Hi,
Does anyone know the state of progress with the EDSAC replica project. There has been nothing from TNMOC in ages. Their web site says they are looking to run its first programs in 2019.
The most recent info I can find is from a team member on his personal YouTube channel showing them testing the long transfer tank, and that was a year ago.
One can be a supporter of TNMOC (£50 annually) and then get a quarterly newsletter… and we did get a mention of EDSAC earlier this year. I won’t quote directly, but I will say that there are reports of checking the correctness of functional units, of single-stepping, and of running simple test programs.
Photo of the display unit from that update:
I’d recommend keeping an eye on Computer Resurrection which also comes out quarterly, which might well have news from time to time. Here’s a recent article on EDSAC (as a computer, not as a replica project).
And from there, a link to a useful page on EDSAC (but again, not on the replica)
EDSAC 1 and after - a compilation of personal reminiscences
Oops, I see in fact there is an update in the latest Resurrection: here’s just a small part - it’s in the Society Activity section which I’d not noticed:
We continue to make good forward progress with EDSAC and continue to enjoy an extended period with very few component failures.
Since the January report the most significant step forward has been to demonstrate writing of the results of computations in the Arithmetic Unit being successfully written back to store (in the right location and with the correct value) via the Transfer Unit. This required some work on the interface between the Arithmetic Unit and the “Master Output Bus” and between there and the Transfer Unit. The MOB is where several data sources are merged for machine functions that write to store (data input via I order, arithmetic results via T and U orders, initial orders and engineers’ switches). Each source has to be adjusted to bring signal levels into line, and timing adjusted to ensure each is brought back into synchronisation with the main store, whose timing effectively controls the whole machine.
…
We have also overcome the previously reported problems with running short programs, accessing operands with addresses close to the current order, and single-stepping through programs. Many of these problems were eventually attributed to the predominance of AC coupling in the EDSAC design and the consequent build up (or leakage) of charge in circuits holding flip-flops on or off and making their behaviour sensitive to the timing and sequence of orders being executed.
Thanks EdS,
That’s exactly the info I was looking for, I have bookmarked CCS for future reference. The last time I was at TNMOC, they were using Arduino based probes for signal injection. I see that has come a long way with the EDSAC Distributed Logic Analyser, and good progress is being made.
Hopefully Andrew will find time to make another video at some point. I’ve followed those with interest over the years.
1 Like