IBM 1401 from 1959

[Homer J Simpson]

wiki;

https://en.wikipedia.org/wiki/IBM_1401

[Lightages]

That was a fun watch, but I would have liked to have seen some specs in the video to see what it could do back then.

[EEVblog]

Great video.
Loved the smoke'n'mirror demo!
Would love to hear the full interviews with those old design guys.

[Richard Crowley]

In the late 1960s, we had an IBM 1401 and an IBM 1620 at Orange Coast College where I got my 2-year degree (before going on to uni).
The 1401 was an 8-bit, BCD machine (CBA8421M) and ours had 4K of magnetic core memory.
(This was the BCD era, and "4K" actually means 4000, not 4096)

We had 4 of those tape drives, and toward the end of its life-cycle, they added 2~3 IBM 1311 disk drives.
The disk "packs" were removable stacks of six platters, ~14 inches in diameter, and weighing ~10 pounds.
But they would hold a whopping 10MB!  More than a whole room full of tape drives. And instantaneous random-access!

It had the card reader/punch for 80-column Hollerith cards which we used essentially only for input.
Output in an educational setting was mostly hard-copy, and it was rare to punch cards for output.

We had a 1403 chain printer which was one of IBMs most popular printers ever made.
It had a life cycle well beyond the 1402 and 1620, and well into the 360 era and beyond.

The 1402 was aimed more at the business market, and they created the 1620 for scientific computing.
The 1620 used the same card reader/punch for I/O, and also used the 1311 disk drives.

In my first job, I ended up supporting the 1620 (at a different school) because there was nobody left at the local IBM office that knew how to support them
I designed and installed a Centronics parallel print port for the 1620.
We connected it to a drum printer which had all the characters in 120 rings around the drum, and the wide ribbon between the drum and the paper.
There were 120 solenoid-fired "hammers" which pushed the paper (and ribbon) against the drum at the exact instant the desired character came by.
The 1403 chain/train printer and our off-brand drum printer were noisy beasts, but they printed paper REALLY fast.

https://en.wikipedia.org/wiki/IBM_1401
https://en.wikipedia.org/wiki/IBM_1403
https://en.wikipedia.org/wiki/IBM_1620
https://en.wikipedia.org/wiki/History_of_IBM_magnetic_disk_drives#IBM_1311

Several interesting stories to tell about those "good old days".

[Don Hills]

Ken Shirriff has done several blog posts about the 1401, including using it to mine bitcoin:

http://www.righto.com/search/label/ibm1401

[TimFox]

In my first three years at college, we also had a 1620 for students and a 1401 in the business office.  They replaced the student computer with a PDP-series unit in my senior year.
I remember that the card-punch units we had used 25L6 tubes to drive the solenoids.
This allowed four heaters in series across the AC mains.
Many years later, I needed two 25L6s for spares on an E H Scott WWII maritime-marine short-wave receiver, and the unitss I obtained surplus had IBM markings.

[retrolefty]

In my first job, I ended up supporting the 1620 (at a different school) because there was nobody left at the local IBM office that knew how to support them
I designed and installed a Centronics parallel print port for the 1620.
We connected it to a drum printer which had all the characters in 120 rings around the drum, and the wide ribbon between the drum and the paper.
There were 120 solenoid-fired "hammers" which pushed the paper (and ribbon) against the drum at the exact instant the desired character came by.
The 1403 chain/train printer and our off-brand drum printer were noisy beasts, but they printed paper REALLY fast.

 I was a field service engineer in the 70s for a mini-computer company. I had to work on 132 column drum printers, Data Products was one manufacture I recall, the drum was engraved in Switzerland and it alone weighed about 60 pounds. The printer cost more then the home I owned at the time, around $45K. As said there were 132 magnetic hammer drivers that would fire when that columns desired letter passed by the hammer.

As stated in the video (or other related video, I don't recall) the main problem with drum printers was wavy printed lines as hammer alignment deteriorated over time.  The human eye is very sensitive to straight line patterns and wavy characters showed up pretty easily. The adjustment (done quarterly even if no customer complaints) involved a diagnostic test that would fire all 132 characters at the same time (same character, say all A) and on a 1200 line per min rate would sound like a 50 Cal machine gun and would dim the overhead lighting at every firing in this alignment test. Then a 2 channel scope was hooked up to hammer  position one for channel one and the second channel was moved one hammer at a time for hammers 2-132. One looked for a small counter emf pulse (hump really) from the hammer coil and looked at it's position relative to the hump on the channel 1 hammer. If they differed in time each hammer had a plastic adjustable stop position screw that one used a Allen wrench to make it match. That was one ugly, noisy, and time consuming procedure that one learned to hate after just one attempt. The paper moved so fast when performing form feeds to the printer that static electricity would build up on the paper and could cause intermittent problems like 'falling off line" or fault status. So they made the paper pass by metal foil whiskers to bleed off static charges.

