Re: FPGA for PCI based servo control board

[John Kasunich <jmkasunich-at-worldnet.att.net>]

Robin Szemeti wrote:

>I'm just confused by the whole thread really.
>
>What it seems to boil down to is this:
>
>EMC works fine, but for high pulse rate applications its give the CPU a
>hard time, so offloading that to an external device would be a good plan.

Yes.

>Jon has already done this with his external stepper driver boards and
>found the market to be in the region of 5 a year ho hum.

Sad but true.  However it is fixed at four pulse outputs, four encoders,
and a fixed set of digital I/O.  I'd buy it myself, but I want more 
flexibility.

>The thrust of the new project seems to be to split Jon's board in half,
>move the FPGA inside the PC and leave the IO external to the PC

Overly simplified, but yes.  IMHO, the key is that the I/O becomes
completely open and flexible.  Need 5 encoders, 1 analog output,
3 step/dir outputs, 8 digital outputs, and 16 digital inputs?
No problem.  You pay for only what you use.  The FPGA needs
to contain the full set of stuff whether it is used or not, but FPGA
gates are cheap and getting cheaper.  The sad fact is that the
terminal blocks, relays, optos, etc., needed to connect the signals
to the outside world cost more than the logic gates needed to
generate the signals.  Even the PCB space occupied by those
parts is expensive, especially in small quantities.

>  ... this avoids the 'problem' of using the printer port to drive the 
> external
>card, but adds interfacing to the PCI bus,

Actually I've advocated keeping the parallel port interface, but the
design could support PCI, ISA, parallel, and ethernet interfaces
between the PC and the FPGA.  The first two would locate the
FPGA inside the PC, the other two would locate the FPGA outside.

>and then providing a custom interface to the IO board.

The interface to the I/O board is plain 'ol digital signals on wires.
The FPGA and I/O board would always be close together, no more
than three feet.

>Ethernet has been suggested

Not between the FPGA and I/O.  Between the FGPA and PC, yes.

>Perhaps using another PC at the other end to handle the comms.

Not by me.  I don't think Craig or John Sheahan suggested that
either.

>Now, call me a silly old goat, but it seems to be adding a bunch of extra
>hardware, to solve a problem thats already been solved in a slightly
>different way.  I dont see its any different to Jon's existing board, except
>its going to be in 2 boards, with the added 'fun' of a custom interface
>between the two.

Not as much added hardware as you seem to think.  The EPP version
will have a component count very similar to Jon's, except that there
will be two connectors and a cable to connect the FPGA board to the
I/O board.  And some of the parts won't be installed unless you need
them.  Jon has to install everything, because he doesn't know what
you need.

Jon's board does four axes of step/dir for $200.  Additional axes cost
another $200.  If you need analog, Jon's solution is about $650, and
the STG board is $800+.  I expect that this project will cost about
$250 for a 4 axis EPP version.  It can provide any mix of analog and
digital.  Additional axes cost maybe $20 each, for the I/O parts.
ISA version might cost a little less, PCI and ethernet will cost a
little more.

>   I'm not saying that it won't be a fun project, (heck, I'll
>most likely build one or two if it does come off)

That's the spirit.  If it ain't fun, why do it?

>I'll be watching with interest though ... a PCI pulse gen card with external
>IO does sound rather apealing for some odd reason ...

Openness and flexibility are what appeals to me.

John Kasunich