It looks like the larger 4k core plane was part of a core stack used in a Saratov 2 (Soviet PDP-8 clone): rusue.com/cemetery-of-soviet-computers/ . If someone can identify the 1k core plane, that would be terrific.
@@subutayozselanikli The 4K array is now identified as likely coming from a Saratov 2. Still don’t have a clue on the 1K. It is not very dense, so maybe you are right!
@Subutay Özselanikli That’s almost it! Ralph Mirebs below confirms the 1k is an M-4 memory stack (but not from an EVM computer) and the 4k is Saratov-2. Thanks a lot for the great website link www.mirebs.com/ . My core planes can be found on this page: www.mirebs.com/fozu/ru/_mod_ru.html
this would probably shock you, but, here, in Riga, Latvija, some boilier houses, even the bigger plants work on computers so old, that are equipped with theese memory drives. year 2021...
Love it! I too have some old core planes but I never thought anybody would actually get them going in a demonstrable form such as yours, wonderful preservation of a near lost technology
This man is teaching circuit theory in greater detail. It's amazing how generous are these explanations. An exceptional class of engineering and circuit theory.
Im russian. There is good french subtitles that can be autotranslated to any known lang. So i understand everything he's talkin' btw, i had a peace of such memory when i was a boy in late 80s Nice idea to make some arduino gate to USB from this mem, and test its speed
Especially if he reads it out, and you are able to continue the program from where it was last stopped all those years ago, the big advantage of core memory being non volatile.
@@SeanBZA Unfortunately the LVDC had up to 8 of these core modules, so whatever it holds (if anything) will only be a fraction of the program. And since we only have a single surviving example of an LVDC flight program (for the Saturn IB) that doesn't even assemble without errors, there is sadly probably not a lot that could be done with its contents.
Hi Curious; I have a Univac core plane from a state of the art memory we were testing in the late ‘70s. This was a “2 wire” core memory. We were getting ready for market. This memory banged the X & Y lines with very short but quite high currents. The trick was there was a delay from when the pulses were applied to when the core actually switched. There were separate gated sense amps on each of the X & Y lines that were strobed after the ringing to recover the bit. It was a marvel this actually worked at all. Shortly after, the project was abandoned because Intels solid state memory was cheaper. Though not as reliable as the core memory. Those were the days. redrok
smakfu The approach to random runtime corruption was: let it crash, and if it crashed too often you used core or doubled banks with parity selection, but the need for that was very limited - semiconductors did get better very quickly. Bit rot was a thing, and still is. But the reliability issue wasn’t all in the read error rates - in fact those were within usable ranges IIRC. The reliability issue was the chips just failing - whether bits dying, or rows/columns failing, or entire chips just catastrophically overheating (the power supply rails were totally unforgiving). All in all it was improving almost by the month when the Intel stuff was new and they were adjusting the process.
Brilliant, I love figuring out old technology it's kind of like Reinventing the wheel, I don't experiment with vintage computers I usually work with Old Tube radios different technology same thrill, I've been a subscriber for a few years enjoy your channel thank you for your work!
I find it so entertaining and heart-warming that you have a bench-grinder just next to the electronic test gear. You are a true all around renaissance man sir. My lab is very much the same way (albeit smaller and messier).
I bought one of those core planes on eBay a while ago and it is in one of my display cases in my office. Allegedly mine came from a Soviet clone of a DEC computer. The LVDC drop at the end - great bonus and tease - can't wait!!
I am quite curious to see what it would show if you tried to read every bit. Would require some massive circuirty but would be interesting. They probably erased it in one way or another, but it should be interesting nonetheless...
Ah, good old days when gods walked the Earth and Keysight was not even Agilent, but HP! (I still have somewhere in my attic a stack of those thick HP instrumentation catalogs and HP Journals, right from the time they decided to turn their instrumentation control computers into a general-purpose 1000 series minis.) Wonderful demonstration!
Yep! Today equipments are becoming so annoying. I just saw that in this video the agilent scope shows the same annoying bug my has. It always enable the cursors when doing automatic measures !! WHY !?? You can see the cursor jumping around, over the green trace
Agree! That wiggling cursor that you can’t turn off is so annoying! Could use a few more knobs instead of Russian doll menus too. That said I love this scope. It is so responsive and accurate.
@@CuriousMarc Yeah! I love it too. I will record this annoying bug and send the video to agilent. Ops, keysight 🤦♀️A friend of mine have the new Touch version and it do not have the bug 🙄
"We have done like three months of research in five minutes" and 50 years of improvement in test and signal generation equipment :P Love seeing how technology worked back then. Keep up the good work.
First, that is so beautiful. Secondly, I had always heard that we went directly from using vacuum tubes, right to using silicon chips; and this is a lot bigger I guess than silicon chip-but a LOT smaller than using vacuum tubes! Actually a very cool intermediary thing. Although it's pretty big for 1k, when I was a 17-year-old teenager the first PCs came out-and we were so excited that they were a whopping 64k!!! So, you can imagine stringing 64 of these together, and the box would be fairly large, but not that large. So, I think this is very cool.
Nice video. I worked on a couple of commercially available devices, made in America by Sperry Univac, in the early 1970s. One was a terminal, Univac Uniscope 100, which had core memory for its display memory. The memory boards wore manufactured at the facility I worked at in Salt Lake City, Utah. The core was later replaced by integrated circuit shift registers. The other device that CuriousMarc might be interested in was a Univac DCT 1000 printing terminal. A bit of a beast, but it had a couple of interesting features. One: it had character memory to allow buffering. The memory was a glass delay line module made by Corning glass. It had two flipping buffers holding 160 characters each. That was real fun to troubleshoot. The other feature of interest was a built in modem. This was back in the day when you could only get modems from Bell. Univac designed and manufactured their own modems. They got a huge push back from the Bell monopoly for hooking this foreign device to the Bell system who would never admit to the defects on their system even though we had the proof. Just a bit of nostalgia for me and thanks again for sharing and triggering that.
Ah, the Uniscope 100. You're probably very familiar with Engineering Procedure #1 to get a non-functioning U100 working: lift the unit an inch or so and drop it. That jiggled the PC board connections enough to restore good connectivity for a few days. Some DCT series devices were used as console printers for several generations of the 1100 series systems.
