Core Memory (magnetic memory)

Discussion in 'The Projects Forum' started by Art, Jun 17, 2015.

  1. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    Hi Guys :)
    I have been learning about core memory, and would like to try one of my own just for a fun project.

    I’ve learned about as much as I can watching Jeri Ellsworth’s video, and several pages out there
    explaining how it works, how to drive the matrix, and also interface to a micro controller:
    http://www.corememoryshield.com/report.html

    I still have a few questions...
    I understand the resistors must set current, and if they weren’t present, every row or column would be a short circuit.
    If I could not find the exact dimensions of ferrite rings used in the above micro controller project,
    is the resistor value the way to compensate to change current flow (so long as the transistors can drive the current)?

    In a noisy circuit, would it be a good idea to shield with copper? and it would matter the copper was not close to the beads?
    I only have in mind an 8x8 matrix to store 2x32 bit floats (a geo location).
    If I went all out with 64 beads I’ll want to mount it in something.
    Cheers, Art.


    This is a closeup of a quarter of a 1024 bit matrix (so only 256 bit shown here)
    [​IMG]
     
  2. DickCappels

    Moderator

    Aug 21, 2008
    2,647
    632
    Yes, good idea unless your driver has internal current limit.

    Get the smallest toroids you can get, made of square loop material. If you can find some old magnetic memory that you can cannibalize, problem solved.

    The tiny cores used as timing transformers in some Compact Florescent Lamps (CFLs) should work well.

    You can read about the characteristics in this Wikipedia article: https://en.wikipedia.org/wiki/Magnetic-core_memory

    If you are in a noisy environment it might be helpful to enclose the magnetic cores and the sense amplifiers in a shielded enclosure. Copper or copper-clad printed circuit board material would be wonderful. If fully enclosed the memory does not need to be in close proximity to the shield.

    If I were going to experiment with magnetic memory I think I would start with one bit and get that right before investing a lot of energy in building an array.
     
  3. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    Thanks for the reply :)
    Are you saying the small ring beads for RFI suppression will not be suitable ferrite material?
    I’m tempted to break a CFL to see.

    Also I wonder about their lifecycle in terms of number of read/write cycles.
    I know some have been running a long time.
     
  4. DickCappels

    Moderator

    Aug 21, 2008
    2,647
    632
    I am not saying that RFI beads aren't suitable because I don't know. Give the RFI beads a try -if they work, great! If not, you learned something.

    I don't think you can wear it out.

    Here is a note about opening compact florescent lamps to get to the parts inside:

    Some compact florescent lamps from Sylvania had failed in my home. After buying new Philips lamps to replace them, I ventured into the garage to take one of the Sylvania lamps apart. The first problem was getting to the electronics in the base of the lamps. In later correspondence, Wolfgang showed me that the base of the lamp can be pried open and the circuit board removed without having to break any glass. Be careful not to break the glass tubes in the lamp, as they contain mercury, which is toxic.
     
  5. takao21203

    Distinguished Member

    Apr 28, 2012
    3,577
    463
    rather get some real ferrites they arent expensive i think the EMI kinds might not work the way you want or may need to adjust the magnetization current/timing.

    6mm is about the limit to add a few windings without too much effort rather use larger cores.

    I sell 13.5mm for $2/12 pcs but can sell lower price for more of them, and also have smaller kinds, like 10mm.
    Youd have to try one, each is different even if they look same, their magnetization parameters can be different a hundred times and more.

    EMI ferrites are only specified in terms of resistance at some frequency.
     
  6. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    Ok thanks, it seems easiest to get a core memory from eBay,
    they don’t seem expensive until they are part of a stack.
     
  7. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    A bit of an update.. My 1mm ferrite arrived and I have woven an 8x8 matrix, but while waiting this is
    probably the first time I’ve written software prior to hardware. The memory should be read/written serially,
    as I plan to ultimately control it’s drive lines with 2x8 bit shift registers in series.
    They don’t even need to be latching because the bridge ICs have enable pins :D



    A trap for younger players like myself... Just because 3x0.25mm gauge wire fits through a 1mm core,
    does not mean it will be as easy at different angles!
    It has come to threading the sense wire one core at a time and have to pull through from the threading end.
    You can see in the magnified photo, the core beginning to strip enamel off the sense wire.
    This should not matter if one wire is partially stripped.
    This is also why the kink between the diagonal cores, but at least that should happen each time, and be uniform.

    Prior to sense loop:
    [​IMG]

    Magnified (sense loop threading) :
    [​IMG]
     
    absf likes this.
  8. DickCappels

    Moderator

    Aug 21, 2008
    2,647
    632
    I am impressed at your progress as well as the fact that you were actually to thread the wires through the cores.
     
  9. alfacliff

    Well-Known Member

    Dec 13, 2013
    2,449
    428
    the last core memory I worked on was the memory in a juke box. sorry my old memory has failed a bit and I dont remember the brand of juke box. the records were on a rack that moved back and forth, with a switch contact moving along. the switch discharged about 350 volts dc into the wire going through one core at a time, selected by the moving switch contact. there were two other wires through the cores, one to write again a 350 volt dc pulse, but of oposite polarity, and one read wire looped through all the cores. the thing used a small thyratron to detect the read pulses and stop the rack to play a record.
     
