IBM did contribute one HUGE advancement to the advancement of IC's once they were involved full swing. And that was Chemical Mechanical Planarization (CMP). This is also called Chemical Mechanical Polishing". But, to be pedantic, planarization (flat, not necessarily shiny) and polishing (shiny, not necessarily flat) aren't quite the same thing. CMP allowed a huge increase in transistor density. Before this, you got higher density by making the dies bigger and/or the transistors smaller. Now, you could stack many layers (up to maybe 10 nowadays). So it is sort of a cousin to being a multi-layer PC board. But the big problem is that at small sizes, the photolithography process step has virtually no depth of field. So the deposition of transistors and metal connects must be on a ludicrously flat surface. When you put a dialectric (insulating) layer over a metal layer, the surface will not be flat as it will somewhat conform to the features of the metal layer underneath. (sort of what a conformal coating over PC board traces looks like.) CMP makes it flat. Then holes are etched where needed to make connections to the layer underneath. Then the next active layer is deposited. And then the next dielectric layer, and so on. There are enormous technical difficulties in achieving good CMP, however. I worked at the company where IBM initially purchased some manual polishing machines (designed for another purpose). They were very secretive about what they were doing. This was in the late 1980's. Then they contracted with us to design and build a quantity of purpose built automated machines for this. Many, many others adopted CMP. For a number of years, the company I was at and one other sort of split the marketplace. Then others got into it, eventually including the "big kid on the block" Applied Materials. it is my understanding that IBM got over 1 billion in licensing fees from the CMP technology.
@@beautifulsmall That is a very good analogy. For some reason, many in the semiconductor industry have moved toward the word "polishing". And I don't know why. But every industry develops its own lexicon. Even if some words are technically incorrect. Such as on the Southern Pacific Railroad they called all their articulated locomotives "Mallets". A "Mallet" is a specific type of articulated engine with both high and low pressure cylinders. A "simple" articulated has all high pressure cylinders. That is what Southern Pacific actually had, but they still incorrectly called them "Mallets".
According to the history of IBM technology by Pugh et al there was a big battle internally between proponents of SLT and ICs. The designer of the SLT series got demoted because the new 360 range was slated in the press for not using ICs. However he was vindicated (and reinstated) a few years later when it was shown that the SLT packaged solutions had a reliability around 100 times greater than IC based systems. They adopted ICs later - built by themselves - when the technology could offer the same reliability.
Those SLTs are more commonly known as "Hybrids", in the industry, at large. They still exist in more specialized applications. Moore's law made the IC win, in the end. Though, in the mid 60s, no one knew that. When I first studied electronics, in high school, in the 70s, hybrids were a common sight, at surplus stores.
Hello Fran!! Yes SLT!! I worked at the East Fishkill N.Y. IBM plant for 31 years, I started in 1977 August and even then SLT was being produced in extremely large numbers! They were produced in BLDG 310!! Those film resistors were trimmed by a process called LASER ACTIVE TRIM! the device SLT was loaded into a computerized machine with the socket structure and the computer would select the correct pins to monitor THEN the laser would start to trim the selected resistor one at a time!! but quickly!! I worked in production control, and had to move the product from department to department and had to keep track of production loses and the type of loss that was incurred !!! Fran you re ignited some cool memories of a bygone era for me thanks!!
@@effexon These SLT devices were used in the IBM 360 and 370 systems, These were very large mainframe systems of the day back in the 60's and 70's. The particular unit shown by Fran was still produced in smaller quantity up until the late 1978 or 79 as replacement parts!
My dad worked for Raytheon. He spent a summer working on a join venture with IBM . It was a nice surprise to see naked chips of the IBM type in books on IBM 360 machines. Dad had a few he kept , including one made into a tie clip. He worked first for Raytheon's tube division then for it Missle division. I had a friend who worked for IBM. He had many card from their smaller systems, using the later versions of these chips, in their metal cans.
I learn almost as much from reading the comments under Fran's videos as I do from Fran herself! Thanks, Fran, for bringing such a wonderful community together.
SLT was discontinued in the late 1960s. There was an argument at IBM when the System 360 was in planning whether to go with hybrid, or fully integrated circuitry. In hindsight it was better to go with hybrid (SLT) at that time, fully integrated would have added at least a year to system 360 development time. The card you have your finger on at 0:29 is NOT an SLT card. If you look at the other side of the card you will see that the packages have 16 rather than 12 pins. There is an integrated circuit inside. It looks very similar to SLT, but it was called MST (Monolithic System Technology). The logic was in fact ECL rather than TTL. System 370, and even some system 360 models used MST. Pull the lid off one of those packages. System 360s did not use the grey card carrier that 370s did.
While some SLT modules (the name IBM used for the square metal packages) only had 12 pins others had 16. One other point was that it was possible during development to implement a SLT module with discrete components on a small board with the pins. Back in the mid-1990s I went to a talk by some of the people who designed the 360/40 CPU and they had some of their prototype boards containing a combination of normal SLT modules and a few made up from transistors, diodes and resistors.
@@jeremybarker7577 SLT (Solid Logic Technology) applied to the technology and packaging system of very small discreet components. Standard SLT packaging used 12 pins although some later versions did use 16. ULD (Unit Logic Device) was SLT miniaturized and packaged for the Apollo spacecraft. Fran displays a lovely example in this video. ULD was not packaged in the square metal packages. There were some later SLT developments that used 16 pins. ASLT had two stacked substrates and SLD (Solid Logic Dense) put resisters on the bottom of the substrate, like the modules for the Apollo. ASLT was used in the s/360 9x series of large computers. IBM refers to the units here as cards, cards are plugged into boards. Boards have no logic on them, but backplane wiring to connect cards with each other. Boards are mounted in gates, gates are mounted in frames. In IBMese none of the devices in this video are boards, they are cards. The System 360/40 was developed in the IBM plant in Hursley, England. The production versions used standard SLT, all devices had 12 pins. I have worked on many 40s. One unique thing with SLT was the method of attaching the active devices to the substrate. Conventional devices connected the devices with small wires. SLT put tiny solder coated copper balls on the substrate, then the device (transistor) was placed on top of the balls. The module was put into an oven to flow the solder. The technique proved highly reliable. When IBM started producing monolithic technology in about 1967 they used the same technique to attach the chips to the same square silicone chip used in SLT. Commercial ICs in DIP packages were connected with tiny wires, IBMs were attached with the little copper balls. This has another advantage that circuit paths can be printed under the chip. The IC version was called Monolithic System Technology (MST). The packages looked identical to SLT, but had 16 pins. The larger card that Fran shows is definitely an MST card, not SLT. Under the metal package will be 1 or more monolithic chips, and no resistors on the chip. I hope she takes the lit off of a package and shows us.
I have the little logic handbooks the engineers used for reference (SLT, ASLT, MST MST-4, etc.). Each is about 3 inches by 4 inches with pale blue, green, red covers
Those cards brought back some memories! When I was a kid my dad bought a load back from work to show me, I wish I still had them. Later at school we could send cards to the local college to run on an ICL 3900 minicomputer, I taught myself to punch cards using one of the 12-key punches that had a key for the hole in each row, and you punched in the EBCDIC code for the characters you wanted. I actually got quite fast at it and pretty accurate, luckily programs where usually less than 100 cards long! Waiting a week or so to find out if you'd punched the cards right was tension making too! They would come back in a bag with the contents printed along the top of the card, so I got in the habit of re-using cards that had standard statements on them so I didn't have to re-punch them, a very early example of code reuse!
Those green cards were "JOB" cards, the first card in a deck. They were green so the operators could separate the jobs easily on the output hopper of the card-reader, wrap elastic bands around them, match them with the job-output from the printer and give the job-decks and output back to the programmers, or, for production jobs, put the card-decks back to run the next night or week or month.
I still work in the mainframe arena. Still Going strong. Biggest problem is a lot of the code is old and those that knew it have retired or passed away. Still a very relevant technology to this day as it is a very robust both hardware and OS wise. As database servers there is little to beat them. As for the punch cards, many large scale printers use them to this very day as measurements for gaps etc are done in IBM punch cards. As such, they are still made. Gaps between the laser and the drum, developer rollers, sensors etc all measured in x punch cards.
The banks cannot afford to loose track - therefore it will probably be a century long maintenance contract for some banks with IBM before mainframes are replaced with something... Maybe cryptos (probably not the traditional Bitcoin or a bit newer Ethereum type though) finally get to sunset all the COBOL-architectures.
My uncle worked at the IBM plant in Poughkeepsie for many years where they designed and made many of those SLT packages. At one point, during the late 60's or early 70's, he arranged a tour for me and my father. I remember thinking at the time that those SLT packages were already a bit obsolete, I had already seen TI DIP packages in other products. Each SLT "can" contained semiconductor components which consisted of a tiny rectangle of silicon cut from a grown wafer. Each silicon rectangle had either two (diode) or three (transistor) very tiny copper or gold spheres bonded to it, and then the leads were soldered to the spheres creating the circuits. By current standards, very low-density stuff, but rugged. The fabrication process was all internally developed by IBM, they were almost fanatical about doing everything in-house. My uncle got an award for the helping develop the process they used to align the little silicon chips properly in the fab, it was an amazing piece of Rube-Goldberg engineering, but it worked. On a different note, those boards in the 360/370 systems may have been reliable, but the back-plane wiring was a nightmare. The interconnect circuitry used something called "try leads", which gave endless trouble and were very difficult to diagnose and repair. At one point, when I was an operator in a large mainframe shop we spent the better part of a shift waiting for the service tech (CE) to find a bad lead on a backplane and get the system running again.
I first saw these in the early 80's in a surplus IBM "data adapter unit" which is a VERY large UART essentially. I still have one of the small modular plugin circuits. Amazing technology - developed originally for the NSA according to their published history at fairly significant expense.
I worked at an IBM manufacturing plant in Charlotte, NC for 11.5 years. This video brings back memories of the 11 years of my career there as a test technician working on theses types of boards. I did incircuit and final test. When the boards failed we did component level repair too. They started switching to SMT boards and components in the late 80's. It saddens me to know the entire IBM Charlotte facility closed and this type of manufacturing was sent to the Pacific rim putting about 6,000 of us out of work. Thanks for the walk down memory lane nonetheless!
11:28 A tip: When desoldering through-hole components it helps to hold the board with the pins facing down on top of the iron held vertically with the tip facing up. That way gravity assists molten solder to flow down from the board to the bottom of the desoldering filter capsule.
These regular grid style boards surely made things easier in production but especially in the design system. So IBM used them even with DIP packages until wire density forced them to change. Also interesting, the flipped and soldered transistor dies have been replaced by ICs that are connected via bonding wires. Today, the "flip-chip package" ist default again.