@@stevew8233 U100 had edge connector board that were not gold plated, so the contacts corroded, hence the drop. The 'fixed' the problem by putting fish oil on the contcts. Oops, then the slippery boards came out of the connectors in shipping. Then they designed a clamp to keep the boards in. Thanks for reminding me, I had completely forgotten.
Dear Marc, I've been a subscriber of your channel for a really long time and it amazes me how you are able to grasp the working and circuitry of such complex electronics so elegantly. Of course there is the documentation, but still getting so used to the circuitry as if you were part of it's development is truly amazing. I'm an upcoming computer engineer and would really appreciate if you have any advice or suggestions for a book that might help with developing a skill like yours ( if possible ) 😀
Oh, how I love watching these videos. Some lives ago I spend two years studying electrotechnics at university, before my life went into a totally different direction. Trying to follow the subjects triggers some deep parts in my brain I used to use in those days. I recognize equipment from the lab or from piles in the corridors. I see schematics and diagrams I remember from books. Fascinating to realize there was I time I was supposed to sort of understand this all. Now I find it utterly relaxing and entertaining to simply watch it. Thanks a lot! Hope you will be given ample time to keep on doing this.
In Air Force, working C-5A navigation computers, we had core memory in both primary and backup computers. One of the technical representatives and I figured out the code to which memory core failed. We were always handicapped by depot only maintenance, and lack of computer parts. We started fixing computers by making one really screwed up computer with not one working memory module, and fix eight or ten and sending them back to supply. Depot got angry, but gave in as we could fix it just as good as they did
You make this so much more understandable than when I was in college. We had a core memory computer but it was only for use by post-grads. Thanks for helping out an old Computing and Electronic Graduate from just after the end of the Apollo era . And now, back to reading the AGC Hardware book....
Some of us are old enough to have repaired the driver boards for these and adjusted the X and Y currents to find the Smoo curve. This curve tends to be a well rounded square of X and Y currents where inside, you don't have bit errors, and outside, you start having bit errors. We fought with one memory for a couple of weeks that would not stay working. turned out one of the carbon composition resistors was defective and would change value if wiggled. We upgraded and replaced the resistors with metal film resistors to fix the problem. These are temperature sensitive. For reliability, they are most often kept in an oven to hold the temperature stable. If you watch the old Disney movie The Computer Who Wore Tennis Shoes", this is why temperature was critical to the computer operating. The small sense area is for the Parity. The array is one bit. A stack of 4 makes a 4 bit nibble and a parity bit in the same array, so in the same oven of the same core batch.
I had a lot of that stuff about ~20 years ago. It was parts from old soviet ES EVM (ЕС ЭВМ) industrial mainframes, equivalents of IBM S/370 (really it was not copies but mostly equivalents with fully compatible interfaces and functionaluty designed on soviet guts). Some modules had dozens of these ferrite planes. Back in the day I was young and had no idea that such stuff is valuable and used it only for parts salvage. Later I cheaply sold the remains. The only things left is empty cases which are still used as big flower pots by my mom in the village☻
So worth watching till the end. I’m so amazed at the ingenuity of scientist yesterday and today. Wow I’m just in awe of what we are capable of. Thank you and Bless you for your all the time and effort you put into this video. I appreciate everything. The “one more thing” just did me in bro.
Reminds me of my days in the 70s at the University of Illinois where work was being done with Control Data Corporation on what was called Auxiliary Mass Storage. It was able to store the equivalent of 8MB (not really, CDC machines used 60 bit words with no parity; thank you Seymour Cray) that we were able to fit in about 9 cubic feet. We used to get memory errors in that bank of memory and our fix was to go over and kick the box low on one side -- there were mechanical relays that would get stuck sometimes.
I should add that the 6500 had another claim to fame. We got it from McDonnell Douglas where it was the computer they used to run the engineering analyses for the Gemini spacecraft. Oh, and they were the predecessors of super computers, not ones themselves.
Lovely! According to certain "experts" on the internet, the Apollo guidance computer couldn't work because core-rope memory wasn't real. Hilarious. And what a beautiful LVDC core... all that yummy stuff!
Mind that a core memory like this one was used as "ram", I don't think ram is the most exact name, but it does rougly the same job: stores data which will change during the mission (like istantaneous velocity or results of calculations); while the core rope stored "non changing" data, like the agc program itself and some constants, it was read only and used a similar physical principle of interaction between magnetic cores and wires but instead of only one you had more than 150 bits stored in each core by passing more wires in it, also because of this it wasn't possible to modify the data stored, it was literally... hard wired.
@@XMarkxyz That is exactly correct: core-rope memory is ROM (read-only, and hard-wired) whilst conventional core memory is RAM, random access memory for read and write operations.
Really? Of all the possible... objections? core rope? I suppose the film of it being made is fake, the modules in museums are empty, and Marc is actually a government agent installing 5G in our Yourtube feeds?
Soviet Ferrite wasn't memory only!! At ferrite logik was made soviet computers for atomic rockets... Only 10 kHz but non sensitive to radiation, vibration and others...600 ferrite rings at small block are 600 triggers
I'm stunned that I understood about 20% of this. I should spend more time learning instead of gaming. It seems so unlikely that we arrived where we have in computing technology from the devices we did. The PCB is so amazing.
As someone who knows nothing about vintage computing, everything you say usually goes well over my head, but I'll always come back for the pure aesthetic, space-age beauty of the technology.
Purpose of the inhibit wire. I found this online somewhere. "Many core memory arrays also have a 4th wire, known as the inhibit wire. Like the sense wire, one inhibit wire runs through all the cores. The purpose of the inhibit wire is to optionally provide current in the opposite direction to the write current, through the core being written. This will cancel out some of the magnetic field, and prevent the core from changing. Why would you want to do that? The answer is that it lets you easily stack multiple planes of core memory, all sharing the same drive wires. Without the inhibit wire, this would cause the corresponding core in each plane to be written. By sending the appropriate current down all the inhibit wires except one, we can ensure the write only affects the desired plane. The inhibit wires is an efficient way to scale up a system by adding multiple memory core planes. The alternative would be to make the matrices bigger and bigger, but eventually current losses and crosstalk make this impractical."
