[net.micro.atari16] The 32081 Floating Point Chip on the ST

braner@batcomputer.TN.CORNELL.EDU (braner) (09/27/86)

[]

After a period of planning and experimentation, we have the National
Semiconductors 32081 floating-point math coprocessor talking to the ST!

("We", since I have collaborated on this with Andreas Andreou at Hopkins.)

Interested parties please e-mail to <braner@amvax.tn.cornell.edu.ARPA>,
or at the address above, or call 607-272-3487, or write me at the address
below.  If the response is good, we might post some details.  (If it is
VERY good, we'll keep it a trade secret :-).

We have the 32081 connected to the cartridge port.  To do that, the port
needed some modifications.  We suggest this mod as a "standard", since it
is useful for many other purposes and designed to be easy to do and to
keep the port usable for ROM cartridges and such too.  Here is the mod:

Cut pins 1, 39 and 40 behind the connector. (No trace-cutting needed.)
Connect the CPU R/!W line to pin 1.
Connect the CPU clock to pin 40.
Build an address decoder that senses the address $3xxxxx (x=don't care)
and is strobed by the CPU !AS signal.  The best way to build this decoder
is with one chip, a 74HCT138.  The "HCT" stands for high-speed, LSTTL-
compatible CMOS.  Connect the decoded address strobe line to pin 39.

Since these pins are just redundant Vcc and ground lines, standard
cartridge port devices can still be used, provided these lines are cut
inside each device (which will then still work on standard ST's).

BTW, if Atari didn't bother to write-protect the ROM area, the decoder
wouldn't be neccessary.  This scheme will work with 2-Meg ST's too, but
not with 4-Meg models, since $3xxxxx is within the fourth Meg.  Addresses
above $3FFFFF yield a bus error!

- Moshe Braner
  Corson Hall, Cornell University, Ithaca NY 14853