Core Rope Memory Built and Explained - F-J's Physics - Video 169
HTML-код
- Опубликовано: 16 ноя 2024
- With the renewed interest in the Apollo missions and the associated hardware, especially the AGC (Apollo Guidance Computer), here I explain the amazing technology that is Core Rope Memory and build a working example! I must stress that this is 'conceptual' and massively simplified but it does demonstrate some of the principles used in Core Rope Memory without getting into the complexity of actual production versions. In fact you may wish to look up IBM's 'Transformer Read-Only Storage' that is very clever and works in a similar way to the example I have built! Correction: I do talk of the cores as being really small but I am of course confusing them with those in Magnetic Core Memory! Those are the really tiny ones!
Have a look at Magnetic Core Memory Explained (Video 179) • Magnetic Core Memory E...
Please consider supporting my work by buying me a coffee at
www.buymeacoff...
Very many thanks, F-J
Please consider supporting my work by buying me a coffee at
www.buymeacoffee.com/francisjonesa
Very many thanks, F-J
dont reinvent the wheel by changing names ... Magnetic core memory and rope core memory are two different things ... no point making rope core memory into 2 names ... you dont call an electrolytic capacitor a liquid filled capacitor and electrolytic capacitor ... its JUST the latter and it is NOT a ceramic capacitor ... stop being a dumb american ...
HI ANTHONY THAT CIRCUIT IS A WORK OF ART IT REALLY IS THANKYOU SO MUCH
Patrick, thanks so much and glad you liked it. As you know I was trying to design something from scratch so it may not be the best kind of circuit to build but it works, is a bit of fun, and looks quite good too! Thanks for watching and taking the time to comment. It was a fun build!
YOU ARE VERY WELCIME I LOVE YOUR WORK SO MUCH THANKYOU
Excellent simple explanation of how this type of memory operates.
Thanks Anthony. I realise it is a massive simplification in many respects but it was a fun build and forms the basis of how this kind of memory works. As I think I said, the Apollo Guidance Computer version was similar in some ways but different in many more! Glad you enjoyed it!
The crazy part is that there's people recovering data once thought lost from ROM core rope memory right now. Despite it being more then 50 years later the data on the memory modules was not only preserved it's fully recoverable. The system was even more robust then the original designers realized.
Thanks - yes it is true that in most cases the coercivity and remanence of the cores is such that they will retain the data indefinitely. Versions that use my method retain the data for as long as the system is in one piece as it only depends on the location of the wiring. As far as the original designers are concerned, that's interesting to hear. I do not have sight of how robust they thought the systems were but from what I know the designers knew they were pretty robust as they were cleared for use in space. I will be doing a video on core memory soon - that too retains its data but one has to be very careful reading old core memories as you have to immediately make sure you write the data back or it is lost from the cores when they all flip to zero!
no they knew the hard wired data was always gong to be there as long as the wires didnt break ... READING it can be done only two ways ... electric poling of working modules or physical tear down and logging of where EVERY wire went ... as they want to preserve the history they used a home made circuit to poll the memory by stepping through every possible address and reading what is there ...
@@AnthonyFrancisJones btw .... you arent sending any AC signals ... you are sending PULSED DC signals ... which is NOT AC ... you dont have a wave rectifier circuit to convert the dc into ac ... you just have dc power being pulsed on and off very fast (abut 60khz you said) if your dc is gong from 0 to 5v t o to -5v then it is ac ... but since you are using 555 timers you are just pulsing the dc between 0 and 5v positive ... not one place in that 9v wired circuit is there an AC signal .... from source to display it is ALL DC the clock side is high pulsed dc and the working side is selected DC by reading the rope core memory and acting on the data contained there in a certain way ... in your case you light a segment of a display ...
@@AnthonyFrancisJones Indefinite is kind of a tricky word when it comes to memory storage, after all, it was widely assumed that laser disks could hold onto data indefinitely only to find out the metallic layer used for data storage oxidized and that they didn't last nearly as long as intended.
I have no doubt every engineer assumes their designs would hold data basically forever but there really isn't that many that actually can. My comment was more meant as a compliment that their ideas worked as intended.
Yes, you are absolutely right! It worked as they intended and as you suggest, even better than expected! A kind of modern day 'Victorian engineering'! By the way, I am tired of the number of my CD's that have packed in and my Ampex reel to reel tape too. My vinyl records are still going strong!
This was excellent.
It zeroed out all yhe ambiguity I'd previously had on the operation of these mechanisms.
Thank you.
Great! Glad it helped! Remember it is a bit of a simplification but it is a good place to start!
Great video, clear yet detailed
Thanks! Glad you enjoyed it.
Maybe someone else can clarify, but afaik the Darlington array and even the BFY51 transistors are really only needed because there needs to be enough current on the other side of the transformers to drive the LED pairs for each segment of the 7-segment display. If you want to just read the binary data from the core rope memory you probably dont need nearly as much current and can therefore trim a lot of the current boosting transition circuitry
Good question and I hope this answers this. Remembering that each core is used as one bit of data and it works as a transformer, you do need an AC signal to make it work. I use the transistors here to switch this AC (square wave) on and off depending on whether it is a lit or not lit LED that is needed. They LEDs or the cores cannot be directly driven from the 7 segment display chip if that makes sense. That is why there are two 555 timers - one to scroll the digits and one to proved the cores' high frequency signal that gets switched on and off. Hope that helps and thanks for asking!
I absolutely love this. You made my morning. 😀
Excellent! So pleased you liked it and I hope it all made sense. Yes I know it is a massive oversimplification but it was a bit of fun and I am quiet pleased with what came of it! Thanks for your kind comments and support of the channel. Hope you enjoy some of the other videos!
Great explanation of core rope memory! Thx
Thanks, I know it is a oversimplification but I hope it's a good place to start. Thank for watching and taking the time to comment.
