TTL homebrew CPUs

Allison ajp166 at
Sat Jul 7 17:16:49 CDT 2007

>Subject: Re: TTL homebrew CPUs
>   From: Lars Hamren <hamren at>
>   Date: Sat, 07 Jul 2007 21:38:15 +0200
>     To: cctalk at
>The first computer built by Computer Automation, circa 1967, was the
>PDC 808. It was an 8-bit TTL-only machine with a very basic
>instruction set. In "Computer Technician's Handbook" Brice Ward
>describes tat machine in great detail. Should be enough to build a
>replica from.
>A scanned version of the relevant parts of that book is available from
>PDC stands for Programmable Digital Controller. The word "computer"
>was avoided because people were afraid of them!
>The "8" means 8 uS cycle time, and "08" means 8 bits. The 208 was
>three times as fast (2.66 uS cycle time, hence the "2").

I've seen a lot of discussion on making a CPU from TTL or MSI (bitslice).
The task is not that bad and the parts to do it are very available still.
The assembly task with out a premade board can be large as it's a lot of 
repetive circuits.  If the word size is kept reasonable but useful and 
that seems to be 8 to 16 bits people are doing it.  The variable word
length 8bit designs are scarce but requires less bus "width" to get a 
workable system and have performance in the zone of faster than serial 
but not as fast as wide word machines.  Using that approach you can 
have a 32bit machine with 8bit busses saving a huge amount of repetetive 

For those that want small and less significant part counts there are
minimalist 8/12/16 designs like TOY or Pilgrim VSC (very simple computers).
A board for one of them would have wide appeal and useful for those that 
wish to see the guts of cpu without probling a die or having a whole PDP-8.

The biggest difference between building a PDP-8 (or whatever) then and now 
is not using TTL (or cmos) or bit slices but the availability of fast and 
large RAM that is easy to use.  Many earlier systems were flavored by the 
ram they could interface or the needs of the memory.  Same for the older 
serial designs to were done to economize on then costly registers and 
arithmetic logic along with serial memories.

There is a lot of "you can do" but building a CPU comes down to what do 
you want to do and can reasonably achieve or can practically play with
once functional.  Unfortunately some want a cpu that can run unix and an
IP stack others want just enough to qualify as a computer for close up 
examination of "just how do they do it?" and that represents a large 
spread of design and architecture possibilities.  Having built machine 
from very simple state logic to an elaborate 16bit bitslice I can say 
all are interesting but some are a lot too much work.


More information about the cctech mailing list