 It's not so much how well those generation printers worked, but rather like the bumble bee it's amazing that they worked at all.     

[Homer J Simpson]

Here is another one with more detail on the 1401.

[Richard Crowley]

I was a field service engineer in the 70s for a mini-computer company. I had to work on 132 column drum printers, Data Products was one manufacture I recall, the drum was engraved in Switzerland and it alone weighed about 60 pounds. The printer cost more then the home I owned at the time, around $45K. As said there were 132 magnetic hammer drivers that would fire when that columns desired letter passed by the hammer.

As stated in the video (or other related video, I don't recall) the main problem with drum printers was wavy printed lines as hammer alignment deteriorated over time.  The human eye is very sensitive to straight line patterns and wavy characters showed up pretty easily. The adjustment (done quarterly even if no customer complaints) involved a diagnostic test that would fire all 132 characters at the same time (same character, say all A) and on a 1200 line per min rate would sound like a 50 Cal machine gun and would dim the overhead lighting at every firing in this alignment test.

Yeah, I do believe ours was a Data Products drum printer. I still remember setting the seven mini toggle switches to the ASCII dash (0101101) so it would print whole lines of dashes (easiest to align into a straight line).  And yes, it sounded like a machine-gun in full auto blasting mode.  Every month I would get my hex key and adjust the print hammers to make straight lines.  It wasn't stable enough to last a whole quarter between adjustments.  It was rather a budget printer of modest capabilities.

[miguelvp]

I used to work in the computer lab and one of our tasks was to clean the drum printer with a metal brush and some kind of solvent.

Since we had DEC equipment (and newer than that 1401 or even the system 36) we had a DEC LP20B 132 columns, apparently it could do 240 lines per minute.

Don't recall ever hearing that the hammers needed to be adjusted.

[AF6LJ]

Good Stuff.
You use to be able to buy those IBM logic cards surplus, as a matter of fact Radio Shack sold a 1LB box of parts and screws and you count on there being one or two of those cards in it.

[matseng]

I've never seen a line- or drumprinter operate in real life.  The closest was a high speed matrix printer back in the beginning of the 80's that had a printhead assembly wide enough to cover the entire 132 column landscape paper and had multiple full columns of needles approximately every centimetre along the full with.  So when printing the head only needed to move just 1 cm to print en entire row of characters.

From what I remember the head moved quite rapidly and printed maybe 15-20 full lines per second or something thereabouts.  It was a noisy beast that actually had to spin up for a couple of seconds before it started to print.
 

[Stonent]

"Leaky transistors" ?

Do you think they meant capacitors, or transistors that had a BCE leak?

[Stonent]

Just an aside, the System 360 that they mentioned that was the replacement for the 1401 is still living on today in the System Z mainframe.  System Z is fully backwards compatible with the 1960s era S/360.

[Brumby]

I entered the computer industry when the S/370 had been around for a few years, so I missed out on a lot of what is being discussed here... but there was an 083 card sorter on site that was used regularly, including by myself.

But I do remember hearing that the 1403 printer had this "feature" that the cover would automatically open when it ran out of paper - and if some poor programmer had left his object deck on top of it, he would be playing a game of 2000 card pickup.

https://en.wikipedia.org/wiki/IBM_1403

These were noisy machines, especially when the cover was raised. Some people were able to create text that used the timing of the print hammers to generate desired frequencies and actually play music when that text was printed.

The machine I worked on was a 370/115 which had a 3203 printer - with the same 5 copies of the character set on a print train.  During my time as an operator I worked this out for myself and put together two tunes - "Whistle while you work" and "We wish you a merry Christmas".  One day the Field Engineer was around and I played them for him - he just cracked up, which was unusual.  Roger never smiled.  I also gave a demonstration to the "EDP Manager" who said "Very clever.  Don't do any more."

[Richard Crowley]

But I do remember hearing that the 1403 printer had this "feature" that the cover would automatically open when it ran out of paper - and if some poor programmer had left his object deck on top of it, he would be playing a game of 2000 card pickup.