Really nice project and yes, Curious Mark is one of my fav channels too.
Have you considered installing a second multiplexer(s) on the bottom right pcb and 10 toggle switches, or 4 sets of 3 switches for Octal switching to allow one to enter most any character or symbol from the ASSCI set. Then you'd have an excellent training tool for others to generate any character/numeral they desired.
I was with DEC, Digital Equipment corp, for years and remember well the core mem bds used on the PDP 8 and other earlier PDP systems. Even with 32 K of core and some massive CDC disk drives with a business with 30 to 40 terminals being used intermittently worked great, However, those poor hard drives took a daily beating doing all of the IO now relocated to mostly huge arrays of RAM. To their credit, those CDC drives could run for years wo any issues as long as the absolute filters were changed and heads were cleaned frequently on a PM schedule.
The 70's -90's were great years to watch the computer industry evolve to the appliance computers we all enjoy these days. Shame how only a small percent of people have any understanding of what happens at the low levels where 0's and 1's still make it all play when you order a new widget on Amazon or watch some video on one of the movie services...LOL
Glenn, thanks for taking the time for such an interesting story. Yes you are right. Those of us that were brought up with computers that had visibly moving and interconnected parts have a much better understanding of what happens 'under the bonnet' (hood!) as it were. I remember as a pupil using the school's ICL mainframe and doing 'dir' requests and hearing the hard drives accessing a room away as the heads moved in and out on the massive drives and disk packs! Really good idea of yours to put switches onto the project to use it to demonstrate ASCII characters. I think I want to leave this board alone as it is just hanging together! I would probably break something! Have you seen my other videos on the magnetic core memory I have? I thing there were links in the description. Thanks again, it is comments and experiences that people like you share that make these objects come to life. Many thanks again, Anthony F-J.
Congratulations!!
Pleasure!
If I understand this concept correctly, the cores aren't actually the memory. The cores have nothing to do with the storage of the bits. The storage is formed by whether or not a wire passes through a particular core. The core is just a way to connect the "read" pulse to the "bit" pulse... or disconnect the "read" pulse from the "bit" pulse, depending on what bit is stored there.
This is actually a pretty clever way (for the era) of storing permanent information, as it acts just like a ROM. And the number of "locations" the ROM can store depends entirely on how many wires you can fit through and beside the ferrite cores.
In summary, "Magnetic Core Memory" was a type of RAM while "Rope" memory was a type of ROM.
That is an excellent explanation! I think it is RAM too as you can directly access the bits in any order without having to go through pervious addresses or location (like tape machines had to). What do you think? I feel it is both ROM (data 'burnt in' and fixed) unlike regular core memory, and RAM as the data can be randomly accessed. Have I misinterpreted this? Thanks for a great comment regardless!
@@AnthonyFrancisJones Technically, all current-day ROM is also RAM but for convenience, we use the two acronyms to distinguish between read-only memory and read-write memory, despite ROM also being random-access memory.
Great. Yes, that's it! Though I would argue that historically punched tape/card was ROM but not RAM which is why I guess I used the expression here but as you say things have moved on!
Nicely explained, thanks 👍
Thanks and glad you liked it. As many will say it is a long way from the Apollo AGC but it gets you started thinking about the world of core memory. I must do another video on a more representative version of the AGC memory method. Thanks for watching.
Very awesome. Thank you sir!
Thanks - glad you enjoyed it.
Can you provide the part number or the specs for the ferrite cores that you used?
Sorry for the slow reply. They are not very critical - I just went for ferrite cores that would be large enough to show up on the camera so I bought these off ebay about a year ago. "10pcs 22.5x13.8x10mm Ferrite Toroid Core Transformer Inductor Coil Ring, Black." Hope that helps and if you build something do let us know!
This is a very clean demonstration of pulse-transformer core rope memory. Do you think we could get a demonstration of switching-core core rope memory, such as were used in the AGC for instance? That technique is a bit more involved, but has the advantage that you can actually handle the address decoding directly in the weave with inhibitor lines. It's pretty cool.
Thanks, really glad you liked it. You are right. I need to look into doing another one more closely related to the AGC method of using the cores. I have some core planes which I must demonstrate too. Would be quite a lengthy build! Thanks for watching and taking the time to comment with an excellent suggestion too!
Keen to do this but having a bit of a problem finding a material for the cores. Suggestions welcome. Don't want to take apart any of the core planes that I have and the cores in those are a bit too small to show on video easily. I will get there I hope!
Yes, that type of core rope memory is very interesting because it uses the cores as a clever address decoder. I estimated that if the decoding were done with the AGC's standard IC (a dual 3-input NOR gate) it would have taken thousands, much more than the computer itself. It can be tricky because the address lines need to drive the cores with enough current to saturate them so they won't pass a readout pulse to the sense wire (that may or may not run through the core).
@@philkarn1761 Thanks Phil, that is an interesting way of looking at it!
@@AnthonyFrancisJones It took me a while to realize this myself, because everybody kept comparing core rope memory to conventional read/write core. They're absolutely entirely different things, and the comparison only confuses people. So don't even mention R/W core except to say "forget anything you know about it, this is totally different."
Each core in an Apollo core rope is just a magnetic OR/NOR logic gate with LOTS of inputs, i.e., the address lines (or their complements). Way too many to practically implement in the silicon of the 1960s.
What was going on with the upper segment? It had around 6 different levels depending on the displayed number.
Well spotted! So this was tracked down to the LEDs being the wrong way around. Now this should not matter but as the signals to them are AC and not symmetrical in shape (like a sine wave would be) they are getting under powered compared with the other ones. So I have swapped them around and another segment that was doing the same and all is well. I think that must have been the cause! Thanks for watching and hope you enjoyed the video.
