Document and create a modern replica of the Harwell Dekatron computer known as the WITCH
so i am continuing to get ready for the trip to the National Museum of Computing to demo the WITCH-E prototype! as the WITCH-E is mean to be easily disassembled and transported, i noticed that the flat head screws on the left were a little time consuming and difficult to assemble, so i've replaced them with some anodized aluminum thumbscrews on the right. these are much easy to use on the fly and make reassembly a snap! they also look a lot better than the plain old screws....
WITCH-E has been reassembled with the new acrylic from Ponoko.com and new wiring. I've made the panels smaller and specifically designed them to fit into a Pelican hard case for traveling. the single green board to the side is the register select test board, and since it is a test board, i've used green soldermask to distinguish it from the blue portions of the actual WITCH-E system. As more of WITCH-E is completed, panels will be added to the right where the register select test board is located. we are all set for travel to the National Museum of Computing for the first public demo of the WITCH-E!
Please Consider Helping
Please consider donating to my fundraiser for the National Museum of Computing! I am matching donations to $1500USD. We have a shared computing history and we need to help preserve it as well as keep computers like the Harwell WITCH on display and functional! If you have questions about how the funds are spend or general questions about the museum and it's mission, feel free to contact me directly at firstname.lastname@example.org
In the fall of 2015, on a dull grey Sunday morning in Block-H of Bletchley Park, I was afforded a great honor. Part of the WITCH-E project is to document as many aspects of the Harwell WITCH as possible, and one of the things I found absolutely no information on was how the wiring on the back side of the WITCH was handled. On this particular Sunday morning, prior to the National Museum of Computing opening to the public, I was allowed to take detailed pictures of the back side of the WITCH. Using this information along with the original schematics, I've replicated the "Interconnects" in generic fashion for the WITCH-E. Here are some of the photos of the original WITCH:
NOTE: Pictures are used with the permission of the National Museum of Computing for Non-Commercial purposes: Creative Commons CC-BY-NC 4.0
I took a quick break from "day job" work to solder on the headers for the new Transfer Interconnect Board. This first revision of this board was simply connections between the headers with no components. After working with the original design for a while and doing some scope captures it was pretty clear that i was getting a lot of signal noise and the rise/fall time of the signals was not within the required ranges for the 7400 series devices that were being connected. With that in mind i did a revision of the design adding some non-inverting schmitt triggers to the the design. oddly enough i was preparing the log post when i realized that Hackaday.com had a new post today about using schmitt triggers. The article called About Schmitt (Triggers) , has a great overview of exactly how adding schmitt triggers to this design helps with both the noise and rise/fall times of the signals.
I was able to complete the assembly of the transfer block last night. the transfer block includes the Transfer Unit, Carry Select, and Carry Unit. each of the modules are soldered together with wire between the 0.1" headers and mounted with M2 screws with 5mm brass standoffs....
So i have the latest revision of the transfer unit modules assembled. this is the fourth edition of this module as I've slowly been able to break down more and more of the original schematics for the WITCH and understand how they are actually translated into modern components and functionality. i started off with this simplified diagram from the Electrical Engineering magazine article on the WITCH:
My initial attempt at the transfer unit was to simply identify the functionality of each of the components and match them with a modern equivalent:
I quickly realized that this diagram from the article was a simplified diagram, and began combing through several pages of schematic. I've slowly identified all the core functionality with this being the resulting schematic (minus the Dekatrons which as done separately):
For the next round of testing and debugging, I have a bunch of new modules to build up...
the Carry Select section on a decimal based computer is the section that when when a carry needs to happen, allows for the carry impulse to pass to the receiving location. this "mux" must allow for the initial counting pulse to be selected during a transfer, and then later based on the Carry Block signals to allow the pulses to pass. The OR gate i was using to combine these signals didn't seem to have enough drive strength to properly drive the receiving location logic properly due to the length between them. so on this revision I've added a Schmitt trigger to the design, which seems to solve the issue with both drive strength and line noise.
here is the schematic:
This time last year I was getting ready for a Hackaday event at the Dallas Maker Space. It was a fantastic event with lots of lighting talks, interesting people, and lots of amazing projects. I did a lighting talk about the WITCH-E project (I wish I could find video of it!), and setup a small table to demo some of the early prototypes. Hackaday's Mike Szczys wrote up a short piece about the project the following week entitled "BOOTSTRAPPED TOOLS, LIVE STOPPED MOTION, AND A DEKATRON COMPUTER"