IBM was using these things in the late 1970's. When I worked for DEC, IBM had introduced their own line of minicomputers to compete with the PDP11. DEC bought two of these computers to evaluate. One was used to evaluate IBM software and computer performance. The other machine was taken apart to evaluate the construction. They even took the IC's apart! I remember seeing some of the circuit boards from the IBM 'series 1' mini computers, and they looked alot like the ones you are showing, but the IC's were a bit larger. (Denser chips).
I had the privilege of using PDP 11/70 serial number 1 while I worked for DEC (1978-1993). However, VAX was very far from "the beginning of the end for DEC". For 10 years it was an outstanding success story with massive growth in the business. It wasn't until the late 1980s that DEC started having real problems that ultimately led to it dumping many parts before being bought by Compaq.
@@jeremybarker7577 I left DEC in the begining of 1980. By then the VAX/11-50 was being prototyped,, also the "JAWS" PDP-11-73. They were a great place to work, especially the 'mill'.
Having dedicated layers for power and ground greatly reduces the need for decoupling capacitors. I suppose the additional layer cost was offset by fewer components and higher reliability.
Wow! How those 80 column cards brought back a flood of memories from the mid to late 70's. Learned to program on a 360/30 but also lucky enough to get a job as a weekend operator on an NCR system. Great times. Thanks for sharing.
Fran I was hired IBM in 1978 to work at the Burlington Vermont Plant (located in Essex Junction) in the FET Memory Test Analysis Engineering Dept 576. I worked on ROS (Read Only Storage) in 1 inch metal packages and the Reisling memory in 1/2 inch metal packages. We tested the memory at Chip, Deck (chips attached to ceramic substrate) and Module level. The module may have either one (1K) or two stacked ceramic decks (2K). The "22" marked on your modules is probably the plant ID number. The solder bridge between the chip die and the ceramic substrate is a process called a "C4" joint (controlled collapsible chip connection). The back seal could be the brown silicone or green epoxy. Most 1/2 inch modules were filled with clear "Sylgard" (we called it Snot) at encapsulation.
The ceramic substrated units are referred to as Thick Film technology. The resistors and traces were silk screened on the substrate and subsequently fired in a kiln. There was also Thin Film technology where the substrate was glass, but the passive components were deposited using vapor or sputtering usually in a vacuum chamber using photoresist masking. Lasers could precision trim passive components. Allen Bradley (now Rockwell) did a lot of development of those processes. This silastic sealed units are not considered hermetically sealed. That is only achieved by welding a case to a header.
Oh, yeah. My first foray into mainframes was writing BAL (Basic Assembly Language) on an IBM 360 in the mid '70s. I had programmed microcontrollers in assembler, so the BAL part wasn't particularly unfamiliar, but the JCL needed to run the jobs was worse than the actual code. 0:35, in the very foreground, shows some of the 14" disk drives like we had, about the size of a small washing machine, they would rock 'n' roll whenever they did a series of head seeks (I've still got a platter from one somewhere around in one of my junk boxes).
Both JCL and 360 assembler are so weird, as if they were blindly copied from some alien technology without regard for human thought processes. (BAL also is the mnemonic for Branch And Link, used in a subroutine call)
I was a was a systems programmer on 370's. Did that until around 1888 or so. The first language I learned was BAL (assembler), I worked mostly with the NCP, TCAM and VTAM but dabbled in other stuff too. I used to get calls in the middle of the night, I would fix JCL errors over the phone. Funny thing was half the time I didn't even remember the call when I got to work the next day!
I vaguely remember hearing somewhere that IBM had an in-house designed PCB layout system with an autorouter, which was based on a grid system, and that's why their boards at the time looked like this. I've got an IBM microchannel token ring card from an PS/2 PC dating back to around 1986 which uses the same grid system. It's got four SLT modules, a bunch of regular ICs, and even a couple of good sized PGA ICs too. One of the SLT modules even has its own clip-on heatsink!
@UpLateGeek Did the adapter have a mechanical relay? They were a surprising component on token ring adapters. If the power to the adapter went out the adapter's connection to the adapter had to be shunted for the ring to be complete without it. If not the ring would be interrupted and therefore unusable. The spring in the relay did the trick mechanically.
Very interesting as always. I didn't realise that multilayer boards were being used that far back. The punched cards really took me back. I took an Mechanical Engineering Degree at Portsmouth Poly back in the late 70's, and they had an ICL computer that we had a course on. We programmed in Algol using punched cards that were hand punched in a little desktop punch. You took the whole stack to a card printer which read them and printed what you'd punched along the bottom edge. That had to be manually checked and binned if it was wrong. Once that was all done, you submitted it with an elastic band around it in an envelope and waited a week to get a tiny scrap of printout that usually said something like 'Syntax error - line 2'. Not exactly interactive computing! I was still absolutely hooked though.
Very similar punch-card reminiscences from 1978 in Sydney Australia. ICL computer, submitting decks of cards, awaiting the printout. Friends greatly valued any buggy discards for the way they rolled into a perfect, tight, springy joint filter. Of course, if I ever partook, I never inhaled. Honest.
ditto... Substitute Warwick for Pompey, a Burroughs B6700 and Algol 68 back in 75 to 78 I went on to work for IBM and used the same card technology throughout the 80's when designing test equipment - still have some in my study...
@@8BitNaptime Heck, HP used a 16 layer teflon insulated PC board in 1968 for their HP 9100A calculator. They needed 32kbits of ROM, and ROM chips were just not a thing available to consumer electronics back then. Rather than stuffing a *massive* board with who knows how many diodes... presumably somewhere around half of 32000, they instead created a multilayer circuit board that created closely coupled layers of zigging and zagging traces in one plane, and pairs of sense lines running perpendicular in the other closely coupled plane. They could drive a row, then detect with sensitive amplifiers on the columns, the inductively coupled pulses from the intersecting traces. With a 16 layer board, they had a density of around 1000 bits per square inch. HP's engineer's problem was "We need around 16000 diodes to create a ROM, that's too many, and will be way to big and expensive". I absolutely love that their solution to the problem was "Okay, then don't use any parts to store the data at all. Just use an empty circuit board" XD
So gorgeous and over-engineered, even if less elegant than the path we ended up on. The boards from 76-81 take me back to 82, when I was a little kid pilfering discarded boards from my dad's CS lab.
I worked for IBM for 37 years. I repaired 360 mainframes. The SLT cards had a part number that started with 580 followed by 4 numbers ei 5801234. Very reliable. Bad cards were rare. There was also ASLT which had 2 ceramic substrates stacked in a single metal can. ASLT was used in high end computers. SLT morphed into SLD which used an IC in each can. System/370 used MST which also used ICs but was emitter coupled logic. Thanks for the walk down memory lane.
Hi Fran, it would be perfect fit if you could get hold of a Marshall Time Modulator, It used CCD sensor from spy satellite tech as a sort of bucket brigade to make a lovely flanging effect and much more.
"I.C.," declared the transistor discretely. Assembly interpretation: package indents in the top are position guides for the substrate, with the silicone poured in over the bottom. Seal without the sweat.
I wish I could get my hands on one of those IBM 22s so I could turn them into a simple two LED multi vibrator and frame it lol However, the best part of the video has to be that last shot of all those vintage 7400 ICs… no words.. beautiful…
I would never have thought that miniaturization like this was being made in the 60's. I do recall the introduction of the 'module', which incorporated a number of components in a 5 or 6 pin package.
When I joined IBM in 1982 at the East Fishkill Semiconductor facility after 4 years in the USAF (Avionics) we still had equipment that used SLT, if I remember correctly they were 'brush cleaners' used to clean the wafers between processing steps. The first uP based piece of equipment we had was an SVG brush cleaner using the Zilog Z-80 uP. Also popular in that time frame was the Motorola 6502. At about the time I joined IBM they were introducing their 308X series of mainframes that used water cooled TCM's (thermal conduction modules) to cool the chips that would otherwise get too hot and fail. It is interesting to note that the performance of the 308X series ranged from 5MIPS to14MIPS compared to a modern day MacBook Pro at about 400,000MIPS or about 28,000X the performance of a high end mainframe of the early/mid 1980's. The early SLT and similar modules incorporated about 4 active components per module whereas modern CPU's can have around 100B components, or about 250,000X as many.
Great video! My late father was one of the system programmers for the first 360s here, when a few of the largest norwegian banks got together and established a common data center for all transactions in the early 1960s. He kept on until the age of 78 when he finally retired. Their first CPU had 64k of core memory, IBM meant it would be sufficient for years. After 6 months they had to expand to 128k.
I interviewed at the IBM East Fishkill (NY) facility in 1969. They built SLT'S there, as well as custom silicon. One interesting thing about the fabrication process for SLT: after the transistor dies were bonded, the SLT was put into a test fixture and a puff of air was directed at each transistor. The test fixture looked for noise indicating s defective bond. So I guess bond reliability was a problem. Puffs of air were a clever solution.
Yep, learned FORTRAN-IV in the mid-70s using punched cards AND punched tape (on a dial-up connection with a Teletype Model 33). We kept some of the old cards to use as book marks.
What a coincidence! I learned Fortran using punch cards too in the early '80s. I hated having to throw away and retype a whole card because of one mistake and the long lines in front of the card reader. I dreaded seeing someone with a huge COBOL stack in front me, because inevitably a card would jam-up the reader and you had to wait for a lab tech to unjam the machine.
@@MrWildbill Ahh, but token ring was a pain to fault find. In college for example it was common for students who were late with their homework to take the whole thing down with a pin through the cabling.
I remember a Dilbert strip in which Wally and Dilbert told their “pointy haired boss” that when he unplugged his computer he brought down the whole network because the “token” fell out onto the floor. So the boss was on his hands and knees trying to find the “lost token!”
At Shell Services Corporation in the latter 90's, we had token ring using RJ-45 cabling (don't know if you want to call it thin or thick). It was always a pain, because you'd have to call the other building constantly to get ports unlocked. Also worked with Synaptics Latticenet. The most fun was StarNet, which was pretty high powered coaxial direct runs from a hub. You could get the snot shocked out of you if you weren't paying attention.
I remember that too. I worked on IBM equipment in the 80's, and almost all of the circuits used this style of tracing. Token Ring was a bit of a pain to debug, but unlike Ethernet it coped with up to 100% utilisation without any degradation as new packets were slotted in as data was forwarded between nodes. Ethernet on the other hand would attempt to send data, find a collision, then each node waited a random time before trying again, so once you got above about 70% utilisation your throughput dropped dramatically. Sounds like it wouldn't matter, but at the time Ethernet was only 10Mbit/s, and TR was 4 or 16Mbit/s. I still remember the clicking of the relays as they looped themselves into the ring :)
I have stacks of unused punch cards, haven't the heart to throw them out. In pink blue white and yellow. They came from Ballistic Missile Early Warning site 3 at RAF Fylingdales and were used on their 1963 IBM mainframes. Those massive computers ran 25+ years up to 1989 when they were decommissioned.