QUESTION what is the blue box above your tube tester? I'm going to guess it might be a rotator control box for an antenna. Just curious.
That's a really good question and your guess is way out!! It is a old motor driven, relay controlled school (Griffin) centisecond timer! I should do a video on it! collections.st-andrews.ac.uk/item/centisecond-timer/1005577
that's so sick, ima build one that flashes my area code
That's a great idea! Let us know how you get on!
Absolutely amazing. I can't emphasise how much I love your videos ❤ I love CuriousMark and his video style! I think I first seen this type of memory on Look Mum No Computer and it was used on a automatic phone dialler I think... I can't remember now!
Have you ever considered doing a video on gardening or cooking tips and tricks 😆 I think this is the only subject area I haven't seen you cover that interests me 👨🌾
Great - glad you liked it. Yes, it was on the phone dialler on LMNC and funny as I was building it when I found that video and the one where he uses core rom for his drum machine! Cooking and gardening... there's a thought! More physics ones to come and thanks for being a loyal supporter of the channel!
❤1995 remember working on a project for elevator that used core memory that was a nightmare, the onwner wanted it to work. I remember †he magnets which held info. It worked but every other 6 months , he called. Today, I surf more. Califronian Surfer Downey Cali
Facinating! I had never heard of that application but I guess a lift (elevator) used repeated common commands. I guess they used lots of cam relays too. Thanks for taking the time to comment.
Hi, thanks for the video, a very interesting project. I have some thoughts/questions on your circuitry and would be interested in your feedback.
1. I had the same observation about the NPN emitter as a few others have pointed out. I wonder if your "a" segment is weak because it relies on the breakdown voltage of a specific transistor which varies across the batch. I have a hunch that if you were to swap the driver transistor for the "a" segment with another segment, the problem would move with the swap.
2. In the comments, you mention that when connecting the NPN emitters to ground you weren't able to get it to work. I see that the coil input is AC coupled so it has a very weak DC reference. I think the purpose of the cap is to serve as a high pass filter (differentiator, pulse detector, etc..). But I wonder if you added an appropriate resistor downstream of the cap, if that would resolve it. Because with very small resistance in the coil driving wire, the cutoff frequency may exceed the 56k and cutoff your signal entirely. The impedance of the reverse biased transistor may have accounted for that which is one possible explanation for why it worked in the unintended design.
3. I was wondering if the 1 amp BFY51's are truly needed since the Darlington drivers are 1/2 amp. It certainly doesn't hurt to have the headroom, but the m54562p's are designed to drive coils such as relays. Granted though a single "turn" is a bit different and I think it depends on frequency. It would be interesting to know the current draw of the circuit as that would help answer this question.
4. I think there's no reason why a ULN (low side) driver wouldn't work. High side drivers tend to be a bit more specialized and harder to find.
Thoughts? (thanks)
Thanks for these brilliant observations and it's things like this that make the positive side of RUclips so enjoyable! So a bit of background first... I did not have any circuit diagram of what this in advance, I just built it as I went along from scratch - which is what I like doing. The overall result being something quite new to me - that's the fun. Looking back there is much I would have changed and I learnt a lot on the way too, which is why I like doing this sort of thing! So to Point 1: I have got it fully working now. I had got a few of the segment LEDs back to front and it seemed clear that the AC signal was not symmetrical. Turned them around and they are all now equally bright. Point 2, I think you are right there. I was new to using transistors in reverse active mode and discovered it 'by mistake'. It is interesting that those are very old transistors and newer ones I bought did not work at all (or were they fakes?). Point 3: I am sure you are right about the darlingtons. I had been using them quite a bit in earlier projects and the idea had sort of stuck! I find it hard to imagine those little chips passing quite a bit of current but of course they can! ruclips.net/video/r77hpJ5bUu0/видео.html and ruclips.net/video/srko8SEDpY4/видео.html Point 4: Agreed! I just went with what I knew! In summary, it was a lot of fun to build and I never really knew if it would work but we got there in the end. You are totally right (if I knew a bit more electronics - getting back into it after about a 30 year break) I would make version 2 differently! All of this said, thanks so much for taking the time to comment and sharing your experience and knowledge. Sometimes when building things you wish you were in the same room as others so you could ask for advice and bounce ideas off each other. Hope you enjoy the other videos and do keep the comments coming!
@@AnthonyFrancisJones Thank you for your quick and informative reply, it really clears everything up! After seeing "Look Mum, no Computer"'s videos on rope core memory, I discovered this video and I found your design to be simpler. I'd like to experiment with this just for fun and also be able to say that "I hardwired my computer program" ☺. Glad I discovered your channel, subscribed, and look forward to exploring your content. Thanks again!
Excellent! Glad that helped. You might note that I have some 'nice' but old test gear on my desk. We tend to surround ourselves with such stuff but actually rarely if ever use it so part of this project was to encourage me to use some of my less used and less understood test gear like the Marconi Instruments Universal Timer Counter. It's like buying an old car as an excuse to use your spanners! All part of the fun. Let us know how you get on.
20:33 - I'm really curious how they did this for Apollo 13..... Did they have discrete transistors stable enough then? What about the chips?
In many ways quite different in the AGC set up but this video was designed to get one started on the idea of a core rope type memory. The AGC ones were very complex and not all the same on each launch. If you want to know exactly how it all worked head off to the amazing Curious Marc channel and be astounded by how well they deal with original memory from the actual missons.
@@AnthonyFrancisJones
That's fair. I'll go over there and have a watch :)
Yes, you will be astounded!
They had the first integrated circuits, OP-Amps in little cans. Where this uses an AC signal to brute force drive the LEDS, the AGC was sensitive enough to pick up a single shot of magnetic decay! Much faster and used much less power. For its time it was truly exceptional.
Yes, it was truly lightyears ahead of its time - complex though but very effective!
