Repairing a 1960s mainframe: Fixing the IBM 1401's core memory and power supply | Hacker News Reader

Repairing a 1960s mainframe: Fixing the IBM 1401's core memory and power supply

righto.com

145 pointsfernly18 comments

I've actually seen this machine in person, I have a punch card with my name on it somewhere that it printed out.

As someone who missed this era of computing entirely, but was quite familiar with it, it was very awesome to see in person. One thing that I really wasn't expecting was the smell in that room. I spend a lot of time in data centers but the room at the CHM with the 1401 smelled more like a machine shop than any data center I've ever been in.

The CHM, in general, was awesome to visit, but definitely go on a day when you can get the 1401 demo.

Yes, you get that characteristic machine oil smell from the old machines. The older punch card machines from the 1930s actually have a bunch of metal oil tubes running through them to keep everything lubricated.

My grandfather was an engineer and manager at IBM in the era that the 1401 was built. He worked on the IBM 1620, and always framed them as kind of in competition with each other. The 1401's internal codename was "SPACE", and the 1620's was "CADET" (apparently as a kind of matched pair?)

It was definitely a treat to see it, with the personal connection I had. I'll be going back if I'm ever in Mountain View again.

The owners manual about keeping things oiled and preventing cards from binding (it's read in a mellow British accent in the music piece I mentioned in another comment). I never thought about it until now, but I guess being around this machine would be more like working on a car or factory equipment than what we would consider a data center machine.

That, and a day when they're doing the PDP-1 demo?

Whenever I hear about the IBM 1401, I think back to Jóhann Jóhannsson IBM 1401 symphony. It's in five movements and uses sounds generated from an actual 1401, as well as very interesting readings from the user's manual. It's a truly amazing work:

https://www.youtube.com/results?search_query=IBM+1401+Jóhann...

Kind of mind-blowing that this machine needed two big TO-3 package transistors to drive 4000 bits of memory. Nowadays you could probably power a few dozen Raspberry Pi's with that amount of current.

Remember that the core memory is magnetizing metal rings in microseconds, so it takes substantial current pulses. But not as bad as vacuum tube computers - I've been working with an IBM 705 tube module from the 1950s and you can feel the heat coming off it.

With music amplifiers basically a couple TO-3's with nominal passive heatsink can output the equivalent 50 watts of a pair of 6L6 vacuum tubes.

The tubes run hotter because of the always-on heated cathode which is not needed for solid state amplification.

BTW still have a couple Diablos and a DEC drive, for parts, that could use a home.

Also an HP1000 in a full rolling rack the size of a refigerator, plus a VAX 4000, it's a mini-fridge. These are in good shape.

Fuzzfactor, can you contact me (ken.shirriff@gmail.com) about the HP 1000? I found someone who is interested in it.

Out of curiosity, why couldn't they just replace the bad inductor, instead of swapping out the board entirely?

I replaced the inductor with one from a different bad board, but it took a while to identify the inductor as the problem with the board. Some people wanted to get the system running right away, and there were spare boards around.

Please remove the anchor from the link.

I assumed it was included on purpose to dodge HN's absurdly strict duplicate link policy, but based on a search this link hasn't been submitted before.

Yeah but then please include something like #banana or whatever that doesn't cause scrolling to the bottom ;)

I've changed the anchor in the article page, so the link anchor will now be ignored.

HN is amazing.

Well that works too.

Love your website, by the way.

Reply on news.ycombinator.com