I've still got a bunch, too. We had a PDP-11/70 at university that students had disk space on during the semester, but it was wiped every term. Those of us who wanted to keep stuff around had to punch all our source out to cards, then reload it at the beginning of the next semester. Good times (ha ha).
Oh wow - first time I've seen punch cards since 1984, the last year that they were used in New Zealand within a university course - one that also taught Pascal. I was pleased to hold that little bit of history back then. THANKS FRAN!
I remember those times. At Auckland University, the two mark-sense card readers would read around one ton of cards each per day when assignments were due.
I used to work on a "hybrid chip" assembly "robot". It was mostly pneumatic & mechanical. Yep ceramic mini boards, and a very precise glass bead blaster. It'd have test instruments plugged in and you'd burn away the resistors until they matched spec. Then like a giant turntable, it'd rotate in the next part, dump the "done" one & it'd feed in an uncalibrated one. One of my favorite memories was the "robot" broke, but there were no existing repair people, services, etc. I volunteered and spent a week rebuilding it. It worked until the company folded years later. Even sped up the machine with a few modern tricks. Never knew what those chips were for!
As soon as I saw your thumbnail I saw a thick film resistor. I used to work in the production of thin film resistors primarily on ceramic substrates in most phases of production. Fascinating stuff
In the late 1960's I acquired a small box of these boards from an IBM 360/370. That computer was dropped off the back of a moving truck onto the street. Apparently boards went flying all over the street and the computer was scrapped by a Federal Government sub-contractor. My dad brought home the boards for me to dissect. (I still have a couple of them.) Some of those Hybrid SLT circuits used a clear silicone jell rather than a hard conformal coating like the one you disassembled. These were probably the first surface mount components and the forerunner of the Ball Grid Array. IBM was still using some of this metal can/ceramic substrate packaging and circuit board technology in their PS-2 line of personal computers and their line of hard disk drives.
WOW! What a blast from the past. My dad had a Fortran manual that I’ve got buried somewhere. Maybe I’ll be able to make my museum shelf when I finish my unpacking.
when I was a kid they had "introduction to computers" class, and one of the classes was about punch cards (we got to create one) so I came up with the line "You can't punch a card wearing boxing gloves"
Hey Fran, in the mid-80s I spent a couple of 6-month training placements at IBM here in the UK, and I remember those pin grid cards well. They had unpopulated and un-tracked ones for development and prototyping, with internal ground and power planes connected to regularly spaced pins so you could lay out rows of 14- or 16-pin TTL logic ICs and wire the circuit with fine solid core wire. They had a production facility with rows of workstations where operators with microscopes and dental drills would mine into production PCBs to make cut and jump mods to inner layer traces! I learned a lot there.
Frikin brilliant! It would so cool to see what current day mainframes are made of. Also disk drives, 'member seeing 3380 disk drives, that reminded me of washing machines. Brilliant! Thx so much from a COBOL programming dinosaur.
IBM used boards exactly like this all the way into the early to mid 80's; large greenbar printer machines (3262/5262) had stacks of logic boards in them with chips exactly like these....
I upvoted this within 50 seconds. I've been working with computers for 40+ years and never even heard of SLT. More please! Awesome, fascinating, information!
Our System/3 from the mid/late 1970s also used the same type of cards. It's hard to remember all the terminology now, but I think that the card plugged into cages. The cages were given a letter; one side was BGMS (mnemonic Buy General Motors Stock) and the other DJPU (Dow Jones Price Up). If memory serves, the cages were assembled into gates that were on hinges that you could swing out to service the individual cards. Fans at the bottom blew cold air (from the under-floor air conditioning) up through the cages, exhausting out of the top past thermal sensors. If a gate got too hot, a red light labelled Thermal Check would illuminate and the CPU would halt. Anyway, the service manuals told you what card needed to be swapped by its address: Gate, cage, card. Each card had its own part number, of course, just as each SLT did.
That air trapped inside is full of carcinogenic chemicals no longer allowed to be near human beings, LOL. I worked at IBM Endicott .. it was a well known fact that employees were punished by being forced to work in "Building 18" .. the plating bath lines, nasty chemicals, stuff growing on pipes, a horror show of carcinogens. Nobody wanted to work there for obvious reasons... so they sent the problem employees there ... IBM "promised them a job for life" .. so they poisoned themselves and went to bldg 18 .. then in 1992 IBM started laying off 1/2 their work force (SURPRISE!!! ..no job for life, lol).
So many memories, not all good. I spent a memorable chunk of my life wrestling with card punch machines, typing in Fortran code with a heavy mechanical shift to get digits. My pinky wasn't strong enough to lift whatever inside the machine had to move to access digits, so I had to shift my whole hand over. We usually used 029 machines, because it had more characters, but sometimes you'd get stuck with the older 026 and have to do tricks to get all the characters you need for Fortran onto the cards. (IIRC, there was a chart of characters that you could double-punch to get the right holes for the brackets and whatnot.) Now I'm having flashbacks of things like working with a card punch that had a dry ribbon so you couldn't see what you'd typed, dropping decks, ... . Fun times. Now, kiddies, if you do what your mother tells you and eat all your vegetables, after dinner I'll tell you about typing technical documents on an IBM MT/ST.
@Amber One summer in grad school in the 1960s I worked for a professor who was writing a FORTRAN Primer textbook. She gave me her handwritten copy that I keypunched (029, IIRC) into 80-column cards. She had a 7070 mainframe program that read the cards and performed pagination, orphan/widow control, and other word processor-like functions. Great summer job: easy work, indoors, air conditioning, and in the process I picked up some programming tips that hadn’t been covered in a prior FORTRAN class. Little did I know ... (Twelve months later I was a newly-minted IBM Systems Engineer, in a ten-week training class)
In 1986-87 I worked for a company that did military manuals and provisioning. We had to provide magnetic tape of the parts lists to the government. We had an IBM mainframe for this purpose. Computer access floor for all the cabling, and a devoted air conditioning unit to keep the monster cool.
10:03 those resistors, and how they trim out the film to dial in the value required, is similar to what they do today with normal metal film resistors, they look like normal modern resistors, but they have a helical cut along the cylindrical body of the resistor and that is how they adjust the length of the spiral that the current must pass. I have a few small samples of equipment similar to those, amazing really, and definitely able to withstand adverse environments !
Same here in Australia Fran. I also learned FORTRAN with punchcards as a very young Engineer back in the late Eighties/ early Nineties, and have never used them since. Even back then, punchcards were a bit of an oddity. Was about that time as we started to see 5.25" 64/128k single and double sided floppy drives, even 250MB Tape back up drives.
When I was a Cub Scout around 1970, we made wreaths using old punch cards and gold spray paint. Being a huge nerd even back then, I was far more fascinated with the punch cards than the end product.
Sadly, if we still used punch cards today, the security department wouldn’t allow used cards to be used for anything like that, because tech savvy people could read the cards and steal secrets.
I recall working on one 360/512 in 80s last century. When serviceman had to replace this or that in CPU cabinet, he actually had to step inside into the cabinet sometimes. At that time PDP-11 was just starting to take over the duties.
9:00 - Hybrid modules. Early surface mount transistors & capacitors mounted onto a ceramic substrate. Resistors are "painted" onto the substrate. My mother worked in a hybrid lab, at what use to be Hoffman Electronics. This was a known technology across the electronics industry, though used mostly on more pricey gear. My high school had an IBM 360. It was the only computer in that school, in the 1970s. Its main purpose was the school's administration, though there was one computer class, that used it to. All "batch processing", so the class didn't interfere with normal operation.
I have a board from an IBM Mag Card Typewriter, this was a Selectric golfball typewriter from the 1970/80's attached by a big thick umbilical cable permanently attached to a large and very heavy box, larger than a PC tower case. The box had a Card reader that used media the same size as a punched card but instead of cardboard it was made from plastic and had a magnetic layer, your documents were written to and read from this. It made a magnificent mechanical noise. Eventually I had to take it to bits as I had nowhere to store it. The part that remains is the processor board (I think) in the same square PCB format and similar silver IC's in the video, it is 23cm wide by 33cm tall and hangs on my wall with a Colour RGB LED light display behind it.
Спасибо за ваши ролики, они такие добрые! 🙏 У меня такие IBMовские игрушки в детстве были. У них так легко снималась крышка и можно было заглянуть внутрь, но понятней не становилось 😀 А ещё от ИБМ была уникальная печатная головка в форме шара, на поверхности которой были нанесены литеры. Ах, какое время было! И никакого дела не было до того, что страна разваливается. Ух, пронесло! Теперь вы держитесь! Все умные, честные и добрые на одной стороне! 🙋♀️
This same hybrid technology also went down deep into the earth, to explore and instrument oil & gas wells and also into geothermal wells. See DOE geothermal logging instrumentation programs around 1980.
In the late 80s and early 90s the community college I attended to learn multiple programming languages, data structures, etc, was still operating an IBM 370 in their computer lab for students to code on and other batch jobs. It was used most heavily by the assembly language courses (QuickC on the PC was more popular for C and Borland Pascal for Pascal…then FORTRAN, Lisp, Prolog, and so on). The professor I took assembly language programming with in my very first semester of college (1989, I was still in high school at the time) had to sign an override form for the Admissions Office to let me into the class since I wanted to take the course without any of the prerequisite courses and I drove him crazy all semester because I was the one student who refused to draw up a flowchart for such simple programs - if I can fit the whole program from start to finish in my head then why do I need a sulky flowchart - so he always docked me 5 points for every assignment, but with a 94 being an A a 95 on every exam and homework assignment was still an A, I think it bugged him that I could have been his first student to get solid 100s across the board but settled for 95s! That said, he got a little crazy with our final assignment to write an entire code editor for the IBM 370 entirely in assembly, I was the only student who even got close to a functional editor in the very limited time we had available to write the program for our final, nobody else got anywhere close (not even something that could be executed successfully) but I did at least have a rudimentary editor that accepted user input and could read and write to a file from the command line. A fully fledged editor written entirely in assembly from a dumb terminal and without the benefit of any preexisting libraries that can be leveraged to accelerate development is quite a lot for the average programming student to complete in just a few hours of data entry in an otherwise entry-level assembly language course! Interestingly, we received a very similar assignment in our C Data Structures class for our final exam and a week to code that in QuickC (so we could code from home on a home PC and DOS) and I got much further along in that editor (actually a proper word processor this time), I had even gotten as far as adding spellcheck and populating a learning dictionary and the ability to utilize expanded memory to speed up spellcheck and string searches with large files. If I had more time then I was very close to implementing a graphical windowing library I had written to give it overlaying windows and pulldown menus (and mouse compatibility) while running under MS-DOS, but I ran out of time before I had it all implemented. If one ever wishes to learn to program in the U.S. it is well worth attending a community college for that particular skill development as our Universities tend to focus much more on theory than the practical in computer science, so the university is the place to go to learn how to do logic proofs and other advanced concepts (they gear you towards being the project manager essentially), but community college is where to go to learn your actual programming languages (and that was especially true in the pre-Internet days where there were no online guided courses and few free compilers, interpreters, and editors available).