What A Great Tutorial Sir, Can I Ask Why you used M54562P IC? The 4017 IC Has Enough Power To Drive The Transistors?
Ali, thanks for your kind comments and great question. So, as with much that I do, I did not overthink it or do much in the way of calculations but I felt that as a 4017 is really a voltage device and transistors are current switched devices it would be a good idea to use a darlington driver (M54562P) to make sure that there was enough current available to switch the transistors. I still cannot remember why the I decided to wire the transistors in what looks to be a back to front way, but it works and that is the main thing! I tend not to design and then build but build as I go along and tinker until it works - seems to be more fun! Glad you liked it. I need to do a video on regular core memory at some stage too!
Excellent video. Thank you. I have just one question, why is emitter of NPN transistor connected to +9V and not to GND (0V) in your schematic?
Thanks for the question - yes this is a bit of an oddity so this is what I have said before, hope it helps.
Agreed! I wired it up and got a transistor the wrong way around and it worked!! I was mystified too by this. I have been over and over it and everything is as per the schematic (unless these transistors are not following the agreed layout of pins for this type of base). I do remember that I did work it out at one stage and it was partly due to the fact that they are switching a rather odd shaped high frequency signal so are behaving in a rather odd way. Not an explanation I know but it works and I was equally mystified!
Have a look at this great comment from Mick Healey - he has nailed it! "If the circuit is actually built as shown on the diagram at 18:56, then the transistor is working in 'reverse active mode'. In this mode, the transistor will operate, but (conventional) current flows in the opposite direction (emitter to collector) rather than collector to emitter as would be the case in active mode. A transistor will work in this mode, and even have some gain, but the gain is low. To enter reverse active mode, the emitter voltage must be greater than the base voltage which in turn must be greater than the collector voltage. That exact scenario is happening here. The base resistor ensures that emitter is higher than the base. Since the collector is connected to the 56kHz oscillator, then the base will nearly always be higher than the collector."
Thank you for this answer, it was quick and it explains "odd behaviour" of NPN transistor. I am trying to replicate the design with some changes in high frequency part. I am using 555 at about 65kHz with LC circuit on output, and I managed to get nice sine wave with the same frequency as input and voltage about 15V (my power supply voltage is 6V). Now I am struggling with core windings because the highest voltage I am able to get on the secondary is approximately 900 mV. We shall see how this will end. Also I removed M54562p from the circuit just to simplify it if possible. I am a bit in the dark here, but maybe it will work. Thank you again on awesome video. @@AnthonyFrancisJones
That's brilliant - let us know how you get on - sounds like you are on the right lines. You may need something to boost the current on the output stage though if you want to light an LED. Less than around 2V is too small. Post again when you are further and perhaps start a new comment as RUclips is not good at pointing me to comments that have new answers! Good luck!
Thank you, it seems that my circuit is missing M54562p after all at the decade counter output. Supply voltage to "slow" 555 timer is 6V, output on pin 3 of 555 timer is about 4.5V and output from decade counter is about 2.2V. Now, when I put M54562p it in the circuit, M54562p output is exactly 6V (as supply voltage). Besides, it is interesting to watch transformation of that square wave on the oscilloscope. M54562p eliminates all noise and odd spikes visible on the decade counter output. @@AnthonyFrancisJones
Great - sounds like you are having real success, and like I do, finding out things for yourself rather than just copying other's work. Great way to learn!
Have a BOM and maybe a schematic for this anywhere?
Sorry, I don't have a BOM for this but the parts are all indicated on the circuit diagrams I show in the video and it is a fairly easy thing to copy now I have done the legwork of getting it to work! Sorry not to be more helpful.
29:51 - There's a mistake here that you correct before -- i'd add text in the video... but you were supposed to MISS E and INCLUDE G.
Yes, you are right! I swapped the letter labels around and that led to the ensuing chaos! Corrected it in the final build once I had powered it up and found that it was not displaying recognisable characters! The joys of filming things fairly 'live' as I build them and perhaps 'rushing to publish!' As I have said before, most of my mistakes are caused by rushing to get something done to share my enthusiasm! Well spotted by the way and thanks for watching.
cool, keep it up. 👍
Thanks Frauke. I had great fun making it! Doing all the colours added to its look as well I think!
Thank You! Amazing!
Pleasure Robert and glad you liked it. Fun project!
Im pretty sure 555 can't produce AC signals, its producing pulsed DC signal with probably 50% duty cycle. Correct me if im wrong.
Good question - yes this is normally the case. However, with my circuit I am using the small changing AC current from the 555 timer to switch the transistors. As an aside you can change the duty cycle of the 555 to suit whatever arrangement you want. I am using it in a rather unusual way here but this is the beauty of the 555, it has so many uses other than just producing square wave digital pulses!
A 5V p-p square wave biased ar 2.5V is still an AC signal. Just biased around 2.5V instead of 0V.
Thank you, it seems that my circuit is missing M54562p after all at the decade counter output. Supply voltage to "slow" 555 timer is 6V, output on pin 3 of 555 timer is about 4.5V and output from decade counter is about 2.2V. Now, when I put M54562p it in the circuit, M54562p output is exactly 6V (as supply voltage). Besides, it is interesting to watch transformation of that square wave on the oscilloscope. M54562p eliminates all noise and odd spikes visible on the decade counter output. @AnthonyFrancisJones
Great - well done! This is a very useful chip. I love Darlington drivers and I feel they have been a bit forgotten by people who only understand normal CMOS and TTL logic but there is a real place for them in many circuits. By the way I think you can go up to about 16V (I use 9V a lot) for your 555 etc. That helps too. Don't get hung on the idea that everything needs to be at 5V!
