Markus wrote:

> According to my favourite hex editor, ... first samples should be

> 698.0      1628.0
> 2096.01  -4322.74
> ...

Thanks, I see what's happening.  It's the SoX utility which I use
to convert the .wav to a data stream.  The stupid program assumes
that audio samples, even when using IEEE float formats, are
restricted to amplitude +/- 1.0   We are seeing the severely
clipped output of a stupid program.


--
Paul Nicholson
--