Re: Homebrew STG card

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

stratton-at-mdc.net wrote:

> > Another possibility is to encode all the logic, both ISA interface,
> > digital I/O, and the encoder counters into a programmable gate
> > array device.  Then, the entire board would consist of one big
> > chip, like the Xilinx XC95160 and one or two A/D and D/A
> > chips.
>
> He's right about that... plus with canned PCI interface libraries
> (using these chips is like writing C code) interfacing to that more
> popular bus would no longer be unrealistic.

I'm still looking for a cheap PCI library.  One that I know of is $9000.

> Even with ISA, some programmable logic saves all the address decoders
> and the like.  It would be wonderful to fold in those pricey encoder
> chips.  The only thing I wonder about leaving discrete is an 8255 I/O
> - they are cheap enough that it might be less expensive to keep that
> external and generate a chip select for it and thus save 24 pins on
> the gate array.  Of course toss more than one or two chips on the
> board and it becomes necessary to buffer the bus - but maybe we
> already have to essentially do that for the DAC's.

Well, these XC95xx devices are SO cheap, and have SO MANY pins,
that it is no biggie.  The slower devices are under $10.00 each in small
quantity, and many have well over 100 pins!

> One big decision to make is if the LS7266 (encoder chip) register
> structure should be copied, or replaced with something new.  I found
> it a rather complicated programming model, but it worked fine once I
> made sense of it.

Well, much of this craziness is to save I/O pins.  It is much less of a
problem if a bunch of encoder counters share the same chip.
If combined with a PCI interface, then all the craziness to read
a 24-bit value without any possibility of the value changing before
you have read all the bytes goes away.

Jon