Help - Search - Members - Calendar
Full Version: So what happens between 96 and 112
Hydrogenaudio Forums > Lossy Audio Compression > MP3 > MP3 - General
markalot
I'm using LAME 3.96, I built it myself with MSVC.

I'm not sure exactly how to phrase this question. I don't think it's a bug but I'm hoping someone can explain to me exactly what is happening to make such a huge difference in quality.

background:

I am using oddcast and lame_enc.dll to transcode mostly 192kbps mp3's into a stream I can listen too at my office. I've been experimenting with a lot of different methods to get the clearest sound quality at the lowest bitrate for playback on crappy speakers. I've modified lame_enc.dll to force a low pass filter at 13k.

lame_set_lowpassfreq(gfp, 13000);

I chose 13k simply because it still sounds good on these speakers. So after experimenting with the low pass at a high bitrate I then began to lower the bitrate until I got to the point where the distortion was too much to handle. I got all the way down to 112 and everything still sounded good, then I tried 96K and mercy! The high end was completly swishy, the distortion level was incredibly bad compared to 112.

My question is why is the difference so great? I expected the sound to get progressively worse, not jump from good to terrible. Is this my ears fooling me, or is there some technical explanation for the difference? Could I have introduced or exposed a bug by forcing a lowpass filter?

Thanks!
dreamliner77
LAME isn't tuned for bitrates <128.

at lower bitrates you should look at FHG.
gusnz
Why MP3 streaming? If you're allowed to run your own software at the office, or can sneak a copy of Winamp/fb2K/your favourite player onto a computer there, try Ogg Vorbis streaming?

Without trying to start a flamewar or institute a TOS violation, it's generally agreed that MP3 doesn't have brilliant quality at low bitrates (<128k), but if you can stream Vorbis or AAC or another more modern codec, you may get better results (I've found that 44khz/16bit/stereo Vorbis is listenable on cheap speakers at or below the 96kbps mark, and presume a good AAC codec is similar).

IceCast is a free and capable OGG streamer (if you're not already using it already with your OddCast plugin), or I guess you could just transcode (or preferably re-rip) your files onto a CD or HTTP/FTP share.
markalot
QUOTE
Why MP3 streaming? If you're allowed to run your own software at the office, or can sneak a copy of Winamp/fb2K/your favourite player onto a computer there, try Ogg Vorbis streaming?


Thats a very good question considering I have icecast installed and winamp at my office smile.gif My question stands though, just for curiosity sake. The current LAME is simply amazing at 128k IMO, at least compared to what it was just a year ago.
markalot
Holy explitive.

Streaming ogg now at 64kbps and it sounds nearly perfect!

Thanks for the suggestion, I was set in my ways I guess!
music_man_mpc
I think the lowpass filter of 13kHz might be too much for 96Kbit/s Mp3 . . . thats just a guess, though. Maybe you should try it with a lower lowpass . . . or use --(alt)preset CBR 96.
MugFunky
definitely --alt-preset CBR 96 will sound better (and the lowpass doesn't matter all that much below 16k).

with the alt-presets (now they're just presets) i didn't hear much difference between 96 and 112.

and you can change lowpass at the commandline (--lowpass 13 i think). no need to rebuild.

resampling from 44.1 to 32 helps matters A LOT.

the fabled sfb21 problem is still there even if there's no actual signal in this band.
Gabriel
QUOTE
definitely --alt-preset CBR 96 will sound better

I'd suggest preset 96 instead (abr, not cbr).

Does your oddcast plugin prevent resampling? By default, Lame should resample to 32kHz at 96kbps.
pcatiprodotnet
Agreed; I convert to 96abr, 32khz (lame3.97a3) for my portable flash mp3 player. It's indistinguishable from the source (to me) and very small file sizes (~2.5mb each).
Then I use mp3gain on them so I rarely need to fiddle with the player volume.
-pc
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2009 Invision Power Services, Inc.