building my own relay computer

[Cory Heisterkamp]

Ben, 

I'm planning on 8-bit paper tape for I/O, but I was in touch with a gentleman from Brazil a few months ago that designed his architecture around 5-bit words. This allowed him to use the Baudot code and make use of teletype equipment that can still be found without too much difficulty. 8-bit ASR-33s are worth their weight in gold these days, but the 5-bit model 32 (or a 28 and tape peripherals) is much more reasonable. -C


On Feb 13, 2015, at 7:14 PM, ben wrote:

> On 2/13/2015 6:02 PM, Cory Heisterkamp wrote:
>> Tom, any reason for choosing the Harvard architecture?
>> 
>> I'm currently in the middle of a relay computer build so I'll toss in
>> my 2 cents. The most important thing I hit on early is to define the
>> instruction set, work out a number of potential applications/uses on
>> paper (or excel), and step through each instruction to make sure the
>> machine is capable enough. You'll probably discover a number of
>> efficiency improvements you can make to the hardware while doing so,
>> and a few "hardware subroutines" worth adding. You can also baseline
>> how long a program will take to run if you settle on a clock rate,
>> and determine if there's a better way to do it.
>> 
>> Ultimately, relay computers are SLOW. Don't take 6 cycles to do what
>> could be accomplished with a couple extra relays and a single pulse.
>> Well thought out HW and instructions will pay big dividends when the
>> soldering iron comes out. -Cory
>> 
> The alu has never relay been a problem with relays, but what about main
> storage? Punched tape (paper/film/card) is no longer with us for I/O.
> Ben.