11 Dec
2012
11 Dec
'12
6:58 p.m.
On Dec 11, 2012, at 8:45 PM, John Wilson wrote:
Exactly. Going with discrete parts was 90% for future-proofing;
the only real dependency (since the dual FETs are pretty common)
is the comparator, but I think the MAX9107 has a pretty common
pinout for a quad op amp/comparator, so if it goes the way of
the dodo, I may still be in luck.
Good to get feedback. Thanks!
- Dave
On Tue, Dec 11, 2012 at 02:55:38PM -0500, David Riley
wrote:
Huh, I hadn't thought of that. That's a good idea, and I might be
able to do that with the SMT parts I'm using. If I used generic FET
and comparator footprints, I could at least make a million (well,
a few thousand) little boards and sell them at cost for a few per
dollar.
An 8646 might be a tough sell in such a tiny part; between the
parity generator and the latches, it would be stuffed pretty full.
Some of the simpler driver/receiver/transceiver chips are a lot
easier, like the DS8641.
I might be persuaded to buy a few for
characterization purposes (I'll
let you know off-list if that's the case), but I'm curious to know if
you're working on a RoHS/modern replacement.
Not as a pin-compatible part for repair of existing boards, no. If that's
what you're going for, that's a great idea! An 8646A clone circuit on a tiny
daughterboard might be cool too (for the KS10/VAX crowd). But I've got first-cut PCBs stuffed (and awaiting
firmware) for a couple of
projects which need to talk to DEC busses but don't need the DS8641N pinout:
That's what I'm doing (for QBUS first, since that's what I have the
hardware to test, then Unibus once I have more tuits).
- Unibus battery-backed RAM and clock/cal board, (for
my own use -- I know it's
trivial but my main 11/34a's MS11Ls have gotten super flaky and I always
wanted a TOY clock). This is a dual-height board, 3" tall, which fits in A/B
of a MUD slot. Embarrassingly, it has a CPU, just to save square inches, but
now that it's there, doing a KDJ11E-compatible TOY clock becomes easy, and
doing config through an RS232 port is easy too so there are no jumpers.
CPUs are embarrassingly cheap these days. I'll be using an FPGA
for the QBUS logic, which is overkill, but a CPLD is too small and
an FPGA lets you add things like DMA engines practically for free.
- Unibus universal disk controller (will be a D Bit
product if successful).
First prototype is a 6"x6" board with ribbon cables for a PCI/Unibus
adapter (without the M9014-like Unibus paddle card) and an M9302-like
terminator daughterboard with female Bergs on the back, so I can debug
the firmware using E11 w/o hauling a big noisy BA11K into my little office.
If I get this working, the next rev will be a quad SPC. SATA/USB/SD,
with an 18-bit data path (obviously!).
Interested in reselling a QBUS universal disk controller with
user-programmable logic? Talk to me off-list if so.
Anyway for these I'm seeing if I can get away with
receiving the bus using
74LVC240s (etc.) and transmitting using a gazillion 2N7002PWs with 100 ohms
in series with the gate (hoping that plus the gate capacitance will get the
slew rate into a good neighborhood). Technically none of this is OK but I'm
hoping in practice it works. Once the firmware's taking shape, a scope will
tell all...
I'm actually going for something that's 100% in spec for DEC
buses, which means using real comparators for the input. The
unpleasant part of that is finding sub-35ns comparators that
don't cost an arm and a leg per gate; so far, the one I've
settled on is the MAX9107, which is a quad part that costs
about $1.50 (<$0.40/gate) in 1k+ quantity. Add a 74LVC125 in for
input gating (for tristate, or 74LVC00 for NAND like a DS3841)
and you have a reasonable input buffer.
There are lots of dirt-cheap FETs that work; the 2N7002 looks
decent. I'm probably going with some duals in an SC89-6 package
to save some real estate, but I'm trying to work out whether it's
worth the extra cost per pin (it almost doubles). Your driver
solution should be 100% DEC kosher (at least per the QBUS spec),
since there's no way your rise time is going to be under 5 ns
with an open-drain arrangement.
I'm
currently prototyping
a board using comparators for the receivers and discrete FETs for the
drivers (no ICs I could find had compatible slew rate, capacitance and
drive strength with DEC's specs), but I'd be curious to know if you'd
found anything better.
Yours sounds better than mine, so, nope! Probably just as well to do it with
loose parts though -- you don't want to get dependent on *another* oddball
chip that'll be rare five years from now.