jbm@uncle.UUCP (John B. Milton) (05/01/89)
What you see below is the text of a letter from John Lydic. Return mail to him
should come to me as he is not yet on the net. The meeting refered to below
was attended by: John Lydic (not on the net yet), John Milton (jbm@uncle),
Jim Prior (jep@oink), Gary Sanders (gws@n8emr), Greg Ratcliff (gwr@nz8r) and
James Nugen (jcn@uncle).
John
---
Date: 4/26/89
From: !ka8lvz!jwl
John,
Here is the information on the new 3B1 combo card. The design is
90% complete and is just waiting on the AT&T documents to doublecheck
the last of the design. We'll convene another meeting of the group to
review the schematics in about a month. The layout can start shortly
after that. It looks as though the bus spec for this thing is real
loose compared to other 68K systems I've worked with. I am trying to
avoid PALs if possible to make the board more generic for other hackers
who may not have our resources. With luck, we could see prototype
hardware by mid to late summer.
General Board Spec
------------------
8 serial ports 2 asynch with full modem pinouts
2 ports synchronous [BOP/COP] capable with DMA
SCSI with DMA
4 Meg Memory 2 meg linear addressed for processor memory
2 meg bank addressed for RamDisk
Modem 103/202 compatable connected in parallel with
output from synchronous serial for packet
Details, Details, Details
-------------------------
Serial ports: Three (3) Signetics 2681 Duarts for general serial
communications. One of these is the big pinout
part with all of the modem control lines connected.
One (1) Signetics 68562 Dual Universal Serial
Communications Controller (DUSCC). This part will
support baud rates through 38.4K plus 56K and 115K.
The DUSCC may be used for standard asynchronous
communications or may be used in synchronous mode
for COP (BiSync) or BOP (HDLC) with numerous CRC
algorithms.
SCSI One (1) Western Digital 33C93 SCSI Bus Interface
Controller (SBIC) is used to handle the SCSI port.
The SBIC has its' own 48ma output drivers, and
contains numerous "combination" commands to make
software management and interrupt overhead minimal.
Memory Two (2) megabytes linear addressed processor memory
and two (2) megabytes bank addressed auxilliary
memory for use as RamDisk, etc. Both of the banks
employ 1 megabit ZigZag Inline Packages (ZIP).
The processor memory contains byte parity, and a
refresh scheme both consistant with the standard 3B1
memory scheme. The auxillary memory contains word
parity and may be addressed on word boundaries only.
Parity errors during reads, are detected by reading
a status register after the transaction.
Modem One (1) Rockwell 7910 World Modem Chip is used to
enable amateur packet software to directly control
a transceiver on one channel at 300/1200 baud. This
chip is used in the Kantronics TNC and appears to
work well enough (better than those drifting PLLs
in my HD-4040). In addition, addressable latches
are employed to allow control of two radio channels
(PTT, etc), without having to keep an image of the
latch.
DMA One (1) 684x0 DMA controller is employed in two
possible configurations. If a 68440 (2 channel)
is used, a channel is available for the SBIC as
well as a half duplex channel for one of the DUSCC
ports (The same port with the modem connected).
If a 68450 (4 channel) is used, the two channels from
68440 are available as well as a full duplex channel
pair for the other DUSCC port. Allowing DMA use
on the DUSCCs can enable the 3B1 to do dual-channel
packet radio at up to 56K with an external modem
(or maybe 115K if modems become available).
NOTE that I say modems, NOT TNCs, because the 3B1s'
hardware would handle all levels except the physical
layer.
I/O Obviously all of this I/O presents a connector
problem which will be resolved with the scheme
we came up with at our last meeting. The SCSI,
modem outputs, and radio control lines will be
on the board on standard dual inline headers.
Still trying to decide the number of pins for SCSI
which is normally 50 (2x25), but will fit on 25
(2x13) if you don't mind a funny cable. The modem
and radio control lines will be on the second.