Thank you. I am usung 5V just because I replaced transistors with relays because of possibility that I will use separate AC source for driving core rope memory (if my 555 timer driven AC generator does not do the job) and control voltage for those are 5V in my case, also they have that "vintage" look. 🙂@@AnthonyFrancisJones
Just about to make a video on relays and I have some massive examples from the railways. Also will be doing one on logic gates using relays too. Good luck with your project!
Did I understand correctly that your circuit uses not alternating current, but a pulsating current with a frequency of 60 kilohertz?
Yes, I just created pulses at a high frequency that seemed to work. As you can see from the CRO images the pulse shapes became very distorted but it was a case of play until something worked rather than copy a known system.
Pulsating DC on the input of a transformer will generally become an alternating current on the output. The waveform may be distorted, but it does work. I'm wondering if wiring the two top LEDs (which seem to be dim on some numbers) in the opposite direction will brighten them up, since they're probably rectifying AC to DC and the AC may not be symmetric due to the input being pulsed DC. Wiring them in the opposite direction may give them more current.
Yes, you are right about the LEDs. I did wire them the other way around the other day and all was perfect! Well spotted!
I didn't understand one thing: The emitters of the NPN transistors are connected to the "+" of the power supply and not to the GND? It makes no sense to me.
You've got me! I made this wiring mistake at the start and it worked! I then corrected it and it did not work! It is strange for sure! I did work out why it worked but would need to look back to remind myself too. Well spotted by the way!
@@AnthonyFrancisJones Now I'm trying to work out what's going on, some unexpected properties of those transistors must be coming into play
@@matthewday7565 if the circuit is actually built as shown on the diagram at 18:56, then the transistor is working in 'reverse active mode'. In this mode, the transistor will operate, but (conventional) current flows in the opposite direction (emitter to collector) rather than collector to emitter as would be the case in active mode. A transistor will work in this mode, and even have some gain, but the gain is low. To enter reverse active mode, the emitter voltage must be greater than the base voltage which in turn must be greater than the collector voltage. That exact scenario is happening here. The base resistor ensures that emitter is higher than the base. Since the collector is connected to the 56kHz oscillator, then the base will nearly always be higher than the collector.
Mick, that's brilliant! I did explain it to myself and then forgot that I did understand what I had done! You have captured it beautifully here so thanks! I did not do any calculations just got on and built it and when I miss-connected the transistors from the start, they worked, then turned them the correct way around and it didn't so I went back to this strange but interesting arrangement. I think many are going to question my logic - pun intended!
Nice video - but it doesn't store so you couldn't call it "memory". But it is a nice example how to "manually" wire a 7-segment decoder. I also did it in the past but using diodes.
Thanks Bernd, this is a form of memory we call ROM as it does store data in its wiring structure but and each transistor can independently address each stored bit making it a form of RAM. Unlike decoding a seven segment display input with diodes (which I do in this video ruclips.net/video/hNkMNqXJQhU/видео.html ) in this case there is no electrical connection at all between the driving circuit and the display, only a magnetic one. Have a google of Core Rope Memory and perhaps with respect to the Apollo AGC and you will see what I mean. Hope that explains it. Many have not seen this technology as it was a very early from of ROM but an excellent one for lots of reasons. Thanks for watching and taking the time to comment.
Sure interesting to see, and a bit of a flashback for me :D I wish I’d kept the ferrite for the magnetic memory. I’d have sent you some. The cores I used were ensued from someone in Bolivia of all places. Anyway, you make a better video than I can!
Thanks Brek. It was a fun build and I know a bit of a massive oversimplification! I have really enjoyed watching yours and seeing just how much work goes into your builds! Must do some other core memory stuff at some stage! Thanks for taking the time to watch and comment.
@@AnthonyFrancisJones Hi :) Brek’s autism here. Does the rouge display element have both LED anodes or cathodes tied together?
Thanks Brek. I need to have another close look but I did check before and all seemed fine. It is strange that segment A lights perfectly well for some numbers - 7, 0 but not for the others. I have just noticed that A does not light properly whenever segment G is lit! I will have to dig deeper! Keep the suggestions coming.
@@BrekMartin Had a look at it finally tonight and I think your diagnosis was correct. Even though it is driven by AC I think I had some of the LEDs the wrong way around - swapped some anodes and cathodes and all is well now!
@@AnthonyFrancisJonesHi :) Brek’s narcissism here. I like being right, so this works well for both of us :)
This is cool and all but you should have all but started with the schematic.
Also should be clamping those transients and the "negative" component of the output to protect the LEDs. The reversal and the kick is likely what is causing them to fade out over time. At the very least clamp the reversal. They may "tolerate" it, but it is bad for them, especially when the peak voltage is high. They clamp themselves reasonably well in Vf but the reverse wipes them out.
Thanks for this, Yes, you are right about the schematic. I tried to explain what I did as I went along but now realise the more of these types of video I do that a full and clear schematic would be a great idea. Thanks too about the reverse biasing of the LEDs. Very good point! I learn a lot from the comments too! Thanks for watching and taking the time to comment.
aw yes, I remember the PDP8 core mem
Thanks Cliff. Yes so many have happy memories of working with that machine. I do plan to do a video on how that kind of core memory worked - I just need to get on and do it! Thanks for watching.
I thought the point of core rope memory was the 2D addressability. Your 1D version seems a lot more complex than just directly wiring each address line to the appropriate data bits on the output.
If you wire each address line to the appropriate data bits on the output, you wind up shorting all of the address bits together via the data bits so that any address line input will cause all of the 1 data bits for every address to go out the output at the same time so there's no way of knowing what the correct data actually is.
You can get around this by using diodes to prevent the address lines from feeding back into other address lines via the data bits, but then you need a lot of diodes. The core memory method stores the bits by whether or not the address bit goes through a particular core, which can be more compact. Of course both methods are obsolete due to integrated circuits which do the same thing in far less space.
