Early Mainframe Information/Schematics/Dimensions: IBM 727 tape unit

[Garonious]

Greetings from Sydney
I have had a very long term project, which I’m researching, and that is to build an early mainframe computer. I’d love to make it all valve, especially as most of my experience in electronics design/development/construction has been valve based, like:
https://www.ludd.ltu.se/~ragge/vtc/

But for practical and financial reasons, will probably employ a hybrid system, valve/diode/germanium-transistor, and am considering systems developed between about 1954 and 1959.

At this point I’m concentrating primarily on the reel-to-reel magnetic tape unit, and am considering the IBM 727, IBM 729, and the Burroughs Model 544 DataReader. Although I’d love to build the latter, the former two appear to have a lot more documented information available, as do the related mainframes that employed such tape drives. Also 7-track (or 8-track) heads are available to purchase, whereas 12-track ¾” heads are not so common, not to mention the greater complexity/expense of a 12-track system. My build would have simplified mechanics, as it doesn’t have to operate usefully in the real-world.

I’ve perused the documents available on bitsaver, but wondered if there were any other sources, including recommended books, especially for the IBM 727?

There appear to be at least two versions of the IBM 727, an earlier chunkier looking version, with 2 sets of 2 push-button switches, and a later looking version, with 2 sets of 4 push-button switches. Does anyone know any details in this regard? Would any know the dimensions of the earlier looking version (I have 69 1/2” x 28 15/16” x 27 9/16” for the later)? Would anyone have a higher-resolution image of this version? Any personal stories about the IBM 727?

Please alert any members you consider appropriate

Kind regards, Gary

[Chuck(G)]

The CHM has a YT video of the internal workings of (IIRC) a 729 that they've restored on a 1401. The people working on that project probably have some very useful information.

If you want something a bit simpler, consider the Bendix drives.

[KC9UDX]

Valve based: binary, decimal, or other?

[Chuck(G)]

I think it'd be a nice long-term project to duplicate a Univac Solid State. Not really valve or transistor, but rather core.

[Garonious]

Thank you, Chuck, I'll have a look at that, and ask some questions. The manuals for the 729 contain a lot of detail, and so I may go down that path. I'd limit the mechanics to fixed heads with single speed play and rewind; I may attempt the vacuum columns. With only the one tape speed this should be straight forward.
I'd love to do a Univac, but I'm not sure if there is much in the way of schematics available. Was there a particular model you had in mind?

I guess some of these early system use quite a blend of magnetic core logic and memory, various delay lines, relays, and so on, making them quite interesting. I like the fact that a lot of this technology can be made at home, with relatively inexpensive equipment.

At this point I had only really considered a binary based system, something like an IBM 701 (1953 - valve, 36 bit word), or an IBM 1401 (1959 - transistor, serial based computation). Items like the system-clock could be valve based. I'd be aiming at modularity, so that each module can be developed, built, and tested as small project.

[KC9UDX]

36 bits is kind it ambitious.

[Chuck(G)]

If you've got a thing for glowing bottles, how about the B205. It could use tape drives. I'm not certain about smaller machines, like the LGP30. The CDC 160A, which is on the small side, could also handle tape drives.

Early vacuum column tape drives were pretty massive affairs, usually massing about a half-ton or more.

[FPM-III]

I wonder what the electric bill will be to power and cool a vacuum tube computer.

[KC9UDX]

I wonder what the electric bill will be to power and cool a vacuum tube computer.

How big is it? There is much room for scalability.

[Chuck(G)]

I'd love to do a Univac, but I'm not sure if there is much in the way of schematics available. Was there a particular model you had in mind?

Either the Solid State 80 or 90; they differed only in punched card format (IBM vs. Univac). There's plenty on bitsavers:

http://bitsavers.informatik.uni-stut...df/univac/uss/

The problem with most early machines is that drum was used for main memory (e.g. IBM 650). Core was new and very expensive.

I don't recall that the SS80 or 90 could take a tape drive, however.

Another interesting logic family was Goto's Parametron; used on the NEC 1100 series computers