I think we bought the same 4k bit core memory! I also bought mine from Russia on eBay and looks identical to yours'. But I don't have the equipment you do to test it out. It's just a talking piece alongside my deck of punched cards and other paraphernalia
I'd always wondered about old memory from devices from years gone by and seen them as big towers/nests of wires and magnets but this was a lot easier to understand when it was in a single plane. Nice
That was done using a combo of Adobe Illustrator and DaVinci Resolve. Took a lot of effort. Resolve is very hard to use and quite buggy, but hey it’s free, and very capable, so it’s hard to complain about it. On the other hand the new Adobe subscription model is not even close to be worth the value for me, I mean by a mile. And they are buggy and overly complicated too. The Apple tools might be the right middle.
I got some of it....either way, I was hooked. Great video. I love old technology, ill be watching more. Thank you for taking the time to make these videos. True technology of old, compared to todays garbage. An I phone is lucky to be working after a year, this memory plate, as old as it is, was built to last. Amazing.
Brilliant lab setup - shows many principles of work - wish I had those during my engineering studies back in the 80s great scholary engineering video - thank you !
I still have the core memory out of a Wang 720C calculator I was gifted as a teenager (1980s). It had an intermittent fault and full of ICs and parts I could not identify. It had a ferrite transformer ROM too. To this day, I kick myself that I broke up the 720 as I’m sure I could have fixed it now, knowing that the ICs were often standard TTL with Wang part nos instead.
Th circuit you used for combining the pulses onto one wire is nearly the same as what you use to add in a stereo aux input to a vacuum tube radios pre-amp tube. Solid circuit.
Our PLCs where I used to work, were still using core memory right up until the 90s. The memory units were getting rare by that point, and expensive to service!
They used these in their knockoff of the DEC PDP-8/M, the Saratov-2. Also possibly the BESM-6 supercomputer. Another Soviet computer gadget to look out for are punched card readers, they kept on doing batch processing and other punched card stuff to the end of the USSR because nothing but tubes ever went obsolete in Soviet computing. I have seen Soviet punched cards for sale online dated "1987" - they weren't laminated like the IBM-Honeywell-Burroughs cards, just rough card stock.
It did take a fair amount of space in the Instrument Unit ring atop the Saturn V. There is the Launch Vehicle Digital Computer (LVDC) computer, multiple core stacks like this one, the Launch Vehicle Data Adapter (LVDA), and the Flight Control Computer (FCC) plus inertial platforms, sublimator cooling systems and telemetry gear.
Just wondering - would an MR sensor, such as from a disk read head, be able to read the core magnetic flux? Small enough for high spatial resolution. Donor hard drive could even supply the read amp (?!) . Back in the day we played with some super-sensitive MR sensors that could read the mag field of a Nd magnet disc being flipped in someone's pocket some ten feet distant.
For those wanting to “See” the magnetic polarity in the core I’m afraid this wont work. The problem is the magnetic flux is quite tightly contained in the core. Very little leaks out. So there is nothing to see. redrok
Hi joob; I seriously doubt the cores magnetic field would be visabe to a magnetic field sensor. These work because they work by sensing a magnetic field that has normal North & South poles. Toroidal cores don’t have this. Their fields are oriented around the core. So they don’t have visible poles. redrok
Wow 4kbit is actually about the size what the the bootrecord today is :-D there are games that fit in it so it would be funny to program a game in these things :-D
Would magnetic viewing film show the ferrite coils polarity without affecting the memory? It would be cool to take a picture before and after the state change if it worked.
@@CuriousMarc I would think the film would display a slight pattern difference in the orientation angle but It might prove interesting to see if you had a chance to post the results.
@@tootalldan5702 I don’t have film handy, but it’s very interesting suggestion, I’ll see if I can find some. Maybe we’d see a “twitch” when the core flips!
A great demo! Do you happen to know, what machines they're from? I've used to have a module from ES-1020 (a low-end model from the early IBM/360 clones, that particular machine was decommissioned around 1993), and I think it looked like your lower-integration module, but it was a long time ago, so my memory is fuzzy :-) It would be interesting if they turn out to be from different machine series, say ES and BESM (I think the series SM that included clones of HP and DEC PDP-11 machines used the integrated memory from the start).
@@CuriousMarc Are there any markings on the modules? Those might be helpful in identifying the machines. BTW, there also were machines made in at least GDR and Poland, GDR was making quite a few of high-end IBM mainframe clones, so the machines might turn out to not be Soviet too :-)
The first module looks like the one from soviet computer "Promin" (unfoirtunately, no English traslation): ru.wikipedia.org/wiki/%D0%9F%D1%80%D0%BE%D0%BC%D1%96%D0%BD%D1%8C_(%D0%BA%D0%BE%D0%BC%D0%BF%D1%8C%D1%8E%D1%82%D0%B5%D1%80) The second one looks like memory from ES series (IBM Syetem/3xx equivalents), probably early models, but I'm not sure.
Only problem is the size and slowness, but the FRAM IC's have gotten past the size and speed quite well, using a very similar storage based on magnetic dots.
@@CuriousMarc Wonder if they ever recover one from the watery grave, or from the lonely stage drifting out in space, if it will still be readable, containing the last instruction executed, which is likely a somewhat destructive to core memory of a jump back one word, if there is no actual halt instruction available.
With core memory, are the magnetic fields strong enough that you can see them using some "magnetic viewing film" sitting right on top of the core array, or would you basically see nothing/no activity at all?
Do we have any information about the Soviet textile workers / artisans who would have hand woven these core memory planes? I ask because here in the US there has been some documentation of the weavers working in the Raytheon factories in MA who hand crafted the core rope memory that was used in the computers that sent the Apollo to the moon.
Don’t know, I also thought of it for a second. But it’s unlikely: that would be way too many bits, and they are also somewhat distributed all over the place.
check this - from 3:59 - that's about ODRA Polish Computer - he was sold to Soviet - there you have original movie from this memory assembling : ua-cam.com/video/aDvtmjWQrzM/v-deo.html&ab_channel=Wytw%C3%B3rniaFilm%C3%B3wO%C5%9Bwiatowych PL language only ..
@@CuriousMarc thank you very much for your movie. For your work. Without this video, was for me impossible to understand how this memory works. thank You Sir.