Tamas, Melkior's reply is excellent. It is the rugged nature of these memories that make them so effective. Back in the day a fault on a single diode could spell disaster as could a cosmic ray strike but here the data is the wiring and the cores enable it to be sent where it is required. The seven segment display is a bit of a red herring here. It is just to show the data output. I could have used a logic analyser or similar. Hope that helps and thanks for watching,
@@melkiorwiseman5234 thank you for the detailed response, these are very good points! I suppose adding a bunch of diodes was a very different proposition back when this technology was trending :)
@@AnthonyFrancisJones Thanks for expanding on the previous already excellent response. In any case, please also make a video with a 2D addressable circuit :)
Thanks. I do plan to do a video on x,y addressed core memory and I even have some 3D core 'cubes' too!
Genial !!
Many thanks!
But how do you know if its a 1 or 0? dose it act like a inducter? How dose the computer know if it has changed?
Good question. So I had better explain. The circuit on the left act as a clock, stepping though each line of code. A word of code is encoded in each wire depending on whether it is in a coil (1) or out of a coil ring (0). The secondary winding is driven high if it experiences a 1, wire through it. In this case I have used each 'segment' of a seven segment display to indicate if a bit value was high (1 - on) or low (0 - off). Those wires could of course gone to a CPU instead so it could execute that line of code. Or to memory etc. Hope that helps! Note that this has 10 words of memory (the ten different coloured wires) each containing 7 bits as the are 7 coils that each wire passes through or around.
And, yes, the rings have a primary winding - the wire passing through it, and a secondary winding - the red coil that goes to the seven segment display output. So they are really transformers which are a type of inductor, but inductors usually only have one coil not a primary and a secondary.
@@AnthonyFrancisJones ah but how dose it know if say only ring A is magnetized and not ring B and C?
@@CLOCK-WORK I think you may be getting confused with Core Memory where the rings are magnetised - have a look at my video on that one. ruclips.net/video/T8VreNzldWc/видео.html In Core Rope Memory a pulse of current in one wire if it passes through a core will result in a pulse out of that core - a 1. If the wire has a current pulse in it and does not pass through the core then the core secondary winding registers a 0. Does that help? The cores don't have any memory in themselves.
@@AnthonyFrancisJones yea but how dose the computer read that information when it's set do you just pull it to cycle the code?
I thought the rings were made from "molly-what's-its-name" and had isolated layers to reduce eddy currents?
Not sure if we are thinking of the same thing here. Transformer cores are made of a laminated series of cores to reduce eddy currents. In the care of the rings, they are made of a fairly homogeneous ferrite typically cobalt iron oxide. Were you also thinking of molybdenum steel?
Hi, yes molybdenum steel I think was the term. I thought the cores were made out of a laminated version of this material for the Apollo mission. Maybe I need to go back and watch Curious Mark again to check what he said. Edit: Molybdenum Permalloy was the material used in the Apollo Core Rope Memory.
Yes, that's right laminated cores of Molybdenum Permalloy for the Apollo core rope memory to make switching fast, low current and low eddy current loss. You will see that the method of using the cores in my build is different from the Apollo AGC but my video is to get you started thinking about how this type of memory works. More to come soon I hope and thanks for your interest and enthusiasm. Curious Marc is in a league of his own!
Just updated my comment below on the cores - hope I have made a better job of explaining it!
If you wanted to do it according to the core rope memory idea, three ferrite cores would be enough
Beniamin. could you elaborate a bit further - do you mean using binary coded decimal - that would need four cores. What am I missing here. Do let me know.
I guess i misunderstood the purpose of the cores. I don't see how they're used as memory. I guess i keep comparing them to magnetic memory cores
Good question. So the secondary (wire coming out of the coil) produces a pulse. Now the memory bit is really the primary wires. One single wire can be laced through each coil or not, meaning that one wire can when switched on (addressed) can control the output of seven cores (in my case) resulting in a 7 bit binary output. Every single wire going into the cores can be laced to give an output of any 7 bit binary number. So the 'memory' is not in the core itself but in how the wires are laced. Address those input wires in an order that gives the output of 7 bit words that you want and you have a running programme. Remember there is another type of core memory where the actual 1 or 0 is not 'wire through core/wire not through core) but instead a core magnetised in one of two directions - that is more like the Apollo one. Hope that makes sense! Thanks for asking, Anthony.
@@AnthonyFrancisJones that's what it seemed to me. At first it seemed the cores could just be bypassed in that setup with the wires going straight to the segments
Yes, that's it - you need the cores to reduce the signal to 7bits from what could be literally 100s of wires (individual lines of code) threaded through them. You can see why it is such a clever method!
In simple terms, the magnetic cores are being used as couplings between each address wire and each data bit. If the address wire goes through the middle of the core, then the core's data wire will output a 1 when the address wire is active. If the address wire bypasses that core, then the core's data wire will output a 0 even when the address wire is active. The data isn't stored in the cores, but in whether each address line goes through or bypasses each core.
Perfect - thanks for this. It does take a while to get your head around the fact that the memory is not really stored in a physical or silicon location but is just in the path of a wire! The whole plan of this video was to try and show this in a working - and I hope not to complex - model.
💖💖💖💖
I've been in electronics for 45 years, and I don't see how the transistors are ever going to be switched on since the emitter is connected to the positive supply, and the base drive voltage can never exceed that supply the way the schematic is drawn. Something's not right somewhere.
Agreed! I wired it up and got a transistor the wrong way around and it worked!! I was mystified too by this. I have been over and over it and everything is as per the schematic (unless these transistors are not following the agreed layout of pins for this type of base). I do remember that I did work it out at one stage and it was partly due to the fact that they are switching a rather odd shaped high frequency signal so are behaving in a rather odd way. Not an explanation I know but it works and I was equally mystified!
