The few times I’ve had the lid off of my 5100 have all been anxious moments, as I have no idea where I’d find replacements for any of the ICs or SLT modules inside the machine. I resolved early on that my recovery of the 5100’s non-executable ROS – the ROS that contains the programming for the 5100’s BASIC and APL interpreters – would be as minimally-invasive as possible.
In accomplishing this recovery I may have used more compute than all the IBM 5100s ever built have carried out over the past 44 years.
Recording a screen dump and writing a program to OCR it is clever and all, serious props to the author for eventually getting it done. It just seems like total overkill. Wikipedia mentions a communications adapter and 9600buad research device coupler, would those have worked? Even lacking those, I’d imagine it would have been much easier to devise a simple protocol and writing a 10-15 instruction program to cycle through memory and bit-bang it onto the bus where he could capture it using his favorite micro-controller.
He mentioned encountering the limits of his logic analyzer, but I still feel he could have gotten results far quicker sticking with that route. Or even with the OCR route, he could have made it more robust by displaying much easier patterns to detect without the need for OCR neural nets.
All in all, he got the job done in a crazy way, and despite my critique that’s what I find so amusing! Articles like this are fun to read. I could probably share some of my own stories, though I automatically assume everyone would just find it too boring anyways.
https://www.osnews.com/story/129744/decoding-photographs-of-the-ibm-5100s-executable-ros/
This is incredible work, and is something that is going to be usable in the figure as older SCADA and embedded systems fail, and we need to recover data from them.
My hat is off to him!