To: menelli@tellabs.com
Subject: PG controller board
Cc: glove-list@karazm.math.uh.edu
Message-Id: <9110211649.AA12004@roi.ca41.csd.mot.com>
Date: 21 Oct 91 16:49:43 PDT (Mon)
From: Lance Norskog <lance@roi.ca41.csd.mot.com>

This is excellent!  The software interface provided in high-res
mode is pretty weak.  For serious work I would much prefer your
outboard controller.  I have some software that does continuous
gesture recognition for mice, based on characterizing two single
input streams.  It mentions something about doing gloves via
multiple streams.  I can't use it with the glove because it
only does 2 bits.  (The software is on emsworth.andrew.cmu.edu
in /gestures.  It's by Dean Rubine, see the SIGGRAPH '91
proceedings.)

A point: the serial port approach does involve delays.  A parallel
port version would cut that down even farther.  Does the Motorola
evaluation board include enough parallel lines to make this
possible?  A parallel version would only involve one interrupt
per sample, and short polling loops.  

Another point: I would want a mode where the controller
board did as little interpretation as possible.  Just feed 
the raw counters in and let the big CPU with floating point 
handle X,Y,Z,rot issues.

A third point: does the Motorola chip include timers? A
global timestamp, based on the start of the sample cycle
in the board, would be very useful for doing synchronization
work.  There are algorithms for dealing with irregularly
spaced input streams; a prediction algorithm would need
accurate timestamps to operate properly.

Lance Norskog