Path: utzoo!utgpu!watserv1!watmath!att!bellcore!rutgers!ucsd!swrinde!zaphod.mps.ohio-state.edu!samsung!noose.ecn.purdue.edu!newton.physics.purdue.edu!murphy
From: murphy@newton.physics.purdue.edu (William J. Murphy)
Newsgroups: comp.lang.pascal
Subject: Accurate Delays with Turbo Pascal or C
Message-ID: <3891@newton.physics.purdue.edu>
Date: 11 Jul 90 00:31:01 GMT
Organization: Purdue Univ. Physics Dept, W.Lafayette, IN
Lines: 26


Recently, someone posted a unit for using one of the timers in the PC
architecture.  I compiled it and tried to modify it to create a custom
delay function similar to the one(s) supplied by Borland.  The timer
unit was able to give 10 microsecond interrupts instead of the usual 18.2Hz.
Any ways, I had little luck in getting a consistant delay.  It seems that
there is a rollover that takes place with the timervalue which I don't
yet understand.  Thus I would be ok for a few cycles then get stomped by
a short cycle.  

What I need is a very accurate delay that won't interrupt a DMA transfer
that will always be occuring simultaneously.  I am using the delay function
to turn on and off a DIO line which connects to a switch which controls
my stimulus.  The DMA transfer is the A/D conversion of the stimulus plus
a signal which is dynamically affected by the stimulus.

Does the author of the Timer unit have any suggestions about writing such
a delay function?  Borland's delay function is accurate to within +- 0.7
milliseconds, as well it has a periodic behaviour in its measured delay.
I would be happy if I could get +-0.05 milliseconds standard deviation.

Thanks,
-- 
Bill Murphy                                  | Is this a holdup? ...
murphy@newton.physics.purdue.edu             | No, it's a science experiment.
Amiga, it's not a religion, it's a lifestyle |     Emmett L. Brown  BTF III