Well, I will certainly look at an XML representation of pvoc analysis
files with interest, but the main advantages for me of a binary format
are streamability (as for ordinary soundfiles), and flexibility in
defining multifarious pvoc formats (different analysis rates, decimation
factors, window sizes, etc). The binary format will also be smaller than
the text equivalent, by at least a factor of two - assuming that six
decimal places plus the point are used for each value, and one tab, and
one newline (two bytes on disk for DOS). If ad-lib white-space is used,
the file-size will grow accordingly.

At the moment, the Microsoft WAVE-EX format looks extremely attractive,
as it is by definition streamable, and allows new sub-formats to be
defined without the need to register them with Microsoft. One goal I
have is a 'multi-channel' pvoc format (eg for stereo streams and above),
and chaining of frequency-domain processes in real-time as in
DirectShow.

The analogy of editing pvoc files in the way one can edit a single
time-domain sample does not quite hold, as in a pvoc stream of
overlapping frames, several frames will need to be edited together. The
potential for corruption of the format through manual editing and
cut+paste must be very high. The text format will certainly have value
for many purposes, but I cannot see it other than as supplementary to a
main binary format. 

BTW: Martin Puryear has just posted a new version of the Microsoft
WAVE-EX document, and I will send it to my website immediately. It looks
as if the main format is now fixed, as the document concludes by saying
that the format 'can be supported today'.  


Richard Dobson

Michael Gogins wrote:

> 
> I have used a hybrid XML format in which the tags and some data is text, but
> other data is length-prefixed binary. This, for example, forms the
>  element of the Csound structured data file (csd file), which I
> created. But I think the advantages of using binary format for the columns
> of a time/frequency analysis, as opposed to a text printout of the vector,
> are slight.
> 


-- 
Test your DAW with my Soundcard Attrition Page!
http://wkweb5.cableinet.co.uk/rwd