[vorbis-dev] vorbis-utils features ;)

[Kevin Marks]

At 7:11 pm +1100 16/3/01, Michael Smith wrote:
>  >Still shouldn't be needing 200 kbps to code this though; if it's below
>>the threshold of hearing it shouldn't matter whether it's random or
>>not; Vorbis should be throwing it out! Or is there a major mistake in
>>the psycoacoustic model?
>
>Vorbis models the threshold of hearing in a slightly different (and more
>accurate) way than many mp3 encoders. Here's a somewhat simplified 
>explanation:
>
>We don't know the playback level (someone might have their amp 
>turned up REALLY
>high), so a fixed threshold isn't possible. Instead, the threshold 
>is a constant
>level below whatever the current peak level is, more or less. So 
>almost-silence
>is encoded in a way not significantly different from full-level sound. The
>difference is that the almost-silence tends to be very close to noise, so it's
>hard to encode (uses a lot of bits). It's neccesary to do this _because_ we
>don't (and can't) know the playback level.

This is a good idea in abstract. In practice, you want to define a 
noise floor and truncate below that. The 90dB range of 16-bit audio 
is overkill in all real-world playback environments, and preserving 
noise in the low 4 bits is just going to make it impossible to hit 
low bitrate targets.

If you don't do this intelligently in the codec, you are going to end 
up having to do it in the encoder application as pre-filtering, which 
will leave you where Sorenson are for video - you can get great 
results but you need to spend a long time tweaking parameters to do 
it.

--- >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-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.