We do destructive reading of RAM all the time up to now. Not only core memory uses destructive read, but dynamic RAM (as the core technology of all DDR RAM today) uses destructive reads, too. The main difference is that dynamic RAM uses electrically charged capacitors to store the bits, whereas core memory uses "magnetically charged" cores. Magnetism and electric charge are dual to each other: Magnetism is related to to current flowing, and you need to apply a voltage to change the magnetization. On the other hand, electric charge is related to voltage between some points, and you need to have current flowing to change the charge. That's why you apply current to the core memory and monitor the voltage, whereas for usual DRAM, you apply a voltage (0V, i.e. you discharge the capacitor) and monitor the current (to find out whether the capacitor was charged before. You don't need to care about the destructiveness of reading DRAM, because the fix for that is integrated in the DRAM chips. When you "open" a page (also called a "row"), all capacitors of that row are discharged, and the measured previous charging state is recorded in the "row buffer". When you "close" a page, the row buffer is written back to the row. Early DRAM chips (i.e. nearly everything before SDRAM) "opened" a row when you pulled /RAS low, and the address pins told the chip which row to open, and "closed" the row again when /RAS went high again. Thus there is no way to forget to undo the destructive read, because one row automatically gets closed (on the rising edge of /RAS) before another row can be opened (with the next falling edge of /RAS).
That would be very cool indeed. I actually tried. Unfortunately the magnetic field is many orders of magnitudes to small for the regular film to work. I tried with much more sensitive oil-based viewers instead, and that did not work good enough either (way too slow), although I saw a hint of something. Then I tried to contact some professional outfits that do magneto-optics films and cameras. That would work, but that's many 10's of thousands of $ worth of equipment and they were not interested in sponsoring a loan.
In school for my certifications, we had a few boards of core memory as examples. They were like 2x2 Ft, about 3000 cores on it. It's crazy when you remember they were hand wound.
It looks like 6 us on the scope. However you could still clearly see the core flip in-between the pickup spikes. If you had a gated sense amp, you could just restrict the sense to that time window and make it run almost as fast, at the expense of an even more complicated setup. Unfortunately my sense amps lack the gating feature.
Thank you for explaining this! 2:48 this may be the quantum mechanical wave behavior we've always worked around (destructive read means "the act of observing changes the state"). Spooky stuff! I have a multi head 3D printer. But I do not know how to 3D design. Is there a way to create a fully 3D printable version of one of these, if someone has a multi-head extruder? How to create the iron toroids, while lacing conductive XY traces. The wire is magnet wire, coated, right? 8:02 beautiful pulse! Is the sense and inhibit lines fed like a string through each toroids for practical reasons, or could you route it through all of them in parallel and get lower latency?
It looks like the larger 4k core plane was part of a core stack used in a Saratov 2 (Soviet PDP-8 clone): rusue.com/cemetery-of-soviet-computers/ . If someone can identify the 1k core plane, that would be terrific.
Probably Ural 1 (Урал 1). If there is any identifier on the module, it is possible to confirm.
@@subutayozselanikli The 4K array is now identified as likely coming from a Saratov 2. Still don’t have a clue on the 1K. It is not very dense, so maybe you are right!
No, 1K module from EVM M-4 (ЭВМ М-4) says mirebs dot com. You may look via translate, you will find Saratov-2 (Саратов-2) and more.
@Subutay Özselanikli That’s almost it! Ralph Mirebs below confirms the 1k is an M-4 memory stack (but not from an EVM computer) and the 4k is Saratov-2. Thanks a lot for the great website link www.mirebs.com/ . My core planes can be found on this page: www.mirebs.com/fozu/ru/_mod_ru.html
this would probably shock you, but, here, in Riga, Latvija, some boilier houses, even the bigger plants work on computers so old, that are equipped with theese memory drives. year 2021...
18:30 And from "it works!" we came to "let's see how good is it to overclock"
haha omg i just realized that's basically what he did
In soviet russia, computer remember you
doesn't need a password
lmao
omg hhahahahah
😂😂😂😂
That’s not funny my grandpa was a Russian computer
Love it! I too have some old core planes but I never thought anybody would actually get them going in a demonstrable form such as yours, wonderful preservation of a near lost technology
Just in case you haven't come across this series ua-cam.com/video/2KSahAoOLdU/v-deo.html
7 nanometers ? Pfff , 1 milimeter ! )
1 is smaller than 7 so its better B)
@@CPSPD look at the units
@@artisticyeti22 /woosh
@@artisticyeti22 /woosh
@@artisticyeti22 /woosh
This man is teaching circuit theory in greater detail. It's amazing how generous are these explanations. An exceptional class of engineering and circuit theory.
Anyone else here and watched the whole thing and don’t know anything he’s talking about, you sir got my attention and I have no clue why lol
i thought i was alone
Im russian. There is good french subtitles that can be autotranslated to any known lang.
So i understand everything he's talkin'
btw, i had a peace of such memory when i was a boy in late 80s
Nice idea to make some arduino gate to USB from this mem, and test its speed
For me, I thought he was going to play with that big ass core with games. lmao
Yes
Marc drops LVDC on us at the very end! Looking forward to that!
Especially if he reads it out, and you are able to continue the program from where it was last stopped all those years ago, the big advantage of core memory being non volatile.
@@SeanBZA I'm sure they're already working on peripherals for all the LVDC input/outputs :-) I wouldn't expect anything less!
@@SeanBZA Unfortunately the LVDC had up to 8 of these core modules, so whatever it holds (if anything) will only be a fraction of the program. And since we only have a single surviving example of an LVDC flight program (for the Saturn IB) that doesn't even assemble without errors, there is sadly probably not a lot that could be done with its contents.
Spoiler 😝😝
Hi Curious;
I have a Univac core plane from a state of the art memory we were testing in the late ‘70s. This was a “2 wire” core memory. We were getting ready for market.
This memory banged the X & Y lines with very short but quite high currents.
The trick was there was a delay from when the pulses were applied to when the core actually switched.
There were separate gated sense amps on each of the X & Y lines that were strobed after the ringing to recover the bit.
It was a marvel this actually worked at all.
Shortly after, the project was abandoned because Intels solid state memory was cheaper. Though not as reliable as the core memory.
Those were the days.
redrok
Amazing I didn't know! Thanks for telling the story.
smakfu The approach to random runtime corruption was: let it crash, and if it crashed too often you used core or doubled banks with parity selection, but the need for that was very limited - semiconductors did get better very quickly. Bit rot was a thing, and still is. But the reliability issue wasn’t all in the read error rates - in fact those were within usable ranges IIRC. The reliability issue was the chips just failing - whether bits dying, or rows/columns failing, or entire chips just catastrophically overheating (the power supply rails were totally unforgiving). All in all it was improving almost by the month when the Intel stuff was new and they were adjusting the process.
