Re: Homebrew STG card - CRAZY-8

[Jon Elson <jmelson-at-artsci.wustl.edu>]

"Arne Chr. Jorgensen" wrote:

> I don't really understand what you mean Jon,  -  but I guess you didn't
> understand me either.  Not sure if I do myself :)
>
> I got a little out of control,  and I probably "talked"  this to death.  What
> really started me off,  were the facts that this will work, and I have done it
> before.  I have a whole computer system made,  with rtkernel and os.  And I can
> "mount"  it more or less the way I explained unto a PC.  But to do the same
> thing with a whole PC, - sounded a bit too wild.  That is why I started out
> this with the "CRAZY" in the subject title.  But the more I thought of it, the
> more it dawned on me - this will work.

Well, the NIST people probably don't intend for people to need to raid
a dumpster to get a 286 machine to set this up with.  2 computers involved
in the system make it vastly more difficult to get everything working.
ISA is a dead (or at least declining) technology.  Since the main Linux
CPU has all the power needed, I see no reason to add another computer.

It is something to consider, but after considering it, I think there are
several better ways to go.  Actually, a bus adapter system hooked to the
parallel port, with current-loop servo amps driven by the CPU, and velocity
read from high resolution encoders, sounds like a real good solution.
The only problem is coming up with 4000 to 10,000 cycle/rev encoders
to put on 5 TPI leadscrews, so the encoder resolution will not cause
nasty hunting in the servos.  If you want system resolution to be .0001",
then you need an encoder resolution of .00001" or so, or it will have to
hunt a lot to hold the machine within .0002" tolerance.  The buzzing would
drive me nuts.

By the way, my servo amps could be retuned to turn them into current loop
amps (instead of velocity loop) by just leaving out a few parts, and
placing one jumper.  This would eliminate the need for a DC tachometer.

Jon