Relay computers

From: Dwight K. Elvey <dwight.elvey_at_amd.com>
Date: Fri Sep 24 12:19:41 2004

>From: "Ade Vickers" <avickers_at_solutionengineers.com>
>
>At 22:50 23/09/2004, you wrote:
>
>>On Wed, 2004-09-22 at 11:17 -0700, Vintage Computer Festival wrote:
>> > On Wed, 22 Sep 2004, Jules Richardson wrote:
>> >
>> > > Apologies if someone's mentioned this one before. Quite possibly the
>> > > coolest gadget I've seen in a long time, though:
>> > >
>> > > http://www.cyberniklas.de/pongmechanik/indexen.html
>> >
>> > Awesome. It would be great to get stuff like this exhibited at the VCF.
>> >
>> > The web page is pretty smart also. What terrific technical and design
>> > work.
>>
>>Amazing, huh? If I read that right, it's 52 relays though (I don't know
>>any German) which makes it sound rather like a 'simple' relay control
>>system rather than an actual relay computer. I fired off an email to
>>them to see if they'll let me have a nose at the wiring diagrams.
>
>I did some back-of-a-fag-packet calculations (i.e. they might be completely
>& utterly wrong, in which case I'd appreciate corrections) on a relay
>computer...
>
>Assume you want a Z80-type CPU. This has ~8k gates. Typically, it seems to
>take 1 relay per input to implement any given gate. Now, I don't know how
>many "x-input" gates there are in a Z80, so I'll assume that - on average -
>it will require 3 relays/gate. Thus, we need ~24,000 relays to implement
>the Z80.
>
>If each relay needs, say, 25mA _at_ 6v to operate, then the peak current draw
>of our R80 (as I shall call it) could be around 600A (I think). And that's
>before we've added memory, i/o, etc.
>
>As for the heat/noise - well, IMHO it's worth building it just to
>experience that! Mind you, you'll need a lot of room: If you use 30mm by
>13mm relays, then the board space you need is at least 9.36 square
>metres... Still, if you assume that each board needs approx 40mm of space
>incl. airflow room, then you should be able to fit the R80 into 2 400mm by
>400mm by 2000mm cabinets (internal w/d/h)...
>
>The relays I've been looking at typically quote around 25ms to operate
>(either way), so I don't see how you could clock the R80 at anything faster
>than around 40Hz; and you'd probably want to drop to 20Hz to be on the safe
>side. The same relays quote a typical lifetime of 10e7 operations; so at
>20Hz, your R80 should last a little under 139 continuous hours of operation
>before relays started failing...
>
>Creating a screen driver should be interesting....
>
>Question: Wouldn't it be easier to implement an OR gate with no relays at
>all (just two wires joining together)? Or would you need to use the relays
>to keep the output voltage/amperage regulated?
>
>Additional: Hunting around for a suitable CPU to implement in relays, I
>came across the P8 CPU design (http://www.rexfisher.com/P8/P8_TOC.htm).
>This uses a 74LS181 4-bit ALU, which I reckon would require 149 relays to
>replicate. The only thing that confuzzles me is: what use, exactly, is a
>1-input AND? Several of these appear on the 74LS181 schematic...
>
>
>Cheers,
>Ade.

Hi Ade
 I think your counts are way off. As I recall, I did a schematic
a while back for a single bit of an alu and it only took about
4 or 5 relays. I did use DC and diodes for steering.
 The only thing you need a relay for is to invert. All AND and
OR functions can be done with the contacts of the relays. Even
the diodes are not needed if you have enough contacts.
 The life time of the relays are for maximum current conditions.
A relay computer would not be running quite to that level.
Also, relay coils can run at lower currents and voltage once pulled
in. The clocking circuits can be built to provide a pulse of
high voltage to pull the relays in and then a reduced holding
current to maintain.
 I still consider building a relay computer as a practical project
but like Karl Zuse, I think that memory is still the biggest issue.
Unlike normal silicon, single relays can be used to buffer both
inverting or non-inverting. This means that a single relay can be
used as a storage element. Silicon requires the use of two inverters
for a static memory.
 I've thought about an array of reed switches for memory with
bias magnets. These would be adjusted such that there would
be just enough magentic field to hold the reed closed once closed
but not enough to cause it to close by itself. The coils would
then be used to open and close by applying a counter or aiding
field to the reed/magnet. The magents would be arrange with alternating
directions such that there was no accumulating of the magentic
field across the array of the reed RAM. ROM can be done with
and array of diodes and switches.
 I'd keep the ALU simple since most times, it doesn't need to
do arithmatic. A single bit ALU with a few serial storage
shift registers would be sufficient for most math operations.
Dwight
Received on Fri Sep 24 2004 - 12:19:41 BST

This archive was generated by hypermail 2.3.0 : Fri Oct 10 2014 - 23:37:31 BST