Howdy,
On Thu, 30 Sep 1999, Steve Ashcraft wrote:
> The
> serious problem is the variation of the latency in the computer. This is a
> factor of both the hardware (which is probably not a problem) and the
> software that responds to the interrupt. I'd love to hear some opinions
> about how DOS handles interrupts and if there is a problem there that is
> unacceptable.
Sitting out here in the cheap seats (its been quite some time since I
played with DOS and any sorta realtime stuff), I can't see how what DOS
does or doesn't do would really matter, since you're gonna go in and
juggle around the interrupt mask and clock interrupt frequency anyway (or
at least I would).
In terms of the archetecture of a good system, I would think you could
combine what everyone's been saying by creating a single board computer
running a custom kernal (and DOS when its not?) with a few switches and
leds. That box would be resposible for keeping accurate time and
extremely basic functionality in terms of stand alone operation (one car
at a time, show the time, whatever). Then have that box talk to a user
interface computer running win32 stuff. Have that computer keep track of
the data, deal with multiple cars on course, etc. etc. etc.
Having said all that, I'm sure I could build a really kickass system for
$10k (single board computer & misc crap for $2k, laptop for $2k, misc
hardware, profit, etc.)... Which is probably 10x what people wanna pay for
timers :-)
I'd be interested to know what system architectures and designs folks out
there really end up using for timing systems.
Mark
|