Sounds like what dwave is doing
That's really cool!
Brilliant, I love figuring out old technology it's kind of like Reinventing the wheel, I don't experiment with vintage computers I usually work with Old Tube radios different technology same thrill, I've been a subscriber for a few years enjoy your channel thank you for your work!
I find it so entertaining and heart-warming that you have a bench-grinder just next to the electronic test gear. You are a true all around renaissance man sir. My lab is very much the same way (albeit smaller and messier).
Hehe. You never know, you might have an urge to grind a resistor ;-)
@@CuriousMarc Haha! I suppose there are cases where it might be necessary. The art of electronics never ceases to amaze and supprise me.
@@juliussokolowski4293 "Percussive, Abrasive Tuning"
Noticed that too. Just because I bought a bench-grinder 2 days ago. lol
I bought one of those core planes on eBay a while ago and it is in one of my display cases in my office. Allegedly mine came from a Soviet clone of a DEC computer. The LVDC drop at the end - great bonus and tease - can't wait!!
I am quite curious to see what it would show if you tried to read every bit.
Would require some massive circuirty but would be interesting.
They probably erased it in one way or another, but it should be interesting nonetheless...
Man, the more I watch your videos the more electronics click for me. Thanks a lot for making them! This is fascinating.
Ah, good old days when gods walked the Earth and Keysight was not even Agilent, but HP! (I still have somewhere in my attic a stack of those thick HP instrumentation catalogs and HP Journals, right from the time they decided to turn their instrumentation control computers into a general-purpose 1000 series minis.)
Wonderful demonstration!
Yep! Today equipments are becoming so annoying. I just saw that in this video the agilent scope shows the same annoying bug my has. It always enable the cursors when doing automatic measures !! WHY !?? You can see the cursor jumping around, over the green trace
Agree! That wiggling cursor that you can’t turn off is so annoying! Could use a few more knobs instead of Russian doll menus too. That said I love this scope. It is so responsive and accurate.
@@CuriousMarc Yeah! I love it too. I will record this annoying bug and send the video to agilent. Ops, keysight 🤦♀️A friend of mine have the new Touch version and it do not have the bug 🙄
@@CuriousMarc I just uploaded a video showing and I will send to keysight 😂😂
It seems in the good old days, HP made a tool for every conceivable need. Does keysight / agilent or any other have this breath of tools today?
"We have done like three months of research in five minutes" and 50 years of improvement in test and signal generation equipment :P
Love seeing how technology worked back then. Keep up the good work.
First, that is so beautiful. Secondly, I had always heard that we went directly from using vacuum tubes, right to using silicon chips; and this is a lot bigger I guess than silicon chip-but a LOT smaller than using vacuum tubes! Actually a very cool intermediary thing. Although it's pretty big for 1k, when I was a 17-year-old teenager the first PCs came out-and we were so excited that they were a whopping 64k!!! So, you can imagine stringing 64 of these together, and the box would be fairly large, but not that large. So, I think this is very cool.
"Here's the skookum version" AvE Reference? :D
Can confirm. Marc also uses “chooch” in other vids.
"So we have done like three months of research in five minutes, thanks to our good equipment here."
Too true! A worthwhile investment.
Nice video. I worked on a couple of commercially available devices, made in America by Sperry Univac, in the early 1970s. One was a terminal, Univac Uniscope 100, which had core memory for its display memory. The memory boards wore manufactured at the facility I worked at in Salt Lake City, Utah. The core was later replaced by integrated circuit shift registers. The other device that CuriousMarc might be interested in was a Univac DCT 1000 printing terminal. A bit of a beast, but it had a couple of interesting features. One: it had character memory to allow buffering. The memory was a glass delay line module made by Corning glass. It had two flipping buffers holding 160 characters each. That was real fun to troubleshoot. The other feature of interest was a built in modem. This was back in the day when you could only get modems from Bell. Univac designed and manufactured their own modems. They got a huge push back from the Bell monopoly for hooking this foreign device to the Bell system who would never admit to the defects on their system even though we had the proof.
Just a bit of nostalgia for me and thanks again for sharing and triggering that.
Ah, the Uniscope 100. You're probably very familiar with Engineering Procedure #1 to get a non-functioning U100 working: lift the unit an inch or so and drop it. That jiggled the PC board connections enough to restore good connectivity for a few days.
Some DCT series devices were used as console printers for several generations of the 1100 series systems.
@@stevew8233 U100 had edge connector board that were not gold plated, so the contacts corroded, hence the drop. The 'fixed' the problem by putting fish oil on the contcts. Oops, then the slippery boards came out of the connectors in shipping. Then they designed a clamp to keep the boards in.
Thanks for reminding me, I had completely forgotten.
Dear Marc, I've been a subscriber of your channel for a really long time and it amazes me how you are able to grasp the working and circuitry of such complex electronics so elegantly. Of course there is the documentation, but still getting so used to the circuitry as if you were part of it's development is truly amazing. I'm an upcoming computer engineer and would really appreciate if you have any advice or suggestions for a book that might help with developing a skill like yours ( if possible ) 😀
The Art of Electronics by Horowitz and Hill would be on top of my list.
Oh, how I love watching these videos. Some lives ago I spend two years studying electrotechnics at university, before my life went into a totally different direction. Trying to follow the subjects triggers some deep parts in my brain I used to use in those days. I recognize equipment from the lab or from piles in the corridors. I see schematics and diagrams I remember from books. Fascinating to realize there was I time I was supposed to sort of understand this all. Now I find it utterly relaxing and entertaining to simply watch it. Thanks a lot! Hope you will be given ample time to keep on doing this.
In Air Force, working C-5A navigation computers, we had core memory in both primary and backup computers. One of the technical representatives and I figured out the code to which memory core failed. We were always handicapped by depot only maintenance, and lack of computer parts. We started fixing computers by making one really screwed up computer with not one working memory module, and fix eight or ten and sending them back to supply. Depot got angry, but gave in as we could fix it just as good as they did
"3 months of research in 5 minutes" this is why I love these videos. Thanks so much for these Marc.
You make this so much more understandable than when I was in college. We had a core memory computer but it was only for use by post-grads. Thanks for helping out an old Computing and Electronic Graduate from just after the end of the Apollo era .
