tag:blogger.com,1999:blog-6264947694886887540.post1572943252397750973..comments2024-03-27T22:51:35.227-07:00Comments on Ken Shirriff's blog: The printer that wouldn't print: Fixing an IBM 1401 mainframe from the 1960sKen Shirriffhttp://www.blogger.com/profile/08097301407311055124noreply@blogger.comBlogger19125tag:blogger.com,1999:blog-6264947694886887540.post-6968191311220892592020-09-05T22:25:53.975-07:002020-09-05T22:25:53.975-07:00Back in the late 1970's, 1980'S I was a p...Back in the late 1970's, 1980'S I was a partner in a computer leasing company. we had about 50 systems's on lease in the nyc area and 500 keypunch machines.<br />(2) Ibm 1440's, (2) RCA 301's(transistor computers), (10) IBM 360 MODEL 30 AND 40'S<br />(10) dEC PHP 11'S and pdp 8's, (3) system 3's, (20) IBM 370's all on lease.<br /><br />I was lead Technician/engineer. It usually took me 2hrs to fix an IBM computer 360/370 MAINFRAME vs 4 to 8hrs for an RCA or borroughs system!! THE IBM DOCUMENTATION WAS GREAT,<br />THE TROUBLE SHOOTING GUIDES WERE CALLED MAPS!! <br /><br />THE EASIEST PART OF FIXING AN IBM MAINFRAME USUALLY CONSISTED OF PULLING A CARD<br />AND CLEANING THE CONTACTS AND RESEATING THE CARD!!! IBM COMPUTERS USUALLY<br />HAD CONTACT PROBLEMS!!!!!! IBM FINALLY SOLVED THE CONTACT PROBLEM IN THE PC<br />BY SCREWING ALL THE CARDS INTO THEIR SOCKETS!!! IT ONLY TOOK IBM 20 YEARS TO COME UP<br />WITH THAT SOLUTION!!!!! SCREW DOWN THE CARDS!!!!!!<br />FOR CHECKING GATES AND SIGNALS ON ALL COMPUTER I USED A LOGIC PROB(RADIO SHACK)<br />OR JUST A VOLT METER!!<br /><br />DEC PDP 11'S AND PDP 8'S USUALLY REQUIRED REPLACEMENT BOARD, USUALLY TOOK 1/2 HR<br />TO TROUBLE SHOOT AND REPLACE A BOARD!!!<br /><br />TROUBLE SHOOTING: ALL IBM COMPUTERS HAD SEVERAL CARDS OF THE SAME TYPE IN EACH MACHINE<br />SO YOU WOULD JUST SWAP THE SAME TYPE OF CARD FROM ANOTHER SECTION OF THE MACHINE AND <br />SEE IF THE TROUBLE DISAPPEARED!!! WE ALMOST NEVER HAD TO ORDER PARTS FROM IBM!!<br /><br /><br />IBM WAS SLOW ON PARTS DELIVERY TO LEASING CO'S TO SABOTAGE OUR MAINTENANCE OPERATION,<br />IBM CLAIMED WE HAD TO BUY PARTS ONLY FROM A PLACE IN INDIANA, THAT HAD NO AIR FREIGHT<br />SERVICE, ONLY GROUND THAT TOOK 1 WEEK FOR DELIVERY!! <br />SO I PURCHASED PARTS FROM RECYCLING CO'S, THEY WOULD SELL US<br />$400 IBM CARDS FOR $5 PER CARD( I BAUGHT 500 CARDS FROM JUNK 360 AND 370 THAT WE NEVER<br />HAD TO USE!!!<br /><br />I WOULD USE AN OSILISCPE ONLY FOR REPLACING DISK DRIVE HEADS OTHERWISE I USED MY<br />LOGIC PROBE(RADIO SHACK) OR VOLT METER TO TRACE SIGNALS!! <br />I MUST HAVE REPLACED 100 TRANSISTORS ON THE RCA 301, I NEVER HAD TO REPLACE ANY TRANSISTORS ON THE IBM 1440.'S<br /><br />AT THE NY TIMES, I WORKED FOR A CONSULTING COMPANY THAT HAD TO KEEP THE TIMES CLASSIFIED COMPUTERS GOING, AFTER THE MFG WENT OUT OF BUSINESS!!<br /><br />WE HAD A POWER FAIL SHORT OUT 2 OF THE 3 SYSTEMS THAT THE TIMES USED TO ENTER<br />CLASSIFIEDS( CAUSED BY A TECHNICIAN MORON, THAT MY BOSS HIRED).<br /><br />MY BOSS AND 3 TECHNICIANS WORKED ON THE SYSTEMS FOR 3 DAYS( TRACING SIGNALS)<br />THEY COULD NOT FIX THE SYSTEMS.<br />FINALLY HE ASK ME TO TAKE A LOOK AT THE SYSTEMS! USING MY IBM MAINFRAME EXPERIENCE<br />AND METHODOLOGY(PULLING CARDS), IT TOOK ME 15 MINUTES TO FIX THE SYSTEMS!!!<br /><br />IF YOU WORKED ON IBM MAINFRAMES WITH THOUSANDS OF CARDS THEN A NY TIMES COMPUTER<br />WITH 100 CARDS IS NOTHING!!