Fran: I worked as a IBM 360 Operator. OS/360-with HASP was a solid opsys. One thing most do not know is the IBM Mainframes were all developed originally by IBM working on the old S.A.G.E. system. SAGE cost a billion bucks in the 1950s and the R&D there was applicable to the later System 360. I worked a university that nursed a 360 along for decades. Back in the day it was supported by a bunch of nerd hippies popping up the false floor and re-working BUS and TAG cables to get the beast to work... Those were the days....
That was fun! Enjoyed seeing the IBM stuff, and the old chips... looks like you were rummaging around in my own junk box... especially that bit at the end :-)
I have some old punch cards from my employer, customized with their crest printed on them, from when we had a zSystem... The blank backs make great note cards/bookmarks for me :-)
My first job in the early 90's was coding RPG on an IBM midrange S/36. That multiuser, multitasking machine had only 256K of memory! Programs were much less complex in those days, and I guess the proper use of the RPG cycle and indicators helped.
The first year at Cambridge University (1979) I programmed in FORTRAN using punched cards to submit jobs to the mainframe IBM 370/165 (including a few cards of JCL incantations at the front of the card stack). The output would be printed on wide aspect ratio white paper by a voracious 1403, and posted by the printer guy into filing cabinet dividers in what was called the "output tank". Every registered user had a username (mine was FB11) and their own labelled divider. Compute time was rationed by an algorithm based on shares and usage history. Lowly users including freshers like me were only allocated one or two shares. I remember going to see Judy Bailey, who was Deputy Director and had a reputation for being a bit of a dragon, to plead Oliver Twist-like for more shares. I think she was a bit bemused by my audacity but was very nice about it. The next year they replaced the output tank with VDU terminals, allowing users to edit programs before submitting them online, rather than the frustratingly slow and error-prone process of punching cards. Lots of fond memories of those days.
In the mid to late '60s, I went from operator in a unit record (punched card) installation, to programming on an IBM 1440 in Autocoder, and then to programming in PL/1 on an IBM System/360 Model 50. I wasn't so interested in the circuitry at the time, but rather IBM's promise of code being able to migratedto more powerful processors without change. Unlike the programs I wrote for the 1440 which couldn't run on the 1401 or 1410, much less on their 7000 series, programming I did on the 360 models moved without change to the 370, 390, Amdahl whatever, and Fujitsu compatibles.
This construction lives on in many types of automotive sensors. They use the ceramic boards, bare dies, trimmed carbon film resistors, and then pot the whole thing the Dow Corning silicone snot. That said, the later ones just use jumper wires similar to the LED "flip-chip" style lights instead of the solder directly up to the dies. The earlier stuff was tough as can be. Have a late 1990's Delphi mass airflow sensor that came from an SUV that was buried under mud after Katrina. After distilled water and a toothbrush, it came back to life, and still is in my car to this day.
Those old IBM cards used xsxp for signal vs power so those were probably 2S-2P meaning 2 signal 2 power. One power for + and one ground. We made some amazing boards in the 1990's. I remember before I took the buyout in 93 we did an engineering job that was 12S 4P.
The resistors might be ruthenium. I worked in early SMT before it was used on PCBs, as it was originally designed for ceramic substrates. The resistors back then were essentially a ruthenium paste printed over the conductor layer then vitrified in a furnace (~850 deg C). Then, as you mentioned, trimmed into value.
Amazing content! In many ways I wish this level of hardening was still applied to today's tech and we wouldn't end up with so much waste or damage to the environment
When IBM began using DIP packaged circuits, the SLT or MST boards got a new designation. VTL for VENDOR Transistor Logic. This also facilitated hand wiring ICs on standardized boards with common voltage and ground anes for industry DIP packages. Or designing less complicated wiring configurations for standard parts. Later on, IBM packaged it's own LSI into MST packages. Monolithic System Technology. Many cards have very mixed sets of components built on to incorporate all the levels of tech, discrete, DTL, SLT, VTL, and MST all on the same circuit board. Your 351497 is a package of four discrete transistors. Found in bitsavers IBM SLT_SLD_Module_Data_Dec1969. 361456 is 8 diodes and two resistors. 316457 is 4 transistors again.
I likewise learned FORTRAN in the mid 80's, at Ohio State University. We were the first class to use PC's to do our coding -- we had to walk to another building to fetch our printouts, which were run on a laser printer that ran at 5 mph, several pages wide, which were slit down to size before being collated and placed in our cubbies. (BONUS TRIVIA: I recall that our TA, Vince Bullock, was researching photoelastic stress analysis, and had developed a program that simulated such analyses -- and rendered visual results using ASCII art!)
In 1979, I attended Ok State U, and learned programming on an IBM 370. The machine had been donated to the University, from a business, and was maintained by the Math College, professors and students. As far as I know, IBM's only interaction was consultant and (sometimes) spare parts. It was fascinating seeing the innards from that machine. Punch cards until my last semester, wherein I splurged on terminal time (not included-- you paid by the minute). But my total terminal time, at semester's end, was less than $200... in '79 dollars, though. Pretty freaking cool. "IBM spared no expense" is quite accurate.
FAIRCHILD developed the ICs which were used in the Apollo computers. And by the way, I personally owned an 370/145 computer (in my house!) which used MST logic. That's what the larger board has on it there. I had been a field engineer for Tandem computers in the early 80s.
Lovely video! What was the typical level of function of each of these pages? Are we looking at something that was, say, a binary adder or register, or did each page do more (or less) than that?
I wondered the same thing, but sadly I bet it was a nibble requiring 3 or more Solid Logic packages to equal 1 bit.. or did you mean the whole page/board? Sorry if I sound dumb, I'm trying to understand how it can retain state for variables not on a punch card... too much to think on
@@purplebrain7421 I'm not sure what you're asking. First, a nibble is typically 4 bits. It's also 4 bits on the 360, which had 32bit words addressed in terms of 8-bit bytes. (It's very familiar!) If you're curious as to how solid-state memory works, google the terms "RS Latch" and "Flip Flop". It's not complicated, but it'll probably keep you busy for an afternoon.
@@recompile In my case I had damaged magnetic media that I had to run back and forth copying the physical space between an upper and lower layer of bits, and had to show it as a "physical" representation of the media data. That was shown as a half bit, upper stream and lower. At the rime we were calling them nibbles, the data was less than 300kb and it took mind numbing days to finish.
Every now and then the UA-cam algorithm recommends something that is truely interesting, made by someone who actually knows what they’re talking about. This video is one of those rare occurrences. Well done!
At 14 min: Writing Fortran on punched cards then putting them into the hopper of the reader. Yes I remember those days too. Using a mainframe across the Atlantic ocean, communicating by satellite link as the internet was not even a figment of imaginations in the 70s. All that heavy duty hardware, with the capacity of a PC running Windows 98. Kept in service decades later due to software that could not be migrated easily as source code was usually no longer available. Employing developers late into their working lives, simply because they knew such things as Cobol, while those around them were doing other stuff. How times have changed, thank heavens.
Southern Ohio College, 1981, Cincinnati Ohio - we were online to a computer ran by Metridata (the company that owned the college) was IBM Model 360 and we used it for Cobol, RPM and 360 Assembler classes. Not one of those computer languages due I use today.
I like that your workbench looks like a Radio Shack exploded nearby. You and I both, evidently, spent a good deal of time, in Radio Shack. (I was a fixture!)
Fran, Regarding those part numbers, I can’t say for sure but it appears the IBM 22 has some other meaning, perhaps the plant that manufactured them? The actual part numbers shown at around the 5:55 minute mark would seem to be the 361456 or 361457, and documentation on SLT devices refers to similar numbers. There is another very interesting UA-cam video on this technology posted by Uniservo. I would really like to get a copy of the specific IBM design brochure he shows in the video, but he apparently never uploaded it to Bitsavers or Internet Archive as he said he was going to? I did find something similar but not quite as detailed or extensive though. Anyway, you can find Uniservo’s video here: ua-cam.com/video/9h0qJdj6CEk/v-deo.html
You are correct. IBM 22 was the plant of manufacture (East Fiishkill, NY). The 361 number is the part number for that device. Each P/N having specific circuity. The r-pacs were designated by the 239**** P/N.
Well for some reason UA-cam clipped off the rest of that after the url. Here’s the rest: 361456 called an AOX-B (Appears to contain 2 dual diodes, a quad diode, and 2 resistors) 361457 called an FTX 9V (Contains 4 transistors) .
I'm always amused when I compare today's electronics to the university mainframe I used in the early 80s. I still remember the awesome performance improvement when they added an additional 16M of memory to their vax 11/780. Students used to wait an hour to compile a 300 line program. Or compare to my high school's pdp-8 with 8K of 12 bit core memory, for three users running BASIC. Or compare to my first microcomputer that I expanded from 256 bytes to 20K. The 16K s-100 board drew a full 3 amps.
IBM did contribute one HUGE advancement to the advancement of IC's once they were involved full swing. And that was Chemical Mechanical Planarization (CMP). This is also called Chemical Mechanical Polishing". But, to be pedantic, planarization (flat, not necessarily shiny) and polishing (shiny, not necessarily flat) aren't quite the same thing. CMP allowed a huge increase in transistor density. Before this, you got higher density by making the dies bigger and/or the transistors smaller. Now, you could stack many layers (up to maybe 10 nowadays). So it is sort of a cousin to being a multi-layer PC board. But the big problem is that at small sizes, the photolithography process step has virtually no depth of field. So the deposition of transistors and metal connects must be on a ludicrously flat surface. When you put a dialectric (insulating) layer over a metal layer, the surface will not be flat as it will somewhat conform to the features of the metal layer underneath. (sort of what a conformal coating over PC board traces looks like.) CMP makes it flat. Then holes are etched where needed to make connections to the layer underneath. Then the next active layer is deposited. And then the next dielectric layer, and so on. There are enormous technical difficulties in achieving good CMP, however.
I worked at the company where IBM initially purchased some manual polishing machines (designed for another purpose). They were very secretive about what they were doing. This was in the late 1980's. Then they contracted with us to design and build a quantity of purpose built automated machines for this. Many, many others adopted CMP. For a number of years, the company I was at and one other sort of split the marketplace. Then others got into it, eventually including the "big kid on the block" Applied Materials. it is my understanding that IBM got over 1 billion in licensing fees from the CMP technology.
@@beautifulsmall That is a very good analogy. For some reason, many in the semiconductor industry have moved toward the word "polishing". And I don't know why. But every industry develops its own lexicon. Even if some words are technically incorrect. Such as on the Southern Pacific Railroad they called all their articulated locomotives "Mallets". A "Mallet" is a specific type of articulated engine with both high and low pressure cylinders. A "simple" articulated has all high pressure cylinders. That is what Southern Pacific actually had, but they still incorrectly called them "Mallets".
