This is a fair question. We went back and forth on this
issue a while ago. Three aspects of the question are
(1) Implementors want to use cheap fixed-point DSPs;
(2) Musicians want to specify instruments in floating-point;
(3) High precision arithmetic is more appropriate, since we
don't know the sensitivity of operations musicians will want.
(2) could probably be worked around if it was the only concern,
but (1) and (3) are pretty much at odds with each other.
We decided, since SA is a technology that's still not likely to
be deployed for 18 months or more, and the trend is certainly
to floating-point DSPs these days, to specify the SAOL semantics
in floating-point. This is only true of the SA part; other codecs
(T/F codec for one) are specified in fixed point.
Note that, like other aspects of the standard, this just means that
you have to get the same answer as if you did the calculation on
a 32-point floating point system. There's no requirement to use
any particular kind of processor.
When you say "our DSP" can you clarify what you're referring to?
When I see Philips today I think Trimedia, which is a floating-point
processor, isn't it?
Best,
-- Eric
This archive was generated by hypermail 2b29 : Wed May 10 2000 - 12:14:10 EDT