And now, back to reading the AGC Hardware book....
Some of us are old enough to have repaired the driver boards for these and adjusted the X and Y currents to find the Smoo curve. This curve tends to be a well rounded square of X and Y currents where inside, you don't have bit errors, and outside, you start having bit errors. We fought with one memory for a couple of weeks that would not stay working. turned out one of the carbon composition resistors was defective and would change value if wiggled. We upgraded and replaced the resistors with metal film resistors to fix the problem.
These are temperature sensitive. For reliability, they are most often kept in an oven to hold the temperature stable. If you watch the old Disney movie The Computer Who Wore Tennis Shoes", this is why temperature was critical to the computer operating. The small sense area is for the Parity. The array is one bit. A stack of 4 makes a 4 bit nibble and a parity bit in the same array, so in the same oven of the same core batch.
I had a lot of that stuff about ~20 years ago. It was parts from old soviet ES EVM (ЕС ЭВМ) industrial mainframes, equivalents of IBM S/370 (really it was not copies but mostly equivalents with fully compatible interfaces and functionaluty designed on soviet guts). Some modules had dozens of these ferrite planes. Back in the day I was young and had no idea that such stuff is valuable and used it only for parts salvage. Later I cheaply sold the remains. The only things left is empty cases which are still used as big flower pots by my mom in the village☻
Typical for barbarians on ruines of ancient civilisation.
So worth watching till the end. I’m so amazed at the ingenuity of scientist yesterday and today. Wow I’m just in awe of what we are capable of. Thank you and Bless you for your all the time and effort you put into this video. I appreciate everything. The “one more thing” just did me in bro.
Your last videos about the HP pulse generators now make so much more sense !!!
You aced everything with this video :-D
Reminds me of my days in the 70s at the University of Illinois where work was being done with Control Data Corporation on what was called Auxiliary Mass Storage. It was able to store the equivalent of 8MB (not really, CDC machines used 60 bit words with no parity; thank you Seymour Cray) that we were able to fit in about 9 cubic feet. We used to get memory errors in that bank of memory and our fix was to go over and kick the box low on one side -- there were mechanical relays that would get stuck sometimes.
Don't tell me it was Urbana, Illinois? That is where the vector computer graphics for Star Wars was made. This is before Lucas created ILM
Relays in your supercomputer? When was this made, when John Glenn orbited the Earth?
@@JohnSmith-eo5sp yup, that's the lab I worked in! Did the attack graphics on the PLATO IV system hosted on a CDC Cyber 73 and CDC 6500.
I should add that the 6500 had another claim to fame. We got it from McDonnell Douglas where it was the computer they used to run the engineering analyses for the Gemini spacecraft. Oh, and they were the predecessors of super computers, not ones themselves.
As I was watching in awe at what you've accomplished, you said *exactly* what I was thinking at 21:50. *Fantastic* video (as always)!
Lovely! According to certain "experts" on the internet, the Apollo guidance computer couldn't work because core-rope memory wasn't real. Hilarious. And what a beautiful LVDC core... all that yummy stuff!
haha! Funny how that works 😂
Mind that a core memory like this one was used as "ram", I don't think ram is the most exact name, but it does rougly the same job: stores data which will change during the mission (like istantaneous velocity or results of calculations); while the core rope stored "non changing" data, like the agc program itself and some constants, it was read only and used a similar physical principle of interaction between magnetic cores and wires but instead of only one you had more than 150 bits stored in each core by passing more wires in it, also because of this it wasn't possible to modify the data stored, it was literally... hard wired.
@@XMarkxyz That is exactly correct: core-rope memory is ROM (read-only, and hard-wired) whilst conventional core memory is RAM, random access memory for read and write operations.
Really? Of all the possible... objections? core rope? I suppose the film of it being made is fake, the modules in museums are empty, and Marc is actually a government agent installing 5G in our Yourtube feeds?
Actually, I had visited a factory and the lady workers were threading this ferrite cores ( at least 40 years ago ). Technologies and time flies.
Soviet Ferrite wasn't memory only!! At ferrite logik was made soviet computers for atomic rockets... Only 10 kHz but non sensitive to radiation, vibration and others...600 ferrite rings at small block are 600 triggers
I'm stunned that I understood about 20% of this. I should spend more time learning instead of gaming. It seems so unlikely that we arrived where we have in computing technology from the devices we did. The PCB is so amazing.
Thanks, Marc. This is excellent content for my core memory experiments.
super cool way to display these and it's great that you preserved the functionality
There's a brief mention about this type of memory in a book i'm reading, called an introduction to microcomputers.
As someone who knows nothing about vintage computing, everything you say usually goes well over my head, but I'll always come back for the pure aesthetic, space-age beauty of the technology.
Nice plexiglas work.... I often have display cases like this in my mind.... but they come out looking a lot more "punk" than this....
Purpose of the inhibit wire. I found this online somewhere.
"Many core memory arrays also have a 4th wire, known as the inhibit wire. Like the sense wire, one inhibit wire runs through all the cores. The purpose of the inhibit wire is to optionally provide current in the opposite direction to the write current, through the core being written. This will cancel out some of the magnetic field, and prevent the core from changing. Why would you want to do that? The answer is that it lets you easily stack multiple planes of core memory, all sharing the same drive wires. Without the inhibit wire, this would cause the corresponding core in each plane to be written. By sending the appropriate current down all the inhibit wires except one, we can ensure the write only affects the desired plane. The inhibit wires is an efficient way to scale up a system by adding multiple memory core planes. The alternative would be to make the matrices bigger and bigger, but eventually current losses and crosstalk make this impractical."
I think we bought the same 4k bit core memory! I also bought mine from Russia on eBay and looks identical to yours'. But I don't have the equipment you do to test it out. It's just a talking piece alongside my deck of punched cards and other paraphernalia
Same here for the 1kb memory! I hope to make a little display that uses a bit of their functionality.
If we all band together we can recreate a full stack :-)
@@CuriousMarc And dump whatever was stored on it when last used. Wonder if there are any Cyrillic "Hello Worlds" in there, lol.
I'd always wondered about old memory from devices from years gone by and seen them as big towers/nests of wires and magnets but this was a lot easier to understand when it was in a single plane. Nice
Magnificent work. I truly appreciate that you invest the time to share your efforts with the world.