The ribbon cables for this I/O will just come out
the back of the 3B1. I think the scheme Gary (N8EMR)
came up with will work OK for the serial I/O. To
refresh your memory, and explain for the net, the
new card will extend about 3-5 inches out the back
of the computer. Eight (8) PCB vertical mount DB25S
connectors as well as eight (8) RJ45 telephone style
connectors will be mounted on this extension. In
addition the layout will contain space for 2 dual
inline headers, one inside the computer, and the
other on the extension. The serial I/O may be used
by simply plugging into the piece of the board that
extends out the back, or the extension may be trimmed
off and mounted separately employing the headers and
yet more ribbon cable to connect it to the main board.
For the NET
-----------
The board design is relatively firm as far as chip selection, etc.,
but we could use some input on a few of the final details.
Since only two of the eight serial ports will support full RS-232
handshaking, we could use 6 DB9S, and that might make room, for
some of the modem or radio signals to be brought out.
The 7910 World Modem Chip is relatively old and other better, cheaper,
more capable chips may be available. This is not my area of expertise,
so let us know if there is any information that may help us, out there.
--
John Bly Milton IV, jbm@uncle.UUCP, n8emr!uncle!jbm@osu-cis.cis.ohio-state.edu
(614) h:294-4823, w:764-2933; AMPR: 44.70.0.52; Don't FLAME, inform!scs@lokkur.UUCP (Steve Simmons) (05/02/89)
I am croggled. That's not a card, that a f*cking universe. But since you folks are laying the universe at our feet... Primary question is $$$$$. Several points on the RAM: First, must the card "subdivide" two and two for main memory and ram disk? I have a 1.5MB oldie, and would appreciate a 2.5/1.5 split, or even 3.0/1.0. From my point of view it's better to have 4.0 main ram + 1.0 ram disk than 3.0 with a 2.0 ram disk. Will the memory interfaces (ZigZag? never heard of it. Is it a SIMM?) for the ram disk support the 4MB inlines? Be nice to convert that 2MB ram to 8MB. Friends who follow the industry closely claim ram prices will nosedive in Sept/Oct, when (literally) everyone's 4Mb chip plants go on line simultaneously. Device driver for the RAM disk? Interfaces: > 8 serial ports 2 asynch with full modem pinouts > 2 ports synchronous [BOP/COP] capable with DMA This looks contradictory. Do you mean 2 full modem asynch, 2 partial, and 2 synch? Can't tell from the details. Will the serial ports use std 3b1 serial drivers? Or can I run my TB at 19.2? Yes, I see the serial chips support 19.2 and higher (higher! on a 3b1. jeez.) > SCSI with DMA Device driver? Will it support synchronous SCSI? > I/O Obviously all of this I/O presents a connector problem . . . No kidding! It sounds like you've been seriously thinking about it, but from the description it's not coming clear. Can someone do up a drawing? Sticking out 3-5 inches from the back sounds decidedly risky -- but it sounds like this can be snapped off, ribbon cabled, and mounted elsewhere. Right? > Still trying to decide the number of pins for SCSI > which is normally 50 (2x25), but will fit on 25 > (2x13) if you don't mind a funny cable. Hmm. If by funny cable you mean something like the Macintosh standard I'd vote for that. Otherwise I vote 2x25. Any savings that require custom cables aren't really worth it. Mondo combo, indeed! -- Steve Simmons ...sharkey!lokkur!scs scs@lokkur.dexter.mi.us "Gordon Way's astonishment at suddenly being shot dead was nothing to his astonishment at what happened next." -- Douglas Adams
ccs@lazlo.UUCP (Clifford C. Skolnick) (05/02/89)
In article <524@uncle.UUCP> jbm@uncle.UUCP (John B. Milton) writes: > Since only two of the eight serial ports will support full RS-232 > handshaking, we could use 6 DB9S, and that might make room, for > some of the modem or radio signals to be brought out. You could also use modular connectors. Another sugestion would be a 50 pin connector that goes to a box with all the real connectors. I've seen both these used for 386 multi-serial port cards. -- "I'd rather stay here with all the madmen, than perish with the sad man roaming free" -- David Bowie "Life is a test, only a test. If it was real, you would have been given much better instructions." Clifford C. Skolnick / (716)427-8046 / ccs@lazlo.UUCP