@@AnthonyFrancisJones Maybe those are PNP transistors?
Definitely BFY51 very old though. It was interesting that some of the newer ones did not work so something odd is going on here as you rightly point out!
Have a look at this great comment from Mick Healey - he has nailed it! "If the circuit is actually built as shown on the diagram at 18:56, then the transistor is working in 'reverse active mode'. In this mode, the transistor will operate, but (conventional) current flows in the opposite direction (emitter to collector) rather than collector to emitter as would be the case in active mode. A transistor will work in this mode, and even have some gain, but the gain is low. To enter reverse active mode, the emitter voltage must be greater than the base voltage which in turn must be greater than the collector voltage. That exact scenario is happening here. The base resistor ensures that emitter is higher than the base. Since the collector is connected to the 56kHz oscillator, then the base will nearly always be higher than the collector."
Could you use a large tv speaker magnet ring for a core memory?
Good question - whilst (as with many things) not totally impossible with the right electronics in this case the answer is basically no. The rings I use have to be capable of changing their magnetism as current passes through the wires. A TV speaker magnet is made out of material that is permanently magnetic so it would not respond to the changing magnetic field due to the wire. You need a material that is not normally magnetised. Interesting other types of core memory (I will do a video at some stage) do use cores that hold their magnetism but are designed so it is easy to erase and reverse. That is not the case with a TV magnet. Hope that helps and thanks for watching.
@@AnthonyFrancisJones So you can't demagnetize the ferrite magnet by heating or using a tool to demagnetize it?
Theoretically but remember that this type of material retains its magnetism once re-magnetised so is not what is needed. What we call 'soft iron' is the material that is needed that loses its magnetism easily and allows for fast changes of magnetic fields (with AC currents). So the TV magnet is not the right type of material! Hope that helps.
@@AnthonyFrancisJones That almost sounds like a ring of rare earth paramagnetic praseodymium.
Ferrite is a pretty common material and had all sorts of uses such as the core for old radio antennas too!
YOU HAVE TEN ADRESS LINES 9-0 LETS SAY YOU WANT TO SELECT ADDRESS LINE 7 WHICH FOR SAKE OF ARGUMENT IS NUMBER THREE ON THE DISPLAY, EACH BFY51 GROUNDS THE PRIMARY CORE WIRE, WHEN THE BASE IS POSITIVE SO FOR TO DISPLAY DIGIT NUMBER 3 YOU NEED TO SWITCH ON 5 SEGMENTS A,B,C,D,AND G, SO A,B.C,D, AND G ARE THE ONLY COIL CORES THAT YOU HARDWIRE THE SINGLE WIRE THROUGH ,WHEN THE 017 GETS TO COUNT 7 THE TRANSISTOR 7 SWITCHES ON GROUNDING THE PRIMARY WIRE AND THROUGH ELECTRO MAGNETIC INDUCTION CAUSES A VOLTAGE IN THE RED SECONDRY COIL LIGHTING UP THE SEGMENT AND ONLY THE COIL THAT HAS THE WIRE THATS HARDWIRED THROUGH THE MIDDLE OF THE COIL WILL PICK UP THE PRIMARYS MAGNETIC FIELD LIGHTING UP ONLY THAT SEGMENT,LETS SAY YOU HAVE ONLY ONE ADDRESS WIRE LINE, AND WITHOUT POWER AND BY LOOKING AT THE COIL CENTERS, IF YOU KNOW WHICH COIL IS FOR WHICH SEGMENT THEN YOU CAN WORK OUT WHAT DIGIT WILL BE DISPLAYED WITHOUT BEING SWITCHED ON SO ANY 1 OR 0 CAN BE RETREIVED WITH COMPLETE ISOLATION AND HIGHLY RELIABLE,
Thanks Patrick. Not easy to explain in words but you have got it! Thanks for this.
AGC core rope used Permalloy rings not ferrite.
Thanks Airton. Yes, this is a bit of a 'loose' representation but I hope it gets some started on how cores can be used. I plan to do a video of a core with read/write/inhibit/sense lines but not got onto it yet! Thanks for watching.
Permalloy is a trademark of a ferrite core. Nickelodeon visa vi puppet show. 🍌
@@co59720
Permalloy is not ferrite. There are various alloys but the composition is usually around 50 to 80% nickel with the balance iron. "Molybdenum permalloy" contains around 2% moly and 80% nickel. These alloys can be rolled into metallic sheet. For cores the sheet is thin and slit into "tape" that is wound to make cores. They can also be made as powders and mixed with binders that create a distributed air gap and formed into shapes, typically toroids. Moly permalloy powder cores are sometimes used for inductors in switchmode power supplies. MPP has excellent properties but has become very expensive.
@@co59720close, but it's spelled vis-à-vis.
Iron is iron, no matter how you slice it.
Literally Ropey...
Hope you feel it is a bit better than a ropey attempt at explaining this!
would be easier to show a diagram
Curtis, have a look at the video from 17:22 and I hope the circuit diagram helps - I realise it is hand-drawn!
10:33 Robots in disguise!!
Indeed!
@@AnthonyFrancisJones I'm going to attempt to gather these components and give this a try. Lots of tools i don't have but i believe i can get some local help on that.
Great - start with one core and see if you can get it working then scale up! Let us know how you get on!
Will do, i have a lot of planning ahead of me. I've been fascinated with this type of memory do to the Apollo missions and watching Marc's video's. Weirdly enough it all makes sense to me and i feel like i can accomplish this. Hopefully i can do this in a relatively timely manner. 😆
Great. I hope you are clear that this is, in many ways very different from the AGC memory and even some versions of core rope memory where the core stored the bit not the wiring like I have done but it is a start and gets people thinking about how these things work. What Curious Marc and his team do is outstanding! Good luck with your project!