Great video,, can I know what you use for producing animations?
That was done using a combo of Adobe Illustrator and DaVinci Resolve. Took a lot of effort. Resolve is very hard to use and quite buggy, but hey it’s free, and very capable, so it’s hard to complain about it. On the other hand the new Adobe subscription model is not even close to be worth the value for me, I mean by a mile. And they are buggy and overly complicated too. The Apple tools might be the right middle.
Oh man this was great! Thank you so much for going over the details of how this works.
Excellent video...!! I've been expecting a video like this on core memory for years..! Thank you..!
I've never seen Silastic applied so precisely and so neatly before :) 22:50
I got some of it....either way, I was hooked. Great video. I love old technology, ill be watching more. Thank you for taking the time to make these videos. True technology of old, compared to todays garbage. An I phone is lucky to be working after a year, this memory plate, as old as it is, was built to last. Amazing.
@6:51 the joy of a pair of repaired HP pulse generators... that was so sweet.
Brilliant lab setup - shows many principles of work - wish I had those during my engineering studies back in the 80s great scholary engineering video - thank you !
I still have the core memory out of a Wang 720C calculator I was gifted as a teenager (1980s). It had an intermittent fault and full of ICs and parts I could not identify. It had a ferrite transformer ROM too. To this day, I kick myself that I broke up the 720 as I’m sure I could have fixed it now, knowing that the ICs were often standard TTL with Wang part nos instead.
Память на ферритовых сердечниках))
интересно)
Th circuit you used for combining the pulses onto one wire is nearly the same as what you use to add in a stereo aux input to a vacuum tube radios pre-amp tube. Solid circuit.
Our PLCs where I used to work, were still using core memory right up until the 90s. The memory units were getting rare by that point, and expensive to service!
They used these in their knockoff of the DEC PDP-8/M, the Saratov-2. Also possibly the BESM-6 supercomputer. Another Soviet computer gadget to look out for are punched card readers, they kept on doing batch processing and other punched card stuff to the end of the USSR because nothing but tubes ever went obsolete in Soviet computing. I have seen Soviet punched cards for sale online dated "1987" - they weren't laminated like the IBM-Honeywell-Burroughs cards, just rough card stock.
Next 6 months: "How we fixed an apollo saturn V launch computer up."
Except the launch computer was absolutely massive
You only need the memory though hehe
It did take a fair amount of space in the Instrument Unit ring atop the Saturn V. There is the Launch Vehicle Digital Computer (LVDC) computer, multiple core stacks like this one, the Launch Vehicle Data Adapter (LVDA), and the Flight Control Computer (FCC) plus inertial platforms, sublimator cooling systems and telemetry gear.
It was a wow! episode, ended with a big woow! thanks Marc!
Roots of our computer memories, Read about but seen First time. Amazing!! Thanks.
11:02 The skookum version? Was not expecting the AvE reference, though I should by now.
Marc uses a LOT of AvEisms. This is definitely not the first time.
Just wondering - would an MR sensor, such as from a disk read head, be able to read the core magnetic flux? Small enough for high spatial resolution. Donor hard drive could even supply the read amp (?!) . Back in the day we played with some super-sensitive MR sensors that could read the mag field of a Nd magnet disc being flipped in someone's pocket some ten feet distant.
Nice. I saw another video where one was tested one and it still worked.
16:45 👨🏻🏫 Can you explain that circuit in more detail? 😍 How does it exactly work? 🥰 Thanks so much! 💕
I think it was explained somewhere during the AGC restoration? (might not be on a video)
Wow - that LVDC core stack should (should in very large finger quotes) still have its last contents, just like the AGC, right?
Marc, thank you. I recently discovered you and I've learned so much already.
Plain perfection Marc! Please go full steam ahead.
Great ! Very good explanation !
For those wanting to “See” the magnetic polarity in the core I’m afraid this wont work.
The problem is the magnetic flux is quite tightly contained in the core. Very little leaks out. So there is nothing to see.
redrok
Would a hall sensor probe be a viable option to safely read it?
Hi joob;
I seriously doubt the cores magnetic field would be visabe to a magnetic field sensor. These work because they work by sensing a magnetic field that has normal North & South poles.
Toroidal cores don’t have this. Their fields are oriented around the core. So they don’t have visible poles.
redrok
Wow 4kbit is actually about the size what the the bootrecord today is :-D there are games that fit in it so it would be funny to program a game in these things :-D
Are those furniture legs holding the pcbs?
Simply fascinating... I love your enthusiasm for basic discovery ... thanks.
Could you build a circuit (just transistors and stuff) that restores the bit after reading? I would be interested to see it!
Would magnetic viewing film show the ferrite coils polarity without affecting the memory? It would be cool to take a picture before and after the state change if it worked.
I think it would be fine and not affect it. However I am not sure it would distinguish between the two core states, as they are both equally magnetic.
@@CuriousMarc I would think the film would display a slight pattern difference in the orientation angle but It might prove interesting to see if you had a chance to post the results.
@@tootalldan5702 I don’t have film handy, but it’s very interesting suggestion, I’ll see if I can find some. Maybe we’d see a “twitch” when the core flips!
Fascinating as always then you drop the bombshell....
I can't wait....
Fantastic display cases!
That is a big stack of fancy equipment to write and read a few bits... Awesome!
A great demo! Do you happen to know, what machines they're from? I've used to have a module from ES-1020 (a low-end model from the early IBM/360 clones, that particular machine was decommissioned around 1993), and I think it looked like your lower-integration module, but it was a long time ago, so my memory is fuzzy :-) It would be interesting if they turn out to be from different machine series, say ES and BESM (I think the series SM that included clones of HP and DEC PDP-11 machines used the integrated memory from the start).
I actually don’t know much about Soviet machines and have no idea which ones these belong too. It would be great if someone could tell.
@@CuriousMarc Are there any markings on the modules? Those might be helpful in identifying the machines. BTW, there also were machines made in at least GDR and Poland, GDR was making quite a few of high-end IBM mainframe clones, so the machines might turn out to not be Soviet too :-)
The first module looks like the one from soviet computer "Promin" (unfoirtunately, no English traslation):
ru.wikipedia.org/wiki/%D0%9F%D1%80%D0%BE%D0%BC%D1%96%D0%BD%D1%8C_(%D0%BA%D0%BE%D0%BC%D0%BF%D1%8C%D1%8E%D1%82%D0%B5%D1%80)
The second one looks like memory from ES series (IBM Syetem/3xx equivalents), probably early models, but I'm not sure.
