Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site mit-eddie.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!harvard!think!mit-eddie!jaf
From: jaf@mit-eddie.UUCP (Joe Frisbie)
Newsgroups: net.arch
Subject: Floating point DAC's
Message-ID: <4811@mit-eddie.UUCP>
Date: Sun, 28-Jul-85 21:22:45 EDT
Article-I.D.: mit-eddi.4811
Posted: Sun Jul 28 21:22:45 1985
Date-Received: Tue, 30-Jul-85 06:12:56 EDT
Distribution: net.arch
Organization: MIT, Cambridge, MA
Lines: 12


|	Floating point DAC's would be very useful.  One of the biggest
pains in DSP with integers is precision.  Floats would let you keep
the same precision independent of signal amplitude.  I don't even think
they're that hard to make.  Use a regular DAC for the mantissa and 
a log DAC for the exponent and run the outputs through an analog multiplier.
	
	16 bits isn't enough if you want to do any processing at all.
All serial digital transmission standards call for passing 24 bits around
(except for CD's maybe).  16 bits is only ~90db which you can hear.

Joe