mellotron samples

[McElhatton, Pat]

Something that works pretty well is to delta encode the PCM (first output   
sample in each channel is
actual 16-bit valueof first input sampel, all following output samples   
are the difference between the
corresponding input sample and the previous input sample) and then use   
ZIP on it.  The delta encoded
values have much more redundancy in the high byte than the original   
source.  Of course, there is no
file format that I know that directly supports this.

  Pat McElhatton

 ----------
From:  KA4HJH

>  ZIPing and then unZIPing wont affect the data in *ANYWAY* at all..

For all practical purposes you're right. You ususally get about 1 to 3 %.
Of course, even that's significant if it's a really big file.

>Seems to me that the tag of the wavefile (info txt file?) is pretty big.
>How about trying it with *.aiff and other files?

The problem is that significantly compressing any kind of PCM audio (or
even worse video) without losing anything is very difficult. Even a   
simple
sound is too informationally complex, and doesn't yield much to Huffman   
and
Lempel-Ziv-
Welch. The repetitions and patterns are all wrong. About the only way to   
do
it is to analyse the sound in some way, then re-construct it. Of course,
that'll probably lose something along the way...


Terry Bowman, KA4HJH
"The Mac Doctor"