Bit reduction question...

[spiritman]

So do you use dithering when going to MP3 or WMA file? I started with 24b 96k WAV file, down as far as I can stand it.
The final product is for an e-mail list. So size is everthing...


So what is the best way to smash a .wav file. I mean to get the smallest size and still sound OK?

[Bob L]

I would mix it down as 16 bit 44.1k wav... then use dbAudioware or any other mp3 converter set to 192 for very decent quality... set to 128 for smaller size... you be the judge on the quality. :)

Most mp3 converter routines are better designed to handle cd type files (44.1k - 16 bit)... which is why I suggest mixing it down in SS with whatever dither you want or not.... don't let the converter do anything except translate directly to mp3.

Bob L

[spiritman]

Thanks Bob.
Also I just D/L the free spell check. Thanks also for the link. Has that been there long? I just saw it tonight.


Anyone else got any input.....About the dither question?

[Bob L]

Try Pieter's native Dither plugin... also try mixing with each of the built-in dither algorithms... listen close... see if you can hear a difference and if you like it or not... only you can decide.

Bob L

[TotalSonic]

There was an interesting recent thread on Brad Blackwood's Mastering forum regarding this exact thing...

http://recforums.prosoundweb.com/index.php/t/9395/2475/?SQ=24cc768d7b016901d2ae125415da5458

Apparently James Moorer (who is involved with Sonic Solutions and a ton of other DSP stuff) wrote a paper advocating encoding from whatever the original bit depth is and there seems to be a number of advocates for doing this for those trying to eek out a tiny bit better sound - you just need an encoder that can work directly from the 24bit file.

Apparently even 5 years ago there were a couple mp3 decoders that can actually play back 24bits -
http://mp3decoders.mp3-tech.org/24bit.html - it'd be interesting to get an updated list somewhere

Does it make a difference??? I haven't evaluated this (it's rare that I get called for mp3 creation and usually use the most recent LAME codec running in SAW via the JMS LACE for it) - but the only way to find out is to encode the same source track a number of different ways and play it back in the various common players (i.e. iTunes, Winamp, etc.) and see what sounds better to you. Which method & encoder will work best might even vary depending on the track material too.

Anyway - if you indeed end up working with LAME inside of LACE (which is probably the most convenient way for SAWyers) -
if file size for the mp3 is an issue sometime VBR (variable bit rate) encoding can sound quite a bit better vs. the same size file using a fixed bit rate.

Best regards,
Steve Berson

[Ian Alexander]

...you just need an encoder that can work directly from the 24bit file.

Apparently even 5 years ago there were a couple mp3 decoders that can actually play back 24bits -
http://mp3decoders.mp3-tech.org/24bit.html - it'd be interesting to get an updated list somewhere

Does it make a difference??? I haven't evaluated this (it's rare that I get called for mp3 creation and usually use the most recent LAME codec running in SAW via the JMS LACE for it)...

JMS LACE is in use here almost daily, but be aware that it does not use bits beyond 16. From the LACE help file:
"Note that MP3s can only be generated from 16bit audio, but it is NOT necessary to use a 16bit F-Res patch. Any patch point will work, data beyond 16bits will simply be truncated."

[Carl G.]

JMS LACE is in use here almost daily, but be aware that it does not use bits beyond 16. From the LACE help file:
"Note that MP3s can only be generated from 16bit audio, but it is NOT necessary to use a 16bit F-Res patch. Any patch point will work, data beyond 16bits will simply be truncated."
However, IF it's premixed spots (Not VO) where there are NO low passages (because of some kind of music/effects) then very little audio level will be decerned below 16bit - therefore why bother dithering? Why not just truncate?
Right?
(I use Lace ALL the time without dither on 24/44.1.... seems to sound better than dithering to 16 first.... well.... least last I 'quickly checked' :) ) Hence, my sincere question above.

[Ian Alexander]

Hey Carl,

It's a good question. MP3 encoding eliminates "masked" sounds, right? It may be that any subtle sounds in those last 8 bits would go away in the data compression anyway. Of course, to determine if it sounds "better", you should do a double-blind, placebo-controlled study with at least a thousand patients in several test centers... wait, never mind. I guess I've been reading too many pharma industrials lately. :o

[Dave Labrecque]

I guess I've been reading too many pharma industrials lately. :o

Ian, you renaissance man. :p