FAX cards for 386 UNIX

[Rick Richardson]

In article <1788 at van-bc.wimsey.bc.ca> sl at van-bc.wimsey.bc.ca (Stuart Lynne) writes:
>In article <391 at icjapan.uucp> jimmy at denwa.info.com (Jim Gottlieb) writes:
>>In article <706 at camco.Celestial.COM> bill at camco.Celestial.COM (Bill Campbell) writes:
>>>The HAYES JT9600 fax/modem card is NOT COMPATIBLE WITH SCSI.
>>>Install the card and the system won't boot.
>>We have a Hayes JT9600 fax card and an Adaptec 154xB in the same system
>>and they both work fine.
>
>The board will respond incorrectly to 16 bit memory addresses. For example
>if the board is addressed at 0xcc000 it will sometimes respond to 0x1cc000
>or 0x2c000. 
>
>I have noticed that more recent mother board designs seem to work better. For
>example the 486 I have now seems to work 100% with any JT9600 I've tried. 

The Hayes card does not, by itself, respond incorrectly to addresses
above 1MB.  The card is an XT bus style card, which means that it
only gets 20 bits of address (not the full 24 bits of the AT bus).
To maintain downward compatibility when the AT bus was designed,
IBM changed the meaning of the MEMR and MEMW control signals on the
original XT connector.  The bus master is supposed to activate
MEMR or MEMW on the XT connector only if the address is below 1MB.

What is going on is that some motherboards are incompatible with
the Adaptec bus master SCSI controller.  Somehow, and I don't
know the exact details, these motherboards trick the Adaptec
into asserting MEMR or MEMW on the XT connector at the wrong time.
This is bad news for any 8 bit cards out there which expect
reliable MEMR and MEMW signals.

As Stuart notes, newer motherboards don't seem to cause the problem
with the Adaptek/Hayes combination.

I even know one fellow who had an errant motherboard/Adaptek
combination who went so far as to move several Hayes boards off
onto an expansion bus, and then added some circuitry to
generate proper MEMW and MEMR signals on the expansion bus.
Don't try this at home.

-Rick Richardson