9 Sep
2005
9 Sep
'05
5:06 p.m.
From: "Ethan Dicks" <ethan.dicks at
gmail.com>
On 9/9/05, Philip Pemberton <philpem at dsl.pipex.com> wrote:
Hi
As I recall, the GPIB chip made by TI ( forget the number )
had a data sheet that explained the states. Not much
to it as I recall, mostly just handshake stuff and address
recognition. It seemed like there was some timed stuff that
might not work well in software but for the most part I'd
suspect that Ethan is right, one could do it all in software.
Dwight
Hi,
Does anyone happen to have any data on the GPIB state machine, or perhaps a
GPIB slave interface implemented in LSTTL?
TNW made a serial interface for the PET that did not contain a
microprocessor. Perhaps you could search for schematics on that.
I've got a copy of the HP "Tutorial
guide to HPIB" and "PET and the IEEE488
Bus", but none of them cover the state machine...
With that as a guide, if you don't find an existing state machine,
there's plenty of information to generate one of your own. You could
even prototype one in software.
-ethan
9 Sep
9 Sep
6:52 p.m.
On 9/9/05, Dwight K. Elvey <dwight.elvey at amd.com> wrote:
As I recall, the GPIB chip made by TI ( forget the
number )
had a data sheet that explained the states. Not much
to it as I recall, mostly just handshake stuff and address
recognition.
The common chips are:
TI TMS9914A
NEC ?PD7210
http://sine.ni.com/nips/cds/view/p/lang/en/nid/1267
It seemed like there was some timed stuff that
might not work well in software but for the most part I'd
suspect that Ethan is right, one could do it all in software.
One issue is to identify what GPIB devices you expect to talk to.
Older ones, especially ones that implemented the protocol via ~1MHz
8-bit microprocessor, are unlikely to be able to handle 1MB/sec speeds
that the newer "IEEE 488.2" devices can. If all you want to do is
talk to old devices at a few KB/sec, it really doesn't matter if you
have a hardware or a software implementation. For the newer,
high-speed stuff, I'd expect you'd need an ASIC to pump the bytes
through in a timely fashion.
OTOH, even Commodore had a little fiddly hardware thing to handle
handshaking turn-around. Andre Fachat mentions that in his extensive
pages covering Commodore hardware and his own home-built
mostly-PET-compatible 6502 computer. I do not know about the
Commodore acoustic couple (C810?), but disks and printers tended to
have 6502/6504/etc software implementation of GPIB.
-ethan
10 Sep
10 Sep
12:32 a.m.
In message <f4eb766f050909115237677199 at mail.gmail.com>
Ethan Dicks <ethan.dicks at gmail.com> wrote:
One issue is to identify what GPIB devices you expect
to talk to.
Older ones, especially ones that implemented the protocol via ~1MHz
8-bit microprocessor, are unlikely to be able to handle 1MB/sec speeds
that the newer "IEEE 488.2" devices can. If all you want to do is
talk to old devices at a few KB/sec, it really doesn't matter if you
have a hardware or a software implementation. For the newer,
high-speed stuff, I'd expect you'd need an ASIC to pump the bytes
through in a timely fashion.
What I'm wanting to do is implement a slave device - in this case it'll be a
data logger. The problem is, I don't want to end up tying up CPU resources
monitoring the bus - it's going to be sharing the bus with a multimeter, a
logic analyser and (if I ever get off my BFA and build the damn thing) a
power supply.
What I'm after is some way to receive data bytes into a buffer, then do some
address-comparing and only interrupt the CPU if data arrives after an address
match is detected. Most of the logic seems to be along the lines of (going
from memory):
ALWAYS @(posedge data_arrived)
IF ATN = 1 THEN
IF data_byte & 0xE0 == 0x20 THEN // TA
IF data_byte & 0x1F == my_address THEN // Talking on
CAN_TALK = 1
ENDIF
IF data_bute & 0x1F == 0x1F THEN // UNTalk
CAN_TALK = 0
ENDIF
ENDIF
// ... same sort of thing for Listen Addressing
ENDIF
END ALWAYS
The problem is working out how to rig up a state machine or something similar
to handle the handshaking...
Trust me, if I had any spare MC68488s or similar, I'd use them. I've got some
Fairchild 96LS488s, but no data at all. Finding a datasheet for the 96LS488
seems to be about as easy as finding 99.9999% pure Unobtainium.
For master-side addressing I'm going to use the PC and a CEC GPIB card.
Speed is not the big issue here - I'm doing this mostly for the "fun
project"
aspect. My idea of fun is, unfortunately, somewhat warped...
Later.
--
Phil. | Acorn RiscPC600 SA220 64MB+6GB 100baseT
philpem at philpem.me.uk | Athlon64 3200+ A8VDeluxe R2 512MB+100GB
http://www.philpem.me.uk/ | Panasonic CF-25 Mk.2 Toughbook
... (((((YOU)))))((((ARE))))((((((FEELING))))))(((((SLEEPY)))))
7050
days inactive
7051
days old
2 comments
3 participants
participants (3)
-
dwight.elvey＠amd.com -
ethan.dicks＠gmail.com -
philpem＠dsl.pipex.com