noise? > > > -- > Naoki Shibata e-mail: shi...@ge... > The problems with track7.wav were always too subtle for me :-) But I did run a test a long time ago where I switched between stereo and m/s every frame. The results were quite bad! Then for other reasons which I no longer remember, we put in some smoothing in the m/s switch so that LAME was less likely to encode a single m/s frame when both neighboring frames where stereo. This did improve some samples at the time (was that for track7.wav?) And it doesn't hurt most other samples where stereo or m/s frames come in long continous blocks. Mark |