Fantastic ! Would it be more reliable than SRAM in the long therm ? I can't see many failures modes 😂😋
Only problem is the size and slowness, but the FRAM IC's have gotten past the size and speed quite well, using a very similar storage based on magnetic dots.
One way it "beats" S RAM is that it retains data on power fail.... old cores from the 60s MIGHT still have the data in them today.
Yes it is extremely good at retaining data for very long times. Doesn’t suffer from cosmic rays either.
@@CuriousMarc Wonder if they ever recover one from the watery grave, or from the lonely stage drifting out in space, if it will still be readable, containing the last instruction executed, which is likely a somewhat destructive to core memory of a jump back one word, if there is no actual halt instruction available.
@@MartianTech I guess it depends on the machine and the design. I remember row drivers as a failure point on the B 3500.
With core memory, are the magnetic fields strong enough that you can see them using some "magnetic viewing film" sitting right on top of the core array, or would you basically see nothing/no activity at all?
That is an excellent suggestion, I had not thought about it. My bet is that you would see them.
@@CuriousMarc yes as long as the grains in the magnetic viewing film were tiny enough to show the field around the rather small core donut.
Do we have any information about the Soviet textile workers / artisans who would have hand woven these core memory planes? I ask because here in the US there has been some documentation of the weavers working in the Raytheon factories in MA who hand crafted the core rope memory that was used in the computers that sent the Apollo to the moon.
"Bro may i borrow your hard disk i need one for school"
"Ok, but you have to find out how to connect by you self!"
The disk:
Have you done a core dump before working on them, just do see if there is some data on it?
On the second array; the blue sense wire which only addresses 1 quarter of the array: is it possible that this could be used for parity checking?
Don’t know, I also thought of it for a second. But it’s unlikely: that would be way too many bits, and they are also somewhat distributed all over the place.
I love the note! "generate with enough oomph" lols
Could you talk about keeping a binded labbook, your strategies, etc? thank you.
Start with ink, never graphite pencil, if you make a mistake one line through the middle, no on page gaps or pages, date each entry ...
Occasionally you can still find small containers of ferrite cores, which were made by VEB Mikroelektronik Gera, on Etsy.
check this - from 3:59 - that's about ODRA Polish Computer - he was sold to Soviet - there you have original movie from this memory assembling : ua-cam.com/video/aDvtmjWQrzM/v-deo.html&ab_channel=Wytw%C3%B3rniaFilm%C3%B3wO%C5%9Bwiatowych PL language only ..
Thanks for the link. A completely manual assembly process for core memory, no fixturing to hold the cores, that’s crazy!
@@CuriousMarc thank you very much for your movie. For your work. Without this video, was for me impossible to understand how this memory works. thank You Sir.
It looks so easy when you show how you do it !
Destructive reading, wow
We do destructive reading of RAM all the time up to now. Not only core memory uses destructive read, but dynamic RAM (as the core technology of all DDR RAM today) uses destructive reads, too. The main difference is that dynamic RAM uses electrically charged capacitors to store the bits, whereas core memory uses "magnetically charged" cores. Magnetism and electric charge are dual to each other: Magnetism is related to to current flowing, and you need to apply a voltage to change the magnetization. On the other hand, electric charge is related to voltage between some points, and you need to have current flowing to change the charge. That's why you apply current to the core memory and monitor the voltage, whereas for usual DRAM, you apply a voltage (0V, i.e. you discharge the capacitor) and monitor the current (to find out whether the capacitor was charged before.
You don't need to care about the destructiveness of reading DRAM, because the fix for that is integrated in the DRAM chips. When you "open" a page (also called a "row"), all capacitors of that row are discharged, and the measured previous charging state is recorded in the "row buffer". When you "close" a page, the row buffer is written back to the row. Early DRAM chips (i.e. nearly everything before SDRAM) "opened" a row when you pulled /RAS low, and the address pins told the chip which row to open, and "closed" the row again when /RAS went high again. Thus there is no way to forget to undo the destructive read, because one row automatically gets closed (on the rising edge of /RAS) before another row can be opened (with the next falling edge of /RAS).
The sense wire maxim... flippin’ eck... so long and thanks for all the memories!
It's Soviet made, so it will work forever (as long as it is intact).
Thank you for this nice and correct explanation!
You should put some of that magnetic viewing film over the cores to see them flip that way, that'd be awesome.
That would be very cool indeed. I actually tried. Unfortunately the magnetic field is many orders of magnitudes to small for the regular film to work. I tried with much more sensitive oil-based viewers instead, and that did not work good enough either (way too slow), although I saw a hint of something. Then I tried to contact some professional outfits that do magneto-optics films and cameras. That would work, but that's many 10's of thousands of $ worth of equipment and they were not interested in sponsoring a loan.
Hold on... You can store -1, 0 and +1 per core?
In school for my certifications, we had a few boards of core memory as examples.
They were like 2x2 Ft, about 3000 cores on it.
It's crazy when you remember they were hand wound.
amazin work... thank you marc
Looks like some interesting things to come....
So how fast was that 1K plane? It looked like maybe 4x slower than the 4K cores, so somewhere around 10 uS?
It looks like 6 us on the scope. However you could still clearly see the core flip in-between the pickup spikes. If you had a gated sense amp, you could just restrict the sense to that time window and make it run almost as fast, at the expense of an even more complicated setup. Unfortunately my sense amps lack the gating feature.
Thank you for explaining this! 2:48 this may be the quantum mechanical wave behavior we've always worked around (destructive read means "the act of observing changes the state"). Spooky stuff!
I have a multi head 3D printer. But I do not know how to 3D design. Is there a way to create a fully 3D printable version of one of these, if someone has a multi-head extruder? How to create the iron toroids, while lacing conductive XY traces. The wire is magnet wire, coated, right?
8:02 beautiful pulse!
Is the sense and inhibit lines fed like a string through each toroids for practical reasons, or could you route it through all of them in parallel and get lower latency?
A long time ago I used at scool a french computeur
Mitra15 from CII Onywell Bull who usees ferite core memoy
I think my dad worked on such machines too.
how exactly do you specify the adress of the bit? wouldnt you need as many wires as there are rows/stores?