A little bit blown away by the biquinary coding used by this machine: at first it seems like a really odd design choice, but as the video explains at ~ 12:00 it makes fault finding for stuck relays much easier.

A really neat example of a design choice which makes a lot of sense for the problem domain in question, but it's also just so awesome that someone could actually see this solution: I can't imagine I would have!

The coding is just like a Japanese abacus!

https://en.wikipedia.org/wiki/Soroban

This, or possibly a variant, was also used in early electronic circuits, because the it would keep the load constant, if I remember correctly.

Additional information about the FACOM 128B:

In a July 2019 article, The Asahi Shimbun reports that Fujitsu technician Tadao Hamada keeps the ancient computer running. http://www.asahi.com/ajw/articles/AJ201907280007.html

On the IPSJ Computer Museum web site, the Information Processing Society of Japan provides more details about relay-based FACOM computers:

FACOM 128B background: http://museum.ipsj.or.jp/en/heritage/facom128b.html

FACOM 128A and 128B technical information: http://museum.ipsj.or.jp/en/computer/dawn/0012.html

Oh, that's a great video. Enough detail to understand how it worked.

As usual in that era, the big problem was memory. Memory in crossbar switches was so expensive that the programs are on paper tape loops. By then, they would have known how to put programs in crossbar memory.

I'd love a video on how that ticket gate machine works and why it's so complicated! https://i.imgur.com/56mmt5c.png

Looking at it, it appears to be a mostly mechanical solution (you'll need clocks, configurability based on location, detection of ticket orientation, etc). They probably got a version working 30 years ago and just stick with what works.

It's infrastructure level time horizons. When you have 10,000 of something built into the foundation of something that needs to be operational nearly all the time, upgrades take years. Look at the new york relay signaling based metro.

Why didnt they use rotary dialers for input. It would have been funnier. LM.Ericsson et.al. had ready-made 10 number dial receivers and input register in a handy 200 kilo rack.

180 words of crossbar memory apparently (a word being 69 bits), not 128 bytes as the name of the computer make me think at first

Loved the video and subscribed to the channel. I really appreciate efforts to keep old computers systems alive.

Wow, pretty impressive. Sounds like its playing own kind of music while its doing the calculations.

loved that video