<br /><br />TODAY WORKING ON SERVERS AND PC, ALL THE PROBLEMS ARE SOFTWARE AND TAKE LONGER<br />TO FIX THAN AN IBMNFRAME!!EDhttp://aaacomputer.homestead.comnoreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-76511313301843384412018-12-07T06:08:52.181-08:002018-12-07T06:08:52.181-08:00This comment has been removed by the author.Alia parkerhttps://www.blogger.com/profile/02326303770549335930noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-89721296450474931452018-11-12T00:35:50.523-08:002018-11-12T00:35:50.523-08:00Excellent diagnosis of a fairly difficult problem ...Excellent diagnosis of a fairly difficult problem (would have been a real challenge if the problem would have been intermittent). I am really impressed the way you have researched the print buffer logic mostly by reverse engineering. I really wish I was there with you to work on this problem.<br /><br />There is DFT called 'print word mark' which we found to very useful in diagnosing and isolating intermittent PC due to a particular position problem (driver card/hammer coil/cable connection etc.)<br /><br />Like always, your posts are really enlightening.bmhttps://www.blogger.com/profile/18254028897901647688noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-37032137495191740852018-09-28T05:45:53.510-07:002018-09-28T05:45:53.510-07:00Interesting article.
I'm pretty sure I used on...Interesting article.<br />I'm pretty sure I used one of these printers (at Rutherford Lab in the UK) in the late 1980's - if not exactly the same model certainly the same family. And boy was it LOUD. The speed of scrolling blank lines or form feeds was impressive.<br /><br />I had no idea it was so old.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-38442709801741405992018-09-22T22:49:24.101-07:002018-09-22T22:49:24.101-07:00Rhialto: that's a good question, how the print...Rhialto: that's a good question, how the print controller steps the memory by 3 and the characters by 2. The short answer is it's complicated and uses a bunch of counters and ring counters. (A ring counter has one bit high at a time, like a shift register.)<br /><br />The first counter is the PSS (Print SubScan) Ring counter, which counts the three subscans in a scan. It is synchronized with the print chain via a magnetic sensor in the printer.<br /><br />Next, the PSS Counter keeps track of the first character in the subscan. It counts through the 48 characters (1 through 12 with two zone bits). Its counting is a bit unusual: it increments after the first and second subscans and decrements after the third subscan, so each scan is one character later. E.g. A, B, C for the first scan, then down to B, C, D for the next scan. The Compare Counter keeps track of the current character on the chain. It loads the value from the PSS Counter at the start of the subscan and then counts by 2.<br /><br />If there is no print buffer, and characters are printed from core memory: the B address register (B star) points to the character in core memory. It is incremented by 3 via the ALU each step. With a print buffer, the memory addressing is entirely different since it uses the print buffer's separate core. The units ring counter goes from 0 to 9 and the tens ring counter goes from 0 to 13. These signals go directly to the X and Y select lines for the buffer core memory. The ring counters increment by 3, which is not too hard with a ring counter.