According to the history of IBM technology by Pugh et al there was a big battle internally between proponents of SLT and ICs. The designer of the SLT series got demoted because the new 360 range was slated in the press for not using ICs. However he was vindicated (and reinstated) a few years later when it was shown that the SLT packaged solutions had a reliability around 100 times greater than IC based systems. They adopted ICs later - built by themselves - when the technology could offer the same reliability.
Those SLTs are more commonly known as "Hybrids", in the industry, at large. They still exist in more specialized applications. Moore's law made the IC win, in the end. Though, in the mid 60s, no one knew that. When I first studied electronics, in high school, in the 70s, hybrids were a common sight, at surplus stores.
Hello Fran!! Yes SLT!! I worked at the East Fishkill N.Y. IBM plant for 31 years, I started in 1977 August and even then SLT was being produced in extremely large numbers! They were produced in BLDG 310!! Those film resistors were trimmed by a process called LASER ACTIVE TRIM! the device SLT was loaded into a computerized machine with the socket structure and the computer
would select the correct pins to monitor THEN the laser would start to trim the selected resistor one at a time!! but quickly!! I worked in production control, and had to move the product from department to department and had to keep track of production loses and the type of loss that was incurred !!! Fran you re ignited some cool memories of a bygone era for me thanks!!
Hi John. And now they brew beer in building 310!
what products used these SLT boards? I havent seen these and doubt any home computer used these.
@@effexon These SLT devices were used in the IBM 360 and 370 systems, These were very large mainframe systems of the day back in the 60's and 70's. The particular unit shown by Fran was still
produced in smaller quantity up until the late 1978 or 79 as replacement parts!
I really like hearing about things like this, thank you for sharing a great piece of history.
My dad worked for Raytheon. He spent a summer working on a join venture with IBM . It was a nice surprise to see naked chips of the IBM type in books on IBM 360 machines. Dad had a few he kept , including one made into a tie clip. He worked first for Raytheon's tube division then for it Missle division. I had a friend who worked for IBM. He had many card from their smaller systems, using the later versions of these chips, in their metal cans.
I learn almost as much from reading the comments under Fran's videos as I do from Fran herself! Thanks, Fran, for bringing such a wonderful community together.
SLT was discontinued in the late 1960s. There was an argument at IBM when the System 360 was in planning whether to go with hybrid, or fully integrated circuitry. In hindsight it was better to go with hybrid (SLT) at that time, fully integrated would have added at least a year to system 360 development time. The card you have your finger on at 0:29 is NOT an SLT card. If you look at the other side of the card you will see that the packages have 16 rather than 12 pins. There is an integrated circuit inside. It looks very similar to SLT, but it was called MST (Monolithic System Technology). The logic was in fact ECL rather than TTL. System 370, and even some system 360 models used MST. Pull the lid off one of those packages. System 360s did not use the grey card carrier that 370s did.
While some SLT modules (the name IBM used for the square metal packages) only had 12 pins others had 16. One other point was that it was possible during development to implement a SLT module with discrete components on a small board with the pins. Back in the mid-1990s I went to a talk by some of the people who designed the 360/40 CPU and they had some of their prototype boards containing a combination of normal SLT modules and a few made up from transistors, diodes and resistors.
@@jeremybarker7577 SLT (Solid Logic Technology) applied to the technology and packaging system of very small discreet components. Standard SLT packaging used 12 pins although some later versions did use 16. ULD (Unit Logic Device) was SLT miniaturized and packaged for the Apollo spacecraft. Fran displays a lovely example in this video. ULD was not packaged in the square metal packages. There were some later SLT developments that used 16 pins. ASLT had two stacked substrates and SLD (Solid Logic Dense) put resisters on the bottom of the substrate, like the modules for the Apollo. ASLT was used in the s/360 9x series of large computers.
IBM refers to the units here as cards, cards are plugged into boards. Boards have no logic on them, but backplane wiring to connect cards with each other. Boards are mounted in gates, gates are mounted in frames. In IBMese none of the devices in this video are boards, they are cards.
The System 360/40 was developed in the IBM plant in Hursley, England. The production versions used standard SLT, all devices had 12 pins. I have worked on many 40s.
One unique thing with SLT was the method of attaching the active devices to the substrate. Conventional devices connected the devices with small wires. SLT put tiny solder coated copper balls on the substrate, then the device (transistor) was placed on top of the balls. The module was put into an oven to flow the solder. The technique proved highly reliable.
When IBM started producing monolithic technology in about 1967 they used the same technique to attach the chips to the same square silicone chip used in SLT. Commercial ICs in DIP packages were connected with tiny wires, IBMs were attached with the little copper balls. This has another advantage that circuit paths can be printed under the chip. The IC version was called Monolithic System Technology (MST). The packages looked identical to SLT, but had 16 pins. The larger card that Fran shows is definitely an MST card, not SLT. Under the metal package will be 1 or more monolithic chips, and no resistors on the chip. I hope she takes the lit off of a package and shows us.
I have the little logic handbooks the engineers used for reference (SLT, ASLT, MST MST-4, etc.). Each is about 3 inches by 4 inches with pale blue, green, red covers
I’m glad that this history is still being shared.
Those cards brought back some memories! When I was a kid my dad bought a load back from work to show me, I wish I still had them. Later at school we could send cards to the local college to run on an ICL 3900 minicomputer, I taught myself to punch cards using one of the 12-key punches that had a key for the hole in each row, and you punched in the EBCDIC code for the characters you wanted. I actually got quite fast at it and pretty accurate, luckily programs where usually less than 100 cards long! Waiting a week or so to find out if you'd punched the cards right was tension making too! They would come back in a bag with the contents printed along the top of the card, so I got in the habit of re-using cards that had standard statements on them so I didn't have to re-punch them, a very early example of code reuse!
Those green cards were "JOB" cards, the first card in a deck. They were green so the operators could separate the jobs easily on the output hopper of the card-reader, wrap elastic bands around them, match them with the job-output from the printer and give the job-decks and output back to the programmers, or, for production jobs, put the card-decks back to run the next night or week or month.
I still work in the mainframe arena. Still
Going strong. Biggest problem is a lot of the code is old and those that knew it have retired or passed away. Still a very relevant technology to this day as it is a very robust both hardware and OS wise. As database servers there is little to beat them.
As for the punch cards, many large scale printers use them to this very day as measurements for gaps etc are done in IBM punch cards. As such, they are still made. Gaps between the laser and the drum, developer rollers, sensors etc all measured in x punch cards.
"I don't know what programming languages will look like in 2100, but I know what it will be called: COBOL 2100." 😁
The banks cannot afford to loose track - therefore it will probably be a century long maintenance contract for some banks with IBM before mainframes are replaced with something... Maybe cryptos (probably not the traditional Bitcoin or a bit newer Ethereum type though) finally get to sunset all the COBOL-architectures.
Yep, former IBM/360 370 Customer Engineer, later Advisory Programmer, System Programmer. Still see jobs available, but I hung em up 20 years ago.
My uncle worked at the IBM plant in Poughkeepsie for many years where they designed and made many of those SLT packages. At one point, during the late 60's or early 70's, he arranged a tour for me and my father. I remember thinking at the time that those SLT packages were already a bit obsolete, I had already seen TI DIP packages in other products. Each SLT "can" contained semiconductor components which consisted of a tiny rectangle of silicon cut from a grown wafer. Each silicon rectangle had either two (diode) or three (transistor) very tiny copper or gold spheres bonded to it, and then the leads were soldered to the spheres creating the circuits. By current standards, very low-density stuff, but rugged. The fabrication process was all internally developed by IBM, they were almost fanatical about doing everything in-house. My uncle got an award for the helping develop the process they used to align the little silicon chips properly in the fab, it was an amazing piece of Rube-Goldberg engineering, but it worked. On a different note, those boards in the 360/370 systems may have been reliable, but the back-plane wiring was a nightmare. The interconnect circuitry used something called "try leads", which gave endless trouble and were very difficult to diagnose and repair. At one point, when I was an operator in a large mainframe shop we spent the better part of a shift waiting for the service tech (CE) to find a bad lead on a backplane and get the system running again.
I first saw these in the early 80's in a surplus IBM "data adapter unit" which is a VERY large UART essentially. I still have one of the small modular plugin circuits. Amazing technology - developed originally for the NSA according to their published history at fairly significant expense.
My grandfather, Sal Calta, was an engineer on the 360. It's wonderful to see his work up close. Thank you Fran.
DSCH53: I worked with both your grandfather (Sal Sr.) and his son, Sal Jr.
I worked at an IBM manufacturing plant in Charlotte, NC for 11.5 years. This video brings back memories of the 11 years of my career there as a test technician working on theses types of boards. I did incircuit and final test. When the boards failed we did component level repair too. They started switching to SMT boards and components in the late 80's. It saddens me to know the entire IBM Charlotte facility closed and this type of manufacturing was sent to the Pacific rim putting about 6,000 of us out of work. Thanks for the walk down memory lane nonetheless!
11:28 A tip: When desoldering through-hole components it helps to hold the board with the pins facing down on top of the iron held vertically with the tip facing up. That way gravity assists molten solder to flow down from the board to the bottom of the desoldering filter capsule.
These regular grid style boards surely made things easier in production but especially in the design system. So IBM used them even with DIP packages until wire density forced them to change.
Also interesting, the flipped and soldered transistor dies have been replaced by ICs that are connected via bonding wires. Today, the "flip-chip package" ist default again.
When I retired in 2000 we still had a 360 system running at HQ for some legacy accounting software at On semiconductor.
That lone old fogey SLT was included on that later board to keep those ornery young DIP chips in line.
These are the FranLabs I like! Thanks!
IBM was using these things in the late 1970's. When I worked for DEC, IBM had introduced their own line of minicomputers to compete with the PDP11. DEC bought two of these computers to evaluate. One was used to evaluate IBM software and computer performance. The other machine was taken apart to evaluate the construction. They even took the IC's apart! I remember seeing some of the circuit boards from the IBM 'series 1' mini computers, and they looked alot like the ones you are showing, but the IC's were a bit larger. (Denser chips).
I had the privilege of using PDP 11/70 serial number 1 while I worked for DEC (1978-1993). However, VAX was very far from "the beginning of the end for DEC". For 10 years it was an outstanding success story with massive growth in the business. It wasn't until the late 1980s that DEC started having real problems that ultimately led to it dumping many parts before being bought by Compaq.
@@jeremybarker7577 I left DEC in the begining of 1980. By then the VAX/11-50 was being prototyped,, also the "JAWS" PDP-11-73. They were a great place to work, especially the 'mill'.
Having dedicated layers for power and ground greatly reduces the need for decoupling capacitors. I suppose the additional layer cost was offset by fewer components and higher reliability.
Wow! How those 80 column cards brought back a flood of memories from the mid to late 70's. Learned to program on a 360/30 but also lucky enough to get a job as a weekend operator on an NCR system. Great times. Thanks for sharing.