That is why the last 5 columns of the card (or the first 5 for FORTRAN) were dedicated to the sequence number so that if the deck was dropped, they could be re-sorted into the correct order using an 083 card sorting machine. 

Those card sorters were the kind of visually-interesting bits of "computing" equipment that movies and TV used to represent a "computer".

The 026 and 029 key-punch machines had an automatic function where they would punch sequence numbers into your cards as you created them.

The machine I worked on was a 370/115 which had a 3203 printer - with the same 5 copies of the character set on a print train.  During my time as an operator I worked this out for myself and put together two tunes - "Whistle while you work" and "We wish you a merry Christmas".  One day the Field Engineer was around and I played them for him - he just cracked up, which was unusual.  Roger never smiled.  I also gave a demonstration to the "EDP Manager" who said "Very clever.  Don't do any more."

Yes, we had several files that would produce different tunes. I believe one of them was "Stars and Stripes Forever".
We also had files that would produce "printer art".  Often called "ASCII art", but on an IBM printer, it was a misnomer because IBM mainframes were (and continue to be?) EBCDIC rather than ASCII like the rest of the industry.

[Brumby]

In my experimenting with the 3203, I tried various character combinations to build up a "scale".  To get the higher frequencies, you would try to get the hammers striking with shorter intervals - fairly obviously - but in my efforts, I came up with a string of characters on one line that the 3203 printer simply could not print.  There was nothing weird involved - it was simply printing characters that were found on the print train - and there were 5 copies of the character set for it to use, but it could not do it.  Every time I tried I got a "Machine check" and the printer would just stop.

I'm pretty sure the problem was the power supply for the hammers.

The character string caused 44 hammers to fire at the same time (columns 1,4,7,10,.....,130) and then - after the print train travelled half a column width - another 44 hammers fired simultaneously (columns 2,5,8,11,...,131).  After a further half column travel by the print train, the last 44 hammers fired (columns 3,6,9,12,...,132).  It would seem logical that such a high demand over such a short period would tax the power supply - but that's only my theory.  I didn't make my discovery known.  Until now. 

[Richard Crowley]

"Leaky transistors" ?
Do you think they meant capacitors, or transistors that had a BCE leak?

That was back in the germanium (or maybe the very early silicon) era.
So it seems quite possible that the power devices weren't anywhere near as good as we expect today.

[T3sl4co1l]

"Leaky transistors" ?
Do you think they meant capacitors, or transistors that had a BCE leak?

That was back in the germanium (or maybe the very early silicon) era.
So it seems quite possible that the power devices weren't anywhere near as good as we expect today.

Yeah, they mentioned germanium.

Germanium transistors have two states: leaky and on.  In fact, it's quite normal to bias a small signal Ge transistor with negative base current, i.e., hFE is so high it wraps through infinity and becomes negative.  A conventional common emitter amplifier circuit, with the base voltage divider, actually has to sink a little current from the base, and the base and emitter voltages float up slightly from the voltage divider's setting.

While this IS IBM we're talking about, it's also a budget machine.  They would've used nothing but the cheapest floor sweepings, which wouldn't have even been all that cheap in 1959, even with a purchase of a few million over the production run of that computer.  (Imagine that today: the quantity pricing you would get on a literal pallet of SOT-23s would be pretty wonderful -- not that you'd be going out and buying that pallet with pocket change or anything.)  A typical Ge transistor that size might switch in fractional microseconds, hence the modest clock speed (110kHz was it?).

Tim

[JoeO]

In my experimenting with the 3203, I tried various character combinations to build up a "scale".  To get the higher frequencies, you would try to get the hammers striking with shorter intervals - fairly obviously - but in my efforts, I came up with a string of characters on one line that the 3203 printer simply could not print.  There was nothing weird involved - it was simply printing characters that were found on the print train - and there were 5 copies of the character set for it to use, but it could not do it.  Every time I tried I got a "Machine check" and the printer would just stop.

I'm pretty sure the problem was the power supply for the hammers.

The character string caused 44 hammers to fire at the same time (columns 1,4,7,10,.....,130) and then - after the print train travelled half a column width - another 44 hammers fired simultaneously (columns 2,5,8,11,...,131).  After a further half column travel by the print train, the last 44 hammers fired (columns 3,6,9,12,...,132).  It would seem logical that such a high demand over such a short period would tax the power supply - but that's only my theory.  I didn't make my discovery known.  Until now. 

There was a "Chain Break Pattern".  A certain pattern of characters, if printed, would cause the chain to break.