<br /><br />Schematics for the print circuitry is in the <a href="http://ibm-1401.info/ILDs_Aug62-Enhanced-TOC.pdf" rel="nofollow">Instructional Logic Diagrams</a> starting at sheet 70. A detailed explanation is in <a href="http://bitsavers.informatik.uni-stuttgart.de/pdf/ibm/1401/ce/R25-1496_1401_Instruction_Logic_Nov60.pdf" rel="nofollow">1401 Instruction Logic</a> pages 88-94.Ken Shirriffhttps://www.blogger.com/profile/08097301407311055124noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-47888014293576609262018-09-22T06:03:30.832-07:002018-09-22T06:03:30.832-07:00The animation is great to clarify how the type slu...The animation is great to clarify how the type slugs line up with the hammers. But what I remain wondering about is how the printer knows which character is present on the slug that is currently lined up to a hammer. (It knows which character to print because it reads from core.) <br />My thought would be that it uses a counter that resets when it has counted all printable characters. But your animation shows that it would need to count up by 3, instead of the simpler counting by 1. <br />So then I thought that they would re-arrange the order of the characters on the chain, to enable counting by 1. The chain would contain something like A . . B . . C (with some other characters in between, to be used in the next 2 subscans). But speaking against this is the photo you show of the chain, which seems in normal order.<br />The only other thing I can think of is that they renumbered their characters in a different order than the print chain. But that would surely be really impractical in software.<br />What is the solution here? I'm very curious.Rhialtohttp://www.falu.nl/~rhialtonoreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-5435381106114727862018-09-21T11:32:15.253-07:002018-09-21T11:32:15.253-07:00When I first started working for IBM in the late &...When I first started working for IBM in the late '70's we used a 1401 with a homemade fixture to program EPROMs. I believe they were Intel 1701's and 1702A's...256 bytes of storage. And it took upwards of 45 minutes to program, if I remember correctly.Wally Hollowaynoreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-64223276852374004362018-09-20T22:43:23.585-07:002018-09-20T22:43:23.585-07:00@Dithermaster:
I've met a couple people who c...@Dithermaster:<br /><br />I've met a couple people who claimed to have encountered a case where a student figured out the character order on a chain printer and sent the 'magic string'. The described result was having to replace the chain and that they were finding links behind various things for over a year.Chakat Firepawhttps://www.blogger.com/profile/02002352879959672258noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-71192157591428492702018-09-20T20:40:20.382-07:002018-09-20T20:40:20.382-07:00Danjovic: I asked some old IBM engineers and they ...Danjovic: I asked some old IBM engineers and they said engineers generally wouldn't have many spare cards (if any) due to the large number of different card types and the high cost of keeping a large inventory. Instead, engineers would usually call the nearest branch office and have someone bring the cards they needed. They would drive halfway and meet up with the person delivering the cards. If the branch office didn't have the parts, they would be sent from the area-level parts depot. For very unusual parts (e.g. a new core array), parts would be flown out from the parts depot in Mechinsburg. Apparently they'd buy two tickets: one for the person delivering the core array and a second seat for the core.Ken Shirriffhttps://www.blogger.com/profile/08097301407311055124noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-18902511974876581122018-09-20T20:31:06.543-07:002018-09-20T20:31:06.543-07:00Excellent reading! Thanks for the article.