Fran I was hired IBM in 1978 to work at the Burlington Vermont Plant (located in Essex Junction) in the FET Memory Test Analysis Engineering Dept 576. I worked on ROS (Read Only Storage) in 1 inch metal packages and the Reisling memory in 1/2 inch metal packages. We tested the memory at Chip, Deck (chips attached to ceramic substrate) and Module level. The module may have either one (1K) or two stacked ceramic decks (2K). The "22" marked on your modules is probably the plant ID number. The solder bridge between the chip die and the ceramic substrate is a process called a "C4" joint (controlled collapsible chip connection). The back seal could be the brown silicone or green epoxy. Most 1/2 inch modules were filled with clear "Sylgard" (we called it Snot) at encapsulation.
16:37 I always think of Custard Creams when I see these packages. It's a UK thing.
The ceramic substrated units are referred to as Thick Film technology. The resistors and traces were silk screened on the substrate and subsequently fired in a kiln. There was also Thin Film technology where the substrate was glass, but the passive components were deposited using vapor or sputtering usually in a vacuum chamber using photoresist masking. Lasers could precision trim passive components. Allen Bradley (now Rockwell) did a lot of development of those processes. This silastic sealed units are not considered hermetically sealed. That is only achieved by welding a case to a header.
I didn't think laser trimming was available in '65. Do you know anything about how and when that was done?
Oh, yeah. My first foray into mainframes was writing BAL (Basic Assembly Language) on an IBM 360 in the mid '70s. I had programmed microcontrollers in assembler, so the BAL part wasn't particularly unfamiliar, but the JCL needed to run the jobs was worse than the actual code. 0:35, in the very foreground, shows some of the 14" disk drives like we had, about the size of a small washing machine, they would rock 'n' roll whenever they did a series of head seeks (I've still got a platter from one somewhere around in one of my junk boxes).
Both JCL and 360 assembler are so weird, as if they were blindly copied from some alien technology without regard for human thought processes.
(BAL also is the mnemonic for Branch And Link, used in a subroutine call)
I was a was a systems programmer on 370's. Did that until around 1888 or so. The first language I learned was BAL (assembler), I worked mostly with the NCP, TCAM and VTAM but dabbled in other stuff too. I used to get calls in the middle of the night, I would fix JCL errors over the phone. Funny thing was half the time I didn't even remember the call when I got to work the next day!
@@bradjohnson9671 1888? Sounds about right... (Honestly, your typo had me in stitches for like 10 minutes.)
@@d00dEEE Wow, I didn't catch that! Some days it sure seems that long ago!
I vaguely remember hearing somewhere that IBM had an in-house designed PCB layout system with an autorouter, which was based on a grid system, and that's why their boards at the time looked like this. I've got an IBM microchannel token ring card from an PS/2 PC dating back to around 1986 which uses the same grid system. It's got four SLT modules, a bunch of regular ICs, and even a couple of good sized PGA ICs too. One of the SLT modules even has its own clip-on heatsink!
@UpLateGeek Did the adapter have a mechanical relay? They were a surprising component on token ring adapters. If the power to the adapter went out the adapter's connection to the adapter had to be shunted for the ring to be complete without it. If not the ring would be interrupted and therefore unusable. The spring in the relay did the trick mechanically.
Very interesting as always. I didn't realise that multilayer boards were being used that far back. The punched cards really took me back. I took an Mechanical Engineering Degree at Portsmouth Poly back in the late 70's, and they had an ICL computer that we had a course on. We programmed in Algol using punched cards that were hand punched in a little desktop punch. You took the whole stack to a card printer which read them and printed what you'd punched along the bottom edge. That had to be manually checked and binned if it was wrong.
Once that was all done, you submitted it with an elastic band around it in an envelope and waited a week to get a tiny scrap of printout that usually said something like 'Syntax error - line 2'. Not exactly interactive computing! I was still absolutely hooked though.
Very similar punch-card reminiscences from 1978 in Sydney Australia. ICL computer, submitting decks of cards, awaiting the printout. Friends greatly valued any buggy discards for the way they rolled into a perfect, tight, springy joint filter. Of course, if I ever partook, I never inhaled. Honest.
Yes they had at least 8 layer boards with flex capability in 1968.
ditto... Substitute Warwick for Pompey, a Burroughs B6700 and Algol 68 back in 75 to 78
I went on to work for IBM and used the same card technology throughout the 80's when designing test equipment - still have some in my study...
@@8BitNaptime Heck, HP used a 16 layer teflon insulated PC board in 1968 for their HP 9100A calculator. They needed 32kbits of ROM, and ROM chips were just not a thing available to consumer electronics back then. Rather than stuffing a *massive* board with who knows how many diodes... presumably somewhere around half of 32000, they instead created a multilayer circuit board that created closely coupled layers of zigging and zagging traces in one plane, and pairs of sense lines running perpendicular in the other closely coupled plane. They could drive a row, then detect with sensitive amplifiers on the columns, the inductively coupled pulses from the intersecting traces. With a 16 layer board, they had a density of around 1000 bits per square inch. HP's engineer's problem was "We need around 16000 diodes to create a ROM, that's too many, and will be way to big and expensive". I absolutely love that their solution to the problem was "Okay, then don't use any parts to store the data at all. Just use an empty circuit board" XD
@@richfiles Sounds like how IBM mainframes had programmable microcode in their high-end machines in the 1960s.
So gorgeous and over-engineered, even if less elegant than the path we ended up on. The boards from 76-81 take me back to 82, when I was a little kid pilfering discarded boards from my dad's CS lab.
I worked for IBM for 37 years. I repaired 360 mainframes. The SLT cards had a part number that started with 580 followed by 4 numbers ei 5801234. Very reliable. Bad cards were rare. There was also ASLT which had 2 ceramic substrates stacked in a single metal can. ASLT was used in high end computers. SLT morphed into SLD which used an IC in each can. System/370 used MST which also used ICs but was emitter coupled logic. Thanks for the walk down memory lane.
Hi Fran, it would be perfect fit if you could get hold of a Marshall Time Modulator, It used CCD sensor from spy satellite tech as a sort of bucket brigade to make a lovely flanging effect and much more.
"I.C.," declared the transistor discretely.
Assembly interpretation: package indents in the top are position guides for the substrate, with the silicone poured in over the bottom. Seal without the sweat.
I wish I could get my hands on one of those IBM 22s so I could turn them into a simple two LED multi vibrator and frame it lol
However, the best part of the video has to be that last shot of all those vintage 7400 ICs… no words.. beautiful…
Great blast from the past - thanks, I left schooll and joined a company in my home town that built computers from TTL DIP ICs.
I would never have thought that miniaturization like this was being made in the 60's. I do recall the introduction of the 'module', which incorporated a number of components in a 5 or 6 pin package.
When I joined IBM in 1982 at the East Fishkill Semiconductor facility after 4 years in the USAF (Avionics) we still had equipment that used SLT, if I remember correctly they were 'brush cleaners' used to clean the wafers between processing steps. The first uP based piece of equipment we had was an SVG brush cleaner using the Zilog Z-80 uP. Also popular in that time frame was the Motorola 6502.
At about the time I joined IBM they were introducing their 308X series of mainframes that used water cooled TCM's (thermal conduction modules) to cool the chips that would otherwise get too hot and fail. It is interesting to note that the performance of the 308X series ranged from 5MIPS to14MIPS compared to a modern day MacBook Pro at about 400,000MIPS or about 28,000X the performance of a high end mainframe of the early/mid 1980's.
The early SLT and similar modules incorporated about 4 active components per module whereas modern CPU's can have around 100B components, or about 250,000X as many.
Great video! My late father was one of the system programmers for the first 360s here, when a few of the largest norwegian banks got together and established a common data center for all transactions in the early 1960s. He kept on until the age of 78 when he finally retired. Their first CPU had 64k of core memory, IBM meant it would be sufficient for years. After 6 months they had to expand to 128k.
I love the way you bring history of electronics back to us that we were never a part of this technology then.
I interviewed at the IBM East Fishkill (NY) facility in 1969. They built SLT'S there, as well as custom silicon. One interesting thing about the fabrication process for SLT: after the transistor dies were bonded, the SLT was put into a test fixture and a puff of air was directed at each transistor. The test fixture looked for noise indicating s defective bond. So I guess bond reliability was a problem. Puffs of air were a clever solution.
Yep, learned FORTRAN-IV in the mid-70s using punched cards AND punched tape (on a dial-up connection with a Teletype Model 33).
We kept some of the old cards to use as book marks.
What a coincidence! I learned Fortran using punch cards too in the early '80s. I hated having to throw away and retype a whole card because of one mistake and the long lines in front of the card reader. I dreaded seeing someone with a huge COBOL stack in front me, because inevitably a card would jam-up the reader and you had to wait for a lab tech to unjam the machine.
I remember around the early 90s the place I worked buying IBM Token ring network boards using this type of construction
Yeah, and I've got some 30-pin RAM sticks from a PS/2 which used them too
@@MrWildbill Ahh, but token ring was a pain to fault find. In college for example it was common for students who were late with their homework to take the whole thing down with a pin through the cabling.
I remember a Dilbert strip in which Wally and Dilbert told their “pointy haired boss” that when he unplugged his computer he brought down the whole network because the “token” fell out onto the floor. So the boss was on his hands and knees trying to find the “lost token!”
At Shell Services Corporation in the latter 90's, we had token ring using RJ-45 cabling (don't know if you want to call it thin or thick). It was always a pain, because you'd have to call the other building constantly to get ports unlocked. Also worked with Synaptics Latticenet. The most fun was StarNet, which was pretty high powered coaxial direct runs from a hub. You could get the snot shocked out of you if you weren't paying attention.
I remember that too. I worked on IBM equipment in the 80's, and almost all of the circuits used this style of tracing.
Token Ring was a bit of a pain to debug, but unlike Ethernet it coped with up to 100% utilisation without any degradation as new packets were slotted in as data was forwarded between nodes. Ethernet on the other hand would attempt to send data, find a collision, then each node waited a random time before trying again, so once you got above about 70% utilisation your throughput dropped dramatically.
Sounds like it wouldn't matter, but at the time Ethernet was only 10Mbit/s, and TR was 4 or 16Mbit/s.
I still remember the clicking of the relays as they looped themselves into the ring :)
I have stacks of unused punch cards, haven't the heart to throw them out. In pink blue white and yellow. They came from Ballistic Missile Early Warning site 3 at RAF Fylingdales and were used on their 1963 IBM mainframes. Those massive computers ran 25+ years up to 1989 when they were decommissioned.
I've still got a bunch, too. We had a PDP-11/70 at university that students had disk space on during the semester, but it was wiped every term. Those of us who wanted to keep stuff around had to punch all our source out to cards, then reload it at the beginning of the next semester. Good times (ha ha).
