Magnetic Core Memory Decoded Part1
HTML-код
- Опубликовано: 28 авг 2024
- This is the first video in a series in which I will be describing the way magnetic core memory works. This series is intended to promote my new book which is entitled 'Magnetic Core memory Decoded'.
I picked up your book. I'm using it as reference to start building a memory controller.. for a vintage set..
It has gone to print so should be available in a few weeks. There is no way of knowing how long it will take different outlets to start listing it as I have no control over that although I will be selling them directly and will post a release video once it has been released.
One question: some core memory have a sense and an inibit wires. Like you say in the video, they could be combined. However, I was wondering why some core memory went to the trouble of having both.
The electronics required to drive large memory arrays becomes increasingly complicated due to the inductance and capacitance and so the supply voltages tended to increase to allow the circuits to work. These days it would be simple to implement circuits using modern devices but back when these systems were developed it was much more difficult. I go into quite some detail in the book on this subject. The problem using a combined sense/inhibit wire is how to arrange to detect very small signals in one direction while allowing large drive currents in the opposite direction. This makes the design of the electronics much more difficult but may be worth the effort in not needing to thread the extra wire through thousands of tiny cores. Early memories used fairly large cores (relatively speaking) so this was not too difficult but as the cores got smaller it became harder to get 4 wires through so most designers moved to the 3 wire option.
Hi Jerry, very excited about this series and definitely looking forward to picking up the book as soon as it releases!
Maybe I missed it (or it'll be covered in a future video), but what is fundamentally different about the larger cores that doesn't allow them to be used? Do they react on the curve tracer more similarly to the inductor you showed?
I will post an announcement once the book is available. It is not the core size that makes the difference but the materials used to construct it. The book covers this but in brief the memory core ferrite material has different flux characteristics than the types use in filters, transformers etc which do not exhibit the magnetisation switch but just run into saturation. I will be demonstrating this in an upcoming video. Larger square loop ferrites could be used but I will explain why this is not a good idea in the videos.
good question! I was wondering the same thing. 😀
I imagine to assemble these, one had to have a steady hand, and a very good microscope. seeing how tiny the Ferrite cores are. this was a truly amazing technology for the time.
A very interesting topic. This kind of memory, as you said, are very reliable. I recently saw a video where they were recovering ram content after more than 50 year of inactivity (and no battery 😬😬).
Yes it is very reliable compared to modern memory. Rope memory is even more reliable. Almost every vintage computer I repair which is fitted with DRAM devices requires several replacement chips but I have only ever had to repair two core memory mats and that was due to physical damage.
Hi, Jerry. Your book just arrived, today! I do intend on getting out a fresh logbook and starting experimentation with a goal to complete a small project in the future. The first and foremost questions in my mind right now are: (1) I think you planning to provide a kit? But I believe I've not found a site, today, that has availability. That may be my failing. But I'd appreciate a suggestion from you, if possible. And (2) how might I track down a good supplier for square loop cores (of small, but varying sizes?)
If you contact me through my web site I can supply the bare boards and cores.
@@JerryWalker001 We communicated recently on this topic. But I haven't received a response to my second letter. So I am re-sending it. Just FYI.
Hello Jerry, always nice to see your step by step explanation.
When will your book be published and where can we buy it? In a previous video I thought you were speaking about a new PCB, is it your intention to combine your theory with a project we can build ourselves?
The book should be available in approx 5-6 weeks. It will be in all the big book stores plus Amazon etc (same as my previous book). I will also be selling them directly on my own website. I have already designed the board and tested it (more on that in the videos) and will be making the boards available as a set along with cores. The board is the one designed in the book. The idea is to make it possible for anyone interested to build one.
@@JerryWalker001 Jerry, when do you expect your book to be available? It is (today) 23 April 2021 and I can't find the book on your site or on Amazon. I've got a small tin of ~10K cores out in my workshop and would like to try building a small core stack! ;)
Hey, so, what would happen if you had an analog signal, like a sound, going through a core memory bead like that? Assuming it's at a current high enough to trigger the non linearity. Would it be just that, a stateless nonlinearity? Or would it be stateful somehow?
Can't wait to buy the book and the kit! Always been fascinated by core memory. Never quite understood how the address decoding didn't end up being humongous. There must be some trick since using a naive approach (74154) for larger arrays would require lots of circuitry. Could you explain how they did the decoding for something like a 64k word core memory?
I will explain this in a future video but in short there were different approaches. First the number of cores is simply increased. So for example a 64 x 64 core mat gives 4096 words (one mat per word bit) or 128 x 128 gives 16k. These were then paged in a very similar way to the way SD cards page the memory space. The memory could then be multiplexed to combine each block of memory. The most common approach was to simply add additional memory assemblies so an 8k memory would be two 4k blocks and each was selected by address range. Magnetic core memory always used a lot of circuitry but as it was just repeated then it was relatively easy to manufacture. The real limitation in manufacture and hence cost was the difficulty in automating assembly of the core arrays.
@@JerryWalker001 Thanks for responding. Even a 64 x 64 array would require a 2:4 and four 4:16 for each set of the 64 lines if done naively. I know there must be a clever trick that gets from 16 address lines to 1 of 65535 different memory elements without having to resort to 65535 16 input AND/NAND gates plus all the inverters. Otherwise more of the chip space would be dedicated to the decoding than to the memory elements.
@@drewtech2 As I said I will explain it in more detail in a video as it is difficult to describe in brief. Remember that it uses current coincidence to flip the cores and all X and Y wires pass through all mats and you do not use separate drivers for each mat. This means that in a 64 x 64 array you can drive the X wires in groups (more than one at a time) . I am not sure where you get the 65535 from as 64 x 64 array would only need 128 drivers even if you had a separate driver for each wire. For a 16 bit address you would split the address into 2 x 8 bit values or split off the lower 12 bits (4k) and use the upper 4 bits to select one of 16 blocks of memory. This is the same method that many much newer devices use to bank memory. For example PIC's use a very similar system to address large blocks of memory. It is also relatively easy to reduce the wire count by interlacing them.
+
Have you come across the book _Square-Loop Ferrite Circuitry, Storage and Logic Techniques, Prentice-Hall Inc., 1962_? It is out of print but can be found through some university Inter-Library Loan programs. It's not real heavy in the maths, but goes into good detail on the technology, and I would imagine a good textbook to cite in your references.
Ooooooo always wanted to do that
This seems to be very similar to how ssd’s work.
Will you do magnetic logic?
I am hoping to cover basic core memory (RAM), Rope memory (ROM) and magnetic logic.It depends on the level of interest. Magnetic logic is very interesting but I figured if I covered standard magnetic memory first then it would be much easier to describe magnetic logic.
I think magnetic logic would be a video series on it's own. I know there were some old computers that used some sort of weird transformer as logic gates.
@@Bata.andrei Yes I totally agree and this is in fact the subject of my next book. I will be building a Rope memory system once the current memory project is complete and then a magnetic logic system.
Would you consider talking about memristors both as digital logic and in analog computing? I mean the kind you can make using point contact junctions with copper sulphates.
In at 3 minutes and he hasn't really said anything.
A comment that has no purpose. I am not forcing you to watch a video so if you are not interested then just don't watch. Some people may find it interesting.
But if you actually watch the full video, you'll learn how they work. It's funny... a lot of videos/speeches/books make you wait for the entire thing before you can understand the entire thing. Funny... things work out that way.
Have attention spans really become so low?
If he hasn't said anything, then be our guest and make a video explaining it better.
@@JerryWalker001 i went to his channel... A single video, 22 seconds. Lol.