Request for Real-time HW/SW Info

[Leonard J. Trejo]

We're planning to buy a microcomputer for conducting
psychophysiological experiments.  We now use a Masscomp 5500 running
RTU 4.0 ("real-time" UNIX) to control presentation of graphics stimuli
and to digitize and store physiological signals.  Problem is that RTU's
time is not "real" enough for some of the things we want to do.  For
example, we want to flash a visual stimulus on a graphics monitor 
for exactly one video field at 125 ms after the onset of data 
acquisition on an A/D converter and then record the response latency 
of a subject (button-press) to the picture with an accuracy of 1 
ms.  Everything must be synchronized to within 1 ms, e.g.


	|------------S---------------R---------------------|
        |            |               |                     |
Time    0            125             X                     1000
(ms)
	Start        Stimulus	     Response              End
        A/D          Onset           Time                  A/D
        Transfer     Time                                  Transfer
 

S should be 125 +/- 1ms and X should be measured to within +/- 1ms.

On the Masscomp we time S by putting a light sensor on the face of the
graphics CRT which sends a pulse into one channel of the A/D
converter.  Although it's a bit awkward, this can provide stimulus
timing accuracy limited only by A/D sampling rate, which can go up to 1
MHz/# channels.  We can time X similarly.  However, this only works
well for simple events, such as the whole screen turning on or off.  It
wouldn't work for the appearance of a character on the screen whose
coordinates vary from trial to trial.  Besides, we want to get rid of
the screen sensor and use a mouse for subject responses.

Our thinking now is to acquire a dedicated PC (e.g. an 80386 based AT
clone) to produce stimuli and measure responses.  The system should be
able to do the following:

	o Be able to send a 16-bit parallel word out which marks the onset 
	  time of a visual (graphic) stimulus (one per video field) 
	  with a latency of 1 ms or less.  This to be followed within
	  100 ms by two words which code the type of stimulus and its
	  screen coordinates.
	o Be able to respond to a mouse button press and send a parallel
	  word out in 1 ms or less which marks the time a mouse button is 
	  pressed (this is probably asking too much).  This to be followed 
	  within 100 ms by two more words which code which buttons were 
	  pressed and mouse coordinates when pressed.
	o Graphics capability of minimum 640 by 480 pixels with
	  independently addressable color lookup tables using 8-bit (
	  or higher) D/A converters (i.e. 16.7 M color palette).
	  Also capable of simple animation (e.g. video game quality).
	  Library of C graphics subroutines.  Sound-producing chip
	  would be a plus.
	o C compiler.  (UNIX development system would be a plus, but
	  is not required.)
	o Cost of hardware and O/S to be about $5K.

Most of this boils down to very tight coordination between graphics
and parallel digital I/O plus fast intterrupt responses.

We'll use this PC as a fancy stimulator which sends/receives event-related
synchronizing digital signals to the Masscomp.  These signals will be
monitored by a fast parallel I/O board controlled by the Masscomp
data acquistion/control processor, which will coordinate the digital
and analog I/O.   Thus, the Masscomp will serve as a fancy A/D 
converter and near-real-time signal processor.

Sorry to be so long-winded, but if you have suggestions please send
them to my by e-mail. 	If there's interest, I'll post a summary.
Thanks in advance.

				L. J. T.

============================================================================
ARPANET : trejo at nprdc.arpa 	UUCP:	ucsd!nprdc!trejo

Phone: (619) 553-7981		Postal Address:	Leonard J. Trejo, Ph. D.
       (AV)  553-7981				NPRDC 
						Code 52
						San Diego, CA 92152-6800