Oh wow - first time I've seen punch cards since 1984, the last year that they were used in New Zealand within a university course - one that also taught Pascal. I was pleased to hold that little bit of history back then. THANKS FRAN!
I remember those times. At Auckland University, the two mark-sense card readers would read around one ton of cards each per day when assignments were due.
@@paulcarter7445 Don't drop the cards on the way.
I still live by the code...
"do not fold, spindle or mutilate" and I don't let others do that to me either.
Back a year later, and enjoyed as much again. Now I call it a Legacy Clip.
I used to work on a "hybrid chip" assembly "robot". It was mostly pneumatic & mechanical. Yep ceramic mini boards, and a very precise glass bead blaster. It'd have test instruments plugged in and you'd burn away the resistors until they matched spec. Then like a giant turntable, it'd rotate in the next part, dump the "done" one & it'd feed in an uncalibrated one.
One of my favorite memories was the "robot" broke, but there were no existing repair people, services, etc. I volunteered and spent a week rebuilding it. It worked until the company folded years later. Even sped up the machine with a few modern tricks. Never knew what those chips were for!
As soon as I saw your thumbnail I saw a thick film resistor. I used to work in the production of thin film resistors primarily on ceramic substrates in most phases of production. Fascinating stuff
In the late 1960's I acquired a small box of these boards from an IBM 360/370. That computer was dropped off the back of a moving truck onto the street. Apparently boards went flying all over the street and the computer was scrapped by a Federal Government sub-contractor. My dad brought home the boards for me to dissect. (I still have a couple of them.) Some of those Hybrid SLT circuits used a clear silicone jell rather than a hard conformal coating like the one you disassembled. These were probably the first surface mount components and the forerunner of the Ball Grid Array. IBM was still using some of this metal can/ceramic substrate packaging and circuit board technology in their PS-2 line of personal computers and their line of hard disk drives.
WOW! What a blast from the past. My dad had a Fortran manual that I’ve got buried somewhere. Maybe I’ll be able to make my museum shelf when I finish my unpacking.
when I was a kid they had "introduction to computers" class, and one of the classes was about punch cards (we got to create one) so I came up with the line "You can't punch a card wearing boxing gloves"
Hey Fran, in the mid-80s I spent a couple of 6-month training placements at IBM here in the UK, and I remember those pin grid cards well. They had unpopulated and un-tracked ones for development and prototyping, with internal ground and power planes connected to regularly spaced pins so you could lay out rows of 14- or 16-pin TTL logic ICs and wire the circuit with fine solid core wire. They had a production facility with rows of workstations where operators with microscopes and dental drills would mine into production PCBs to make cut and jump mods to inner layer traces! I learned a lot there.
Frikin brilliant!
It would so cool to see what current day mainframes are made of.
Also disk drives, 'member seeing 3380 disk drives, that reminded me of washing machines.
Brilliant! Thx so much from a COBOL programming dinosaur.
"It's a bobby-dazzler!!" ... now that's something I haven't heard in years ;-)
There's an old man that handles the metal detector in the TV show "The Curse of Oak Island" that uses that term frequently.
@@cancel1913 Gary Drayton's not that old! He's 60 or 61
Watch Dave from eevblog :)
EEVBlog does it all the time. Probably where she got that from.
ack your reminding me of the difference between hand soldering and wave soldering
IBM used boards exactly like this all the way into the early to mid 80's; large greenbar printer machines (3262/5262) had stacks of logic boards in them with chips exactly like these....
I upvoted this within 50 seconds. I've been working with computers for 40+ years and never even heard of SLT. More please! Awesome, fascinating, information!
Our System/3 from the mid/late 1970s also used the same type of cards. It's hard to remember all the terminology now, but I think that the card plugged into cages. The cages were given a letter; one side was BGMS (mnemonic Buy General Motors Stock) and the other DJPU (Dow Jones Price Up). If memory serves, the cages were assembled into gates that were on hinges that you could swing out to service the individual cards. Fans at the bottom blew cold air (from the under-floor air conditioning) up through the cages, exhausting out of the top past thermal sensors. If a gate got too hot, a red light labelled Thermal Check would illuminate and the CPU would halt. Anyway, the service manuals told you what card needed to be swapped by its address: Gate, cage, card. Each card had its own part number, of course, just as each SLT did.
It'd be interesting to measure CO2 content of air trapped inside and try to correlate with what year it was made.
That air trapped inside is full of carcinogenic chemicals no longer allowed to be near human beings, LOL. I worked at IBM Endicott .. it was a well known fact that employees were punished by being forced to work in "Building 18" .. the plating bath lines, nasty chemicals, stuff growing on pipes, a horror show of carcinogens. Nobody wanted to work there for obvious reasons... so they sent the problem employees there ... IBM "promised them a job for life" .. so they poisoned themselves and went to bldg 18 .. then in 1992 IBM started laying off 1/2 their work force (SURPRISE!!! ..no job for life, lol).
So many memories, not all good. I spent a memorable chunk of my life wrestling with card punch machines, typing in Fortran code with a heavy mechanical shift to get digits. My pinky wasn't strong enough to lift whatever inside the machine had to move to access digits, so I had to shift my whole hand over. We usually used 029 machines, because it had more characters, but sometimes you'd get stuck with the older 026 and have to do tricks to get all the characters you need for Fortran onto the cards. (IIRC, there was a chart of characters that you could double-punch to get the right holes for the brackets and whatnot.) Now I'm having flashbacks of things like working with a card punch that had a dry ribbon so you couldn't see what you'd typed, dropping decks, ... . Fun times. Now, kiddies, if you do what your mother tells you and eat all your vegetables, after dinner I'll tell you about typing technical documents on an IBM MT/ST.
@Amber
IBM MT/ST - The word processor’s grandmother. ;) (Mag Tape/Selectric Typewriter)
@Amber
One summer in grad school in the 1960s I worked for a professor who was writing a FORTRAN Primer textbook. She gave me her handwritten copy that I keypunched (029, IIRC) into 80-column cards. She had a 7070 mainframe program that read the cards and performed pagination, orphan/widow control, and other word processor-like functions. Great summer job: easy work, indoors, air conditioning, and in the process I picked up some programming tips that hadn’t been covered in a prior FORTRAN class. Little did I know ... (Twelve months later I was a newly-minted IBM Systems Engineer, in a ten-week training class)
In 1986-87 I worked for a company that did military manuals and provisioning. We had to provide magnetic tape of the parts lists to the government. We had an IBM mainframe for this purpose. Computer access floor for all the cabling, and a devoted air conditioning unit to keep the monster cool.
10:03 those resistors, and how they trim out the film to dial in the value required, is similar to what they do today with normal metal film resistors, they look like normal modern resistors, but they have a helical cut along the cylindrical body of the resistor and that is how they adjust the length of the spiral that the current must pass. I have a few small samples of equipment similar to those, amazing really, and definitely able to withstand adverse environments !
Those IC's are all familiar they used to be the mainstay of components when I was working with this equipment. Nice to see. Thanks for sharing.
Utterly fascinating
Same here in Australia Fran. I also learned FORTRAN with punchcards as a very young Engineer back in the late Eighties/ early Nineties, and have never used them since. Even back then, punchcards were a bit of an oddity. Was about that time as we started to see 5.25" 64/128k single and double sided floppy drives, even 250MB Tape back up drives.
When I was a Cub Scout around 1970, we made wreaths using old punch cards and gold spray paint. Being a huge nerd even back then, I was far more fascinated with the punch cards than the end product.
Sadly, if we still used punch cards today, the security department wouldn’t allow used cards to be used for anything like that, because tech savvy people could read the cards and steal secrets.
I recall working on one 360/512 in 80s last century. When serviceman had to replace this or that in CPU cabinet, he actually had to step inside into the cabinet sometimes. At that time PDP-11 was just starting to take over the duties.
9:00 - Hybrid modules. Early surface mount transistors & capacitors mounted onto a ceramic substrate. Resistors are "painted" onto the substrate. My mother worked in a hybrid lab, at what use to be Hoffman Electronics. This was a known technology across the electronics industry, though used mostly on more pricey gear.
My high school had an IBM 360. It was the only computer in that school, in the 1970s. Its main purpose was the school's administration, though there was one computer class, that used it to. All "batch processing", so the class didn't interfere with normal operation.
I have a board from an IBM Mag Card Typewriter, this was a Selectric golfball typewriter from the 1970/80's attached by a big thick umbilical cable permanently attached to a large and very heavy box, larger than a PC tower case. The box had a Card reader that used media the same size as a punched card but instead of cardboard it was made from plastic and had a magnetic layer, your documents were written to and read from this. It made a magnificent mechanical noise. Eventually I had to take it to bits as I had nowhere to store it. The part that remains is the processor board (I think) in the same square PCB format and similar silver IC's in the video, it is 23cm wide by 33cm tall and hangs on my wall with a Colour RGB LED light display behind it.
Спасибо за ваши ролики, они такие добрые! 🙏 У меня такие IBMовские игрушки в детстве были. У них так легко снималась крышка и можно было заглянуть внутрь, но понятней не становилось 😀 А ещё от ИБМ была уникальная печатная головка в форме шара, на поверхности которой были нанесены литеры. Ах, какое время было! И никакого дела не было до того, что страна разваливается. Ух, пронесло! Теперь вы держитесь! Все умные, честные и добрые на одной стороне! 🙋♀️
OMG the little plastic base for the transistors! I'd forgotten them.
This same hybrid technology also went down deep into the earth, to explore and instrument oil & gas wells and also into geothermal wells. See DOE geothermal logging instrumentation programs around 1980.
In the late 80s and early 90s the community college I attended to learn multiple programming languages, data structures, etc, was still operating an IBM 370 in their computer lab for students to code on and other batch jobs. It was used most heavily by the assembly language courses (QuickC on the PC was more popular for C and Borland Pascal for Pascal…then FORTRAN, Lisp, Prolog, and so on). The professor I took assembly language programming with in my very first semester of college (1989, I was still in high school at the time) had to sign an override form for the Admissions Office to let me into the class since I wanted to take the course without any of the prerequisite courses and I drove him crazy all semester because I was the one student who refused to draw up a flowchart for such simple programs - if I can fit the whole program from start to finish in my head then why do I need a sulky flowchart - so he always docked me 5 points for every assignment, but with a 94 being an A a 95 on every exam and homework assignment was still an A, I think it bugged him that I could have been his first student to get solid 100s across the board but settled for 95s! That said, he got a little crazy with our final assignment to write an entire code editor for the IBM 370 entirely in assembly, I was the only student who even got close to a functional editor in the very limited time we had available to write the program for our final, nobody else got anywhere close (not even something that could be executed successfully) but I did at least have a rudimentary editor that accepted user input and could read and write to a file from the command line. A fully fledged editor written entirely in assembly from a dumb terminal and without the benefit of any preexisting libraries that can be leveraged to accelerate development is quite a lot for the average programming student to complete in just a few hours of data entry in an otherwise entry-level assembly language course! Interestingly, we received a very similar assignment in our C Data Structures class for our final exam and a week to code that in QuickC (so we could code from home on a home PC and DOS) and I got much further along in that editor (actually a proper word processor this time), I had even gotten as far as adding spellcheck and populating a learning dictionary and the ability to utilize expanded memory to speed up spellcheck and string searches with large files. If I had more time then I was very close to implementing a graphical windowing library I had written to give it overlaying windows and pulldown menus (and mouse compatibility) while running under MS-DOS, but I ran out of time before I had it all implemented.