Solid ...Excellent reading! Thanks for the article.<br /><br />Solid State components should be very very expensive in the 60s but the cost of non-working hours for such machines should be far higher.<br />Then I think that IBM maintenance people carried a bunch of spare cards on their toolboxes, and the troubleshooting was performed by replacing suspect cards while tracking the problem along with the ALD schematics. The cards identified as faulty could be serviced later at the factory using test sets for each model of card.<br /> <br />Today you don't have spare cards, then using a 4 channel digital storage scope is totally fair ! Danjovichttps://www.blogger.com/profile/15384835800874451538noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-72406625499101811122018-09-20T12:17:29.292-07:002018-09-20T12:17:29.292-07:00Microgadgethacker: don't worry, we ended up fi...Microgadgethacker: don't worry, we ended up fixing the bad card with a replacement transistor. So we didn't cannibalize any of the spare cards.Ken Shirriffhttps://www.blogger.com/profile/08097301407311055124noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-82534883986733159332018-09-20T10:22:09.334-07:002018-09-20T10:22:09.334-07:00It's absolutely fascinating that this machine ...It's absolutely fascinating that this machine verified hammer actuation of the particular character being printed. That's an incredible level of fault checking. I worked on fault tolerant computers where a dual core system was really two sets of three CPUs running in lockstep and having hardware check each bus cycle for disagreement. The memory and storage system were ECC protected, as were network communications, but that level of protection certainly was not extended to laser printers! I also appreciate the fact that these vintage machines can be diagnosed and repaired with fairly available parts - even if you had to steal discrete components from spare cards. Microgadgethackernoreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-87368003967482074562018-09-18T13:04:10.083-07:002018-09-18T13:04:10.083-07:00I was the systems programmer for a real mainframe ...I was the systems programmer for a real mainframe the 7040 scientific computer back in 1965. It used a 1401 as it's card/reader/punch and printer. A little bit of commercial accounting was also run on the 1401. The 1410 which was a machine somewhat similar to the 1401 was more of a commercial main frame.<br /> gingorohttps://www.blogger.com/profile/11825691766111067082noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-49860695110579013922018-09-18T08:03:21.861-07:002018-09-18T08:03:21.861-07:00Have you tried turning it off and on? ;)Have you tried turning it off and on? ;)AdamKhttps://www.blogger.com/profile/11874902508171327481noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-7451321532318281892018-09-18T07:05:54.655-07:002018-09-18T07:05:54.655-07:00Wow! A blast from the past. I learned computer mai...Wow! A blast from the past. I learned computer maintenance using a Xerox Sigma 2 computer back in the 1970s. The computer was used in a Westinghouse Prodac-250 at a power station. I remember spending hours and hours wading through logic diagrams in large books to troubleshoot the computer. This sure brought back some of those (now fondly remembered) memories.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-80754651634334436232018-09-17T23:36:42.218-07:002018-09-17T23:36:42.218-07:00LOL, i came here with the same idea, how did they ...LOL, i came here with the same idea, how did they do it without that nice digital stuff :)<br /><br />https://www.oscilloscopemuseum.com/collection-tek.html lots of scopes but don't see any multi channel so a bit more work to find the problem<br />Franchttps://www.blogger.com/profile/17193016208795194087noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-80221239661462202102018-09-17T17:55:02.220-07:002018-09-17T17:55:02.220-07:00Dithermaster: back in the 1960s, field service had...Dithermaster: back in the 1960s, field service had analog scopes. It was easier with a modern scope (in particular, 4 channels made it much easier to figure out what was going on), but I'm sure they did fine with analog scopes.Ken Shirriffhttps://www.blogger.com/profile/08097301407311055124noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-87087547557682484172018-09-17T15:45:33.592-07:002018-09-17T15:45:33.592-07:00This was a very enjoyable article to read, better ...This was a very enjoyable article to read, better than some mystery novels!<br /><br />You used a nice modern storage scope. How much harder would this be with an analog scope? Is that what field service used?<br /><br />Regarding note 2 and the page with animation: The mechanism of scans and subscans not only matches electronic world to mechanical world, but also prevents things like someone like me trying to print a magic line of text the would fire every hammer at the same time, which would surely draw too much current or put too much drag on the system, and blow it up. Instead it spreads out the electrical and mechanical load, and thwarts such attempts. Brilliant!<br /><br />I'm convinced all of these ideas will be needed again when we move to nanotech.Dithermasterhttps://www.blogger.com/profile/07392377429250489728noreply@blogger.comtag:blogger.com,1999:blog-6264947694886887540.post-73224564518022714942018-09-17T14:24:26.637-07:002018-09-17T14:24:26.637-07:00It is so amazing to see how stuff was done at the ...It is so amazing to see how stuff was done at the beginning of electronic computation -- each bit, each gate was precious! (and it could be poked with your finger) I keep that in mind when I launch yet another scientific computation on 12000 cores. You are doing great work!Tüdelbüdelnoreply@blogger.com