Supply voltage for 2N2/256-BSCP

Allison ajp166 at bellatlantic.net
Wed Jan 16 06:53:02 CST 2008


>
>Subject: Supply voltage for 2N2/256-BSCP
>   From: Robert Nansel <bnansel at bigpond.net.au>
>   Date: Wed, 16 Jan 2008 21:24:59 +1030
>     To: cctalk at classiccmp.org
>
>On Wed 16  Jan Brent Hilpert said:
>
>> ...
>> I hope you're not looking to mimic TTL - you'll have trouble sourcing
>> multi-emitter transistors in discrete form. (I know - it's outside  
>> your design
>> constraint anyways.) More seriously, looking to TTL or standard DTL  
>> ICs for
>> design would kind of chew up the transistor count very quickly. Is  
>> there a
>> reason you're not looking to period/original discrete-component  
>> logic designs?
>
>I am using discrete diode-transistor logic;  I'm just trying to  
>figure out what a sensible Vcc should be.  I've seen discrete  
>transistor logic designs with Vcc voltages from 3.6V clear on up to  
>90V, plus and minus (and often both).  I suspect the higher voltages  
>were more out of habit from vacuum tube days, but perhaps there were  
>other reasons.

More in the realm of other reasons.  Transistors of the day could not 
switch much current so I^2R ruled where higher voltage at the current 
they could switch was used to get enough power.  This was balanced 
against device breakdown voltages.  Other factors that are to be 
considered is higher voltages allow larger signal swings for better
noise immunity at the cost of speed (R*C).

>Since I'll ultimately run this whole thing from batteries, I would  
>naturally prefer to deal with 6V or 12V (or even 3V or 4V, depending  
>on the battery chemistry), less the <mumble> dropout voltage of my  
>regulator.

Forget batteries.  The voltage stability will be important and as you
get enough powered elements you'll find that all those milliamps start
to pile up to amps, potentially lots of them.

I'd stick to 5V for interface to current parts where needed but you
might find time to look at old machines and their design.  One of the
things done was -V to reverse bias transistors to help with stored 
charge for faster switching.  The -V was in the range of -3 to -10V.
The most recent of transistor machines like the PDP-8 (link-8 too 
as well as early DEC modules and the TX2 machine) would be an example
to look at.  FYI: the really ond machines used PNP transistors as the
Germainium technology was better able to make those, that only means 
all the votages are "upside down" compared to what we currently expect.

>So far I've been able to keep everything in the ALU data path to a  
>depth of two diode gates.  I'm using the complementary outputs from  
>three flip-flops for true and inverted versions of the A-word and B- 
>word bit streams and Carry inputs to the adder.  It takes more gates  
>this way, 25 diodes but no inverters, versus 12 diodes with two  
>inverters.

Watch fanout, that is output of the FF being loaded with enough current
sinks to inhibit operation or make for lousy noise immuunity.  Many 
designs used enough inverters to avoid needing buffers, again a 
balanacing act.


Allison



More information about the cctech mailing list