If one ever wishes to learn to program in the U.S. it is well worth attending a community college for that particular skill development as our Universities tend to focus much more on theory than the practical in computer science, so the university is the place to go to learn how to do logic proofs and other advanced concepts (they gear you towards being the project manager essentially), but community college is where to go to learn your actual programming languages (and that was especially true in the pre-Internet days where there were no online guided courses and few free compilers, interpreters, and editors available).
Fran: I worked as a IBM 360 Operator. OS/360-with HASP was a solid opsys. One thing most do not know is the IBM Mainframes were all developed originally by IBM working on the old S.A.G.E. system. SAGE cost a billion bucks in the 1950s and the R&D there was applicable to the later System 360. I worked a university that nursed a 360 along for decades. Back in the day it was supported by a bunch of nerd hippies popping up the false floor and re-working BUS and TAG cables to get the beast to work... Those were the days....
That was fun! Enjoyed seeing the IBM stuff, and the old chips... looks like you were rummaging around in my own junk box... especially that bit at the end :-)
I have some old punch cards from my employer, customized with their crest printed on them, from when we had a zSystem... The blank backs make great note cards/bookmarks for me :-)
we used to wrap decks of used punchcards in masking tape. they made excellent paperweights
He can mimic any circuit board to copy same or operate it without connection amazing
My first job in the early 90's was coding RPG on an IBM midrange S/36. That multiuser, multitasking machine had only 256K of memory! Programs were much less complex in those days, and I guess the proper use of the RPG cycle and indicators helped.
The first year at Cambridge University (1979) I programmed in FORTRAN using punched cards to submit jobs to the mainframe IBM 370/165 (including a few cards of JCL incantations at the front of the card stack). The output would be printed on wide aspect ratio white paper by a voracious 1403, and posted by the printer guy into filing cabinet dividers in what was called the "output tank". Every registered user had a username (mine was FB11) and their own labelled divider. Compute time was rationed by an algorithm based on shares and usage history. Lowly users including freshers like me were only allocated one or two shares. I remember going to see Judy Bailey, who was Deputy Director and had a reputation for being a bit of a dragon, to plead Oliver Twist-like for more shares. I think she was a bit bemused by my audacity but was very nice about it. The next year they replaced the output tank with VDU terminals, allowing users to edit programs before submitting them online, rather than the frustratingly slow and error-prone process of punching cards. Lots of fond memories of those days.
In the mid to late '60s, I went from operator in a unit record (punched card) installation, to programming on an IBM 1440 in Autocoder, and then to programming in PL/1 on an IBM System/360 Model 50. I wasn't so interested in the circuitry at the time, but rather IBM's promise of code being able to migratedto more powerful processors without change. Unlike the programs I wrote for the 1440 which couldn't run on the 1401 or 1410, much less on their 7000 series, programming I did on the 360 models moved without change to the 370, 390, Amdahl whatever, and Fujitsu compatibles.
This construction lives on in many types of automotive sensors. They use the ceramic boards, bare dies, trimmed carbon film resistors, and then pot the whole thing the Dow Corning silicone snot. That said, the later ones just use jumper wires similar to the LED "flip-chip" style lights instead of the solder directly up to the dies. The earlier stuff was tough as can be. Have a late 1990's Delphi mass airflow sensor that came from an SUV that was buried under mud after Katrina. After distilled water and a toothbrush, it came back to life, and still is in my car to this day.
I learned RPG inputting my code on punchcards.
Great video Fran!
Punchcard Fortran for me too, circa 1980.
I remember learning to program using an IBM 360 back in my college days. Many a night I would be punching cards trying to get my assignments to run.
Those old IBM cards used xsxp for signal vs power so those were probably 2S-2P meaning 2 signal 2 power. One power for + and one ground. We made some amazing boards in the 1990's. I remember before I took the buyout in 93 we did an engineering job that was 12S 4P.
The resistors might be ruthenium. I worked in early SMT before it was used on PCBs, as it was originally designed for ceramic substrates. The resistors back then were essentially a ruthenium paste printed over the conductor layer then vitrified in a furnace (~850 deg C). Then, as you mentioned, trimmed into value.
Amazing content!
In many ways I wish this level of hardening was still applied to today's tech and we wouldn't end up with so much waste or damage to the environment
When IBM began using DIP packaged circuits, the SLT or MST boards got a new designation.
VTL for VENDOR Transistor Logic. This also facilitated hand wiring ICs on standardized boards with common voltage and ground anes for industry DIP packages. Or designing less complicated wiring configurations for standard parts.
Later on, IBM packaged it's own LSI into MST packages. Monolithic System Technology.
Many cards have very mixed sets of components built on to incorporate all the levels of tech, discrete, DTL, SLT, VTL, and MST all on the same circuit board.
Your 351497 is a package of four discrete transistors. Found in bitsavers IBM SLT_SLD_Module_Data_Dec1969.
361456 is 8 diodes and two resistors.
316457 is 4 transistors again.
I likewise learned FORTRAN in the mid 80's, at Ohio State University. We were the first class to use PC's to do our coding -- we had to walk to another building to fetch our printouts, which were run on a laser printer that ran at 5 mph, several pages wide, which were slit down to size before being collated and placed in our cubbies. (BONUS TRIVIA: I recall that our TA, Vince Bullock, was researching photoelastic stress analysis, and had developed a program that simulated such analyses -- and rendered visual results using ASCII art!)
This is a nice glimpse into the past, when IBM did more than just lay off their employees on a regular basis.
In 1979, I attended Ok State U, and learned programming on an IBM 370. The machine had been donated to the University, from a business, and was maintained by the Math College, professors and students. As far as I know, IBM's only interaction was consultant and (sometimes) spare parts. It was fascinating seeing the innards from that machine. Punch cards until my last semester, wherein I splurged on terminal time (not included-- you paid by the minute). But my total terminal time, at semester's end, was less than $200... in '79 dollars, though. Pretty freaking cool. "IBM spared no expense" is quite accurate.
FAIRCHILD developed the ICs which were used in the Apollo computers. And by the way, I personally owned an 370/145 computer (in my house!) which used MST logic. That's what the larger board has on it there. I had been a field engineer for Tandem computers in the early 80s.
Currently reading The Apollo Guidance Computer by Springer. Fascinating stuff. Never had heard of rope memory before.
I remember seeing the similar boards with the same type of circuit chips on the IBM 1130 computers.
Lovely video! What was the typical level of function of each of these pages? Are we looking at something that was, say, a binary adder or register, or did each page do more (or less) than that?
Likely one bit slice.
...'S
I wondered the same thing, but sadly I bet it was a nibble requiring 3 or more Solid Logic packages to equal 1 bit.. or did you mean the whole page/board? Sorry if I sound dumb, I'm trying to understand how it can retain state for variables not on a punch card... too much to think on
@@purplebrain7421 I'm not sure what you're asking. First, a nibble is typically 4 bits. It's also 4 bits on the 360, which had 32bit words addressed in terms of 8-bit bytes. (It's very familiar!) If you're curious as to how solid-state memory works, google the terms "RS Latch" and "Flip Flop". It's not complicated, but it'll probably keep you busy for an afternoon.
@@recompile In my case I had damaged magnetic media that I had to run back and forth copying the physical space between an upper and lower layer of bits, and had to show it as a "physical" representation of the media data. That was shown as a half bit, upper stream and lower. At the rime we were calling them nibbles, the data was less than 300kb and it took mind numbing days to finish.
I did the computations for my master's thesis with punch cards in Fortran on an IBM 370.....in 1978-79.
I bet the engineer put that 22 and the old ti in the corner for nostalgic reasons, just like the etching art on a lot of other boards.
I had similar thoughts, that the older components placed there as proof of origin badge....
We were doing Fortran, Pascal and Cobol at computer school at 1990
Every now and then the UA-cam algorithm recommends something that is truely interesting, made by someone who actually knows what they’re talking about. This video is one of those rare occurrences. Well done!
At 14 min: Writing Fortran on punched cards then putting them into the hopper of the reader. Yes I remember those days too. Using a mainframe across the Atlantic ocean, communicating by satellite link as the internet was not even a figment of imaginations in the 70s. All that heavy duty hardware, with the capacity of a PC running Windows 98. Kept in service decades later due to software that could not be migrated easily as source code was usually no longer available. Employing developers late into their working lives, simply because they knew such things as Cobol, while those around them were doing other stuff. How times have changed, thank heavens.
Southern Ohio College, 1981, Cincinnati Ohio - we were online to a computer ran by Metridata (the company that owned the college) was IBM Model 360 and we used it for Cobol, RPM and 360 Assembler classes. Not one of those computer languages due I use today.
I like that your workbench looks like a Radio Shack exploded nearby. You and I both, evidently, spent a good deal of time, in Radio Shack. (I was a fixture!)
Fran,
Regarding those part numbers, I can’t say for sure but it appears the IBM 22 has some other meaning, perhaps the plant that manufactured them? The actual part numbers shown at around the 5:55 minute mark would seem to be the 361456 or 361457, and documentation on SLT devices refers to similar numbers. There is another very interesting UA-cam video on this technology posted by Uniservo. I would really like to get a copy of the specific IBM design brochure he shows in the video, but he apparently never uploaded it to Bitsavers or Internet Archive as he said he was going to? I did find something similar but not quite as detailed or extensive though. Anyway, you can find Uniservo’s video here: ua-cam.com/video/9h0qJdj6CEk/v-deo.html
You are correct. IBM 22 was the plant of manufacture (East Fiishkill, NY). The 361 number is the part number for that device. Each P/N having specific circuity. The r-pacs were designated by the 239**** P/N.
Well for some reason UA-cam clipped off the rest of that after the url. Here’s the rest:
361456 called an AOX-B (Appears to contain 2 dual diodes, a quad diode, and 2 resistors)
361457 called an FTX 9V (Contains 4 transistors)
.
I'm always amused when I compare today's electronics to the university mainframe I used in the early 80s. I still remember the awesome performance improvement when they added an additional 16M of memory to their vax 11/780. Students used to wait an hour to compile a 300 line program. Or compare to my high school's pdp-8 with 8K of 12 bit core memory, for three users running BASIC. Or compare to my first microcomputer that I expanded from 256 bytes to 20K. The 16K s-100 board drew a full 3 amps.