DEC bus transceivers (was Re: Unibus disk controller with modern storage)

[allison]

On 10/22/2016 06:40 PM, David Bridgham wrote:
> On 10/22/2016 12:44 PM, shadoooo wrote:
>
>> What kind of bus transceivers did you used for the QSIC, specially
>> because you have
>> to go from 5V open-drain logic to 3.3V logic?
> To add to Noel's answer, here's a picture of our current prototype board.
>
> http://pdp10.froghouse.org/qsic/qsic-wirewrap.jpg
>
> Coming up from the QBUS, the first two rows of chips are the bus
> transceivers.  The next row and a half are the level-converters.  Then
> the two large ribbon cables run off to the FPGA module we're using for
> development.  The two small ribbon cables go to the indicator panel. 
> Just the bus interface takes over half the area of a dual-height board! 
> I've played around with laying out what might be the production board
> (when I get tired of Verilog and want a mindless break, I doodle with
> kicad) and I've got it down to a row of 8641 bus transceivers and a row
> or two of the level-converter chips.  It's better but still a good
> fraction of the entire board.
>
> http://pdp10.froghouse.org/qsic/proto-pcb.jpg
>
> Now I thought, what if my idea of that two MOSFET bus transceiver would
> work?  What would the board look like then?
>
> http://pdp10.froghouse.org/qsic/qsic-smt.jpg
>
> Obviously that could be squeezed down a lot more.  Even another
> transistor or two per bus line would still be fairly small.  Doing the
> bus transceiver and level-conversion in one step makes a big difference.
>
> For the QSIC, we're going to have sufficient room and we're able to find
> enough old bus transceivers to continue on as we're going.  Still, I'd
> sure love to have an option that used production parts and took up less
> board space.
>
>

For bus to card LS241, LS244 and LS245 were used in the day by heathkit
(H11)
For bus to card with hysteresis LS14 was a common choice along with
LS13 or maybe LS132.  

Other considerations the DEC specs were based on available devices of
the day (late 70s)
and in the last 40 years there are a plethora of better devices that
replace those easily.  
The key being Schmidt input (bus to card level) for noise immunity and
robust drive for the
output devices.  Many of those were also used for S100 (less well
thought out and
often noisier.) with good success.  I've copied and used those for
making boards as
needed for Qbus and Omnibus as those represent the common DEC buses
outside of SCSI
in the hardware I own.

One thing I'm adverse to is bus to CMOS without buffering or the reverse
even if
the CMOS can drive the bus.  Unexpected ringing or spikes on the bus can
play
havoc with those as well as poor ESD practices.

Now the problems of board space and available parts may force SMT and CMOS
for the internal circuits and also power conservation (and thermal
management).


Allison