Sorry, but this has nothing to do with a memory. As I understand it, this is just a cumbersome method of controlling a 7-segment display by inducing a voltage in the coils to make the LEDs light up. Here, no ferrite core is specifically magnetized and this state is then read out again.
Yes, you are right! But the 7 segment display is a distraction... The each wire is interleaved through the cores and each core represents a bit of data depending on the position of the wire (through or outside the core). So for example with 8 cores a 1 byte word can be 'stored' with a single wire. The more wires that are added the more bytes can be stored. This was used a lot for memory in telephone autodialers and the like. It of course is not to be confused with magnetic core memory (I have done videos on that as well) where the stored information was not in the wiring but in the core itself. Hope that helps to explain it all a bit better. Thanks for watching and taking the time to comment.
(I wrote this as well thinking that the above answer had been lost!) Interesting. Let me explain. So the 7 segment display is not important at all to the working of this device so don't get too distracted by it. It is just an output device to show the stored data. I use 7 cores but if you imagine me using 8 then each wire threaded through the cores would be able to store a byte of data. Each bit is stored by whether the wire treads inside or outside of the core. As you add more wires you are able to store more 8bit words. So the wires' positions store the data, not the cores! This was used a lot in old programable phone autodialers. It is not to be confused with magnetic core memory (I have done videos on this too - see above in the vides's description) where the data is stored in the core not the treaded wire position. Hope that explains it better and sorry if I did not make it that clear. Thanks for watching and taking the time to comment.
where's my crumpets!
@17:36 ... you show your DC clock pulse circuit ... that is NOT AC it is just a fast on and off of a DC signal ... making it a PULSED DC SIGNAL ... no where in your circuit does that voltage drop below 0v ... no where .... meaning there is NO AC CURRENT IN THE PULSE CIRCUIT .... it's all DC
You make a good point. I have not discussed the direction of current, that is true. I suppose it is best described as having a rapidly varying value of DC voltage value (interesting how we say 'Direct Current Voltage' when we refer to batteries!!). I would be interested to ask what you would call an AC signal with a DC bias - that may well not have a value that is considered to be zero at anytime but would still be considered to be AC as the current is rapidly oscillating, not necessarily through zero. Perhaps I was a bit 'loose' with my language here. Thanks for the correction and for watching. I will have a look at your channel when I get a moment.
@@AnthonyFrancisJones alternating current ... the dc voltage is OMNIdirectional ... you arent flipping the poles you arent even going below 0 ... youre 555 pulse timer is just a switch being flipped very fast so the power is there or not or on or off ... there is no cycle between zero max and zero and negative max ... ac is sinusoidal dc is is there or not or square wave for a basic idea visual
Yes, get your point about the value not passing the zero point. The waveform to the transistors was far from a square wave though which made me describe it as being more AC as it were (I think I showed some of the CRO images but used a DC blocking capacitor to show the AC component - a CRO thus makes it appear as if it is passing the zero point! AC of course does not need to be sinusoidal. I guess in the end referring to AC or DC is a bit misleading as really much of this circuit is using changing potentials/voltages rather than currents (of any magnitude) but of course the cores do need some current to function. The flux in the cores is changing rapidly (inductance etc. stopped the signal from being square in nature) so I guess I used the term AC for that reason but I totally get your point that the current direction does not change, just its magnitude. I hope regardless that there was something of interest in the video for you. If you get one to work do please let me know. It was a fun build!
@@AnthonyFrancisJones that curve is the induced capacitance of the bread boards you are using ... it is slowing the rise time of your pulse ... and will cause some false triggers if it gets too slow ... instead of using the 555 ... a quart crustal would be better then you actually get an ac voltage change ... as the crystal oscillates (about 36khz) ... which is why they are used in clocks ... easy x60 division
Thanks, I had not thought of the capacitance effects. Interesting! I guess it all just came about organically. I tend to build a bit at a time (with what I have lying around) and just get bits to work and never really plan it out in advance! I should do another build sometime and give the quartz oscillator some thought. If people are interested I do need to show other forms of core rope memory particularly one that works more like the AGC one.
Nope your constant use of SO is too annoying
Thanks for pointing that out. I don't read from a script but just narrate as I go along and when I am thinking hard I think I may repeat myself in this manner. I will endeavour to improve my delivery in future videos!
@@AnthonyFrancisJones Its a superb topic and kudos for building it
Pleasure, but nothing unless you and others watch it so thanks for doing so and I am learning from comments too!
@@AnthonyFrancisJones I have a chunk of Russian Ferrite bead memory I dont think its fully working but the skill in making it basically by hand is astounding
Yes agreed! I have some too (but would refer to mine as core rather than bead but I see what you mean). Usually they were not very dense so you can easily see the cores with the naked eye. I always feel it is a shame that comments here do not allow picture uploads as it would help me add to comments and you could share a picture of the item as they are all really intricate and interesting. Imagine what they must have cost back in the day too!
This is not memory, they are just transformers
Yes, that's the clever thing! They are transformers of a kind with multiple and independent primary windings (single wire). If a wire passes through the core it registers as a 1 and if a wire bypasses it it registers a 0. In this way the memory is permanently wired into the system which makes it very rugged and fault tolerant. Hope that explains it a bit better and thanks for watching.
@AnthonyFrancisJones it is not memory it is just power transfer
I see what you are saying but in many ways don't consider power transfer as the currents and voltages are so small that it is not the power that counts here, just if a secondary in a core picks up a magnetic field from the primary (1 - wire through core) or senses no magnetic field (0 - wire outside of the core). I am sorry if I have failed to explain it well but it is an interesting idea and one which some types of early memory relied upon. It reminds me of programming old computers (you may well have done this) were dip switches were used as a form of memory to hold a group of bits at a fixed value though perhaps this is stretching the analogy too far!