Re: PCI card

["alex" <telecomt-at-tht.net>]

If you don't like DOS - don't program in C, program in Assembler.
Very fast!
Alex
----- Original Message ----- 
From: John Griessen <john_g-at-cibolo.com>
To: Multiple recipients of list <emc-at-nist.gov>
Sent: Saturday, February 15, 2003 11:33 AM
Subject: Re: PCI card


> 
> Hi Mark,
> 
> I think industrial ethernet is a better way to think of how to get past
> obsolescence of parallel ports also.  Although it will be a few years at
> the minimum before there is any parallel port obsolescence problem, and
> then it will only be a price problem.  There is a nice VIA mainboard
> that is the smallest mini-ITX size that comes with parallel ports and is
> aimed at the embedded market.  Aiming at the embedded market means the
> serial and parallel ports will stay withthem longer than the mainstream
> motherboards. A VIA EPIA mainbaord costs about $110 at Fry's and has all
> peripherals and one PCI slot, then $30 for memory and $40 for a
> mini-sized case and a $70 hard drive and you have a 600 MHz EMC system. 
> It's even fanless so you can have few dust accumulation problems.  
> Expensive industrial grade computer boards will be available with
> parallel ports for long time frames -- that's part of why some
> industrial boards are specified -- that the maker promises to keep
> delivering and not change them rapidly.  The VIA board isn't quite
> industrial, but close.
> 
> For offloading function from the PC to something closer to the motors,
> industrial ethernet will be good also.  There are even ways to have some
> links be over plastic fiber and compatible with copper twisted pair
> ethernet for the rest, all passing along the same data.   Plastic fiber
> is a bigger benefit than cheap available port cards, since it gives
> noise and voltage drop immunity in an environment where voltage drop and
> noise are real hazards.  Big motors will cause drop, especially when
> some metal cutting tool bites off more than it can chew.  controls to
> many motors are noisy, and someone might be plasma cutting nearby too.
> 
> PFGA's are not a hit with many folks on this list -- it's a rather large
> design task to use them -- lots of resistance compared to
> microcontroller programming.  And then there's resistance to external
> microcontrollers too, since they have their own time and effort hit
> compared to programming on a PC.  The place to study more, (I've been
> studying other stuff for my work and have not dug into it yet), is in
> the EMC documentation about the place in the code where the real-time
> part hands over and receives data from the non-real-time part.  It is
> done in a way that is planned so it can be done in a specific physical
> memory location outside the PC rather than just in an allocated memory
> range setup by software in the PC without you thinking about it.
> Those memory locations can be put at the other end of an industrial
> ethernet network pretty much by the ethernet chips alone, and then
> either hard logic (or FPGA) or a microcontroller could use the ethernet
> chip I/O registers to run motors and listen to sensors.
> 
> John Griessen 
> Austin Texas
> 
> On Sat, Feb 15, 2003 at 12:26:25AM -0500, Mark Pictor wrote:
> > looks to be the way to go, though it would still
> > require someone familiar with the VHDL language.
> > 
> > FPGA? core? VHDL? An FPGA (Field Program
> 
> On Sat, 2003-02-15 at 00:49, John Sheahan wrote:
> > If you are going to do the real-time offload - choosing
> > something with a lot less wires / longer range makes 
> > more sense to me.
> > 
> > The prior suggestions for ethernet seems sensible for this 
> > part. Put the real time external controller in a fpga or real-time
> > cpu or dsp (as suits the individual - I'd personally choose 
> > the fpga).  Keep the PC inside, not in the workshop.
> > 
> > john
>