[John_ITIC]

Gotta love old technology; especially the "leaky transistors" mentioned at 6:53 

[Brumby]

In my experimenting with the 3203, I tried various character combinations to build up a "scale".  To get the higher frequencies, you would try to get the hammers striking with shorter intervals - fairly obviously - but in my efforts, I came up with a string of characters on one line that the 3203 printer simply could not print.  There was nothing weird involved - it was simply printing characters that were found on the print train - and there were 5 copies of the character set for it to use, but it could not do it.  Every time I tried I got a "Machine check" and the printer would just stop.

I'm pretty sure the problem was the power supply for the hammers.

The character string caused 44 hammers to fire at the same time (columns 1,4,7,10,.....,130) and then - after the print train travelled half a column width - another 44 hammers fired simultaneously (columns 2,5,8,11,...,131).  After a further half column travel by the print train, the last 44 hammers fired (columns 3,6,9,12,...,132).  It would seem logical that such a high demand over such a short period would tax the power supply - but that's only my theory.  I didn't make my discovery known.  Until now. 

There was a "Chain Break Pattern".  A certain pattern of characters, if printed, would cause the chain to break.

I'm glad I didn't find THAT one!!

[Stray Electron]

The machine I worked on was a 370/115 which had a 3203 printer - with the same 5 copies of the character set on a print train. 

   Those were chain printers. I was a Field Engineer for Burroughs Corp and we used ones made by ODEC (Ocean Data Electronics Corp).  IIRC the chains held 180 characters total and the number of character sets that you could have on a single chain depended on how many characters you choose to have in each set. IIRC most of our customers used character sets of about 44 characters so could have four sets per chain. Each revolution of the chain generated one timing pulse per revolution and then the printer would "count" as each character slug went past a known position and fire the correct hammer as the appropriate slug past under it. all done in hardware logic!  The problem was that the slugs clipped onto a rubber belt (the chain) and the clips would break and the slug would fly off. From then on, the entire character set and all of the following character sets would shift by one character!  I got REAL good at looking at printouts and recognizing what character was missing and which set it was in! I could usually fix one in a matter of minutes but frequently had to drive 150 miles to the site since I covered a good bit of two states. The interesting thing about our ODEC printers were that they were originally built (I THINK) for IBM since everyone of them had a box mounted on the side that converted ASCII to EBCDIC.  We had models that were rated at 150, 200 and 300 characters per second. IIRC the character set that most of out customers used was the twenty-six character alphabet, 0 through 9 and eight special characters. But as a FE I had to always carry multiple sets of those AND all other character sets including APL.

[Don Hills]

Early model IBM 1403s used a print "chain", with type slugs clamped to a metal wire belt.
Later 1403, and 3203, printers used print trains, not print chains.
A print train consisted of an oval "racetrack" rail. Each print slug had a slot in the back that fitted over the rail. One end of the rail was cut away for a large gear wheel. This engaged with teeth on the back of each slug, pushing them around the track. The gear wheel position was adjustable to set the inter-slug spacing, which was critical - the total inter-slug spacing after taking up all the "slack" was only a few thousandths of an inch. Setting the chain oiler rate was critical. Too much oil, as little as 2 or 3 drops, would cause the train to seize solid due to the oil forcing the slugs apart. Too little oil would cause rapid wear of the rail and slugs, increasing the inter-slug gap. Too much inter-slug gap would allow a slug's teeth to overlap the teeth on the gear wheel, with catastrophic results - "train crash".

[Richard Crowley]

The interesting thing about our ODEC printers were that they were originally built (I THINK) for IBM since everyone of them had a box mounted on the side that converted ASCII to EBCDIC.  We had models that were rated at 150, 200 and 300 characters per second. IIRC the character set that most of out customers used was the twenty-six character alphabet, 0 through 9 and eight special characters. But as a FE I had to always carry multiple sets of those AND all other character sets including APL.

Wow, I didn't think that anybody remembered APL.  And certainly not outside IBM-land.  We had a room full of terminals based on the IBM Selectric typewriter with APL character-set type-balls on them.  We lab-assistants could show-off taking any program in another language and write it in 1/10 the number of lines in APL.  Of course the code was practically unreadable, even to the person who wrote it(!)

Early model IBM 1403s used a print "chain", with type slugs clamped to a metal wire belt.
Later 1403, and 3203, printers used print trains, not print chains.

My recollection was that the lower-speed printers used chains, while the higher-speed models used the train technology.