[vorbis] Low bitrate encoding

[Monty]

On Tue, Jan 09, 2001 at 10:19:54AM -0500, Gregory Maxwell wrote:
> On Tue, Jan 09, 2001 at 10:12:07AM -0500, Gregory Maxwell wrote:
> > On Tue, Jan 09, 2001 at 03:41:58PM +0100, Dr.Joerg Bergmann wrote:
> > > > The other detail is what _analogue_ averaging (AKA low pass filtering) is
> > > > applied. Generally (but not always) frequencies that are above the
> > > > Nyquist frequency are low pass filtered.
> > > 
> > > Thats OK. But sampling itself adds some kind of low pass, an additional
> > > analogue low pass will only reduce signal/noise ratio. An analogue
> > > low pass may be (after fourier transformation) described by some
> > > weighted averaging (weighting function is zero for negative times due to
> > > signal theory, non-zero for positive values, in general none-zero for all
> > > future times!)
> > 
> > No. If you sample a signal with signifant components above the nyquest
> > components these cofeeicents will be aliased and reflected back into the
> > bandpass of your sampled data. Not cool.
> > 
> > To get good quality downsampling (or A/Ding) you must filter first. This is
> > non-trivial especially when the rates in noninteger.
> 
> To get a mental handle on this: Imagine sampling a 15KHz signal at 8KHz.
> Would you get all zeros?

In a perfect world, not limited by finite rolloff characteristics of
lowpass filters, If you don't get all zeroes, you're doing it wrong.

Really, folks, several of you are debating the most fundamental of
signal processing concepts.  Go get an undergraduate level book on
signal processing (eg, _Understanding Signal Processing_ or _Signals
and Systems_) and read it.  This is all covered first month.

Monty

--- >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-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.