  10. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    Thanks for the replies :) This one is torture, but I think the next one would be almost trivial.
    I conceived some tricks that I couldn’t employ because it would make this one non-uniform.
    The most obvious is slightly thinner wire,
    but also, I would thread all Y lines diagonally prior to soldering, and wrap around all posts only once,
    using the wire ends as guy wires for the posts threaded through to the PCB underside.

    That sounds like a freaking heavy ferrite!?
     
  11. alfacliff

    Well-Known Member

    Dec 13, 2013
    2,449
    428
    the last core memory used at NCR while I was there was a mat memory, it didnt use cores, it used rods of ferite woven with the wires into a mat. worked the same as core memory, but was easier for a machine to weave the memory.
     
  12. RichardO

    Well-Known Member

    May 4, 2013
    1,228
    382
    I was told that near the end of the era when core memories were common, the X and Y wires could be threaded by machine but the sense wires still had to be done by hand.

    I was also told that the wires were threaded by pulling them though the cores using a stiff wire. To do this, the copper wire was welded to the end of "piano" wire. The welding sounds tricky to me.
     
  13. RichardO

    Well-Known Member

    May 4, 2013
    1,228
    382
    If you get good at this I can give you some cores. The catch is that the cores I have are smaller than the head of a straight pin. :eek:

    To help, I also have some #44 magnet wire although I don't know if 3 wires will go through one of these very small cores.

    Let me know by PM if you want me to send you a "kit" of cores and wire. I will even pay the postage. :)
     
    SgtWookie and atferrari like this.
  14. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    Hi Richard0, I might take you up on the wire if I have trouble, and thanks for the offer!

    I totally understand all of that, and already thought a sewing needle would be great,
    but impossible to feed through. A rigid wire could be threaded through an entire reasonable
    diagonal row at a time which I currently cannot do.
    I have a finer magnet wire which would flow through easier,
    but it’s not rigid enough to initially feed through a core now (on this board).

    It’s not out of the question to do another one depending how happy I am at the end.
    Also if I break a core now it’s game over for this PCB. I need all 64 bits.
    I plan to make it on 3 boards so that any can be discarded and started again.
     
  15. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    Hi again all :)
    Hopefully this sense wire doesn't cause any problem.
    Very difficult with the wire used.
    Thankfully everything thats us supposed to be conductive is,
    and no shorts despite losing some enamel from the sense wire.

    I have a few questions, but will leave it till I can properly express them.
    I spent 2 days in bed with bad back and then finished this, and hurt it again, and back in bed :(

    [​IMG]
     
  16. RichardO

    Well-Known Member

    May 4, 2013
    1,228
    382
    Nice picture. I think that my eyes would also hurt after wiring like this. :eek:
    By the way, what size are the cores and what gauge is the wire?

    Hope to hear from you soon when you get better.
     
  17. ian field

    Distinguished Member

    Oct 27, 2012
    4,413
    782
    If the TS can find any actual core memory - its probably worth preserving as a valuable collectors item.
     
    RichardO likes this.
  18. Art

    Thread Starter Distinguished Member

    Sep 10, 2007
    785
    61
    @ian field
    As a vintage collector I can tell you that argument will never be resolved.
    Should you smash an old vintage radio up to use the timber for a mock antique desk?
    What about adding hardware for an mp3 player input, and what about when AM MW stations are no longer existent.
    What if you have two identical models, but can only have one original in good condition using parts from both.
    Even if you appreciate vintage stuff, everyone’s answers will be different.

    There are plenty of panes (even damaged ones), separated from stacks. In this case I’d consider them damaged.
    I got mine from eBay as cores (Bulgaria), but even so, I think damaging the memory to recreate memory
    is one of the better justifications ;)

    @Richard0
    These are 1mm diameter cores which means an inner hole smaller than that. X & Y drive lines are 0.25mm,
    and the sense wire is slightly finer gauge than that. If anyone reads this in future, DON'T DO THAT! go slightly finer.

    Hi Everyone :)
    I failed to mention this project is Hardcore! and if you disagree, you are incorrect :D

    I have etched the creases and rectangular cutout for a brass cage for the memory module.
    the blue stuff is just some left over resist, and this whole view is the inside of the cage:

    [​IMG]

    Due to a limitation with mounting, I plan to make a vertical strip on 4 PCBs,
    the middle part being a stack of two PCBs, where the sense amp is also shielded,
    as well as the two wires leading from the sense wire in the core memory,
    past the bridge chips, to inside the sense amplifier cage.

    [​IMG]

    I have all parts, or at least have ordered them, to duplicate the circuitry on Wayne’s Page:
    https://sites.google.com/site/wayneholder/one-bit-ferrite-core-memory
    It’s only one bit, but he has used two drive line axis, and electronically controlled it so it’s expandable.
    I plan to duplicate it exactly, but controlled serially with shift registers.
    Then I do have another idea for a change of my own once it works!

    Some questions I mentioned before, but could not sit at a PC
    comfortably enough to coherently express them :O ...

    Referring to the hand wound toroid transformer, 10 turns primary, 20 turns centre tapped secondary,
    so a pulse of either polarity can be detected by the sense amp. Schema is near the end of Wayne’s page.
    Since my sense amp is at the other end of a long assembly, would it make any difference put the toroid
    either end of the assembly?
    i.e. close to the core memory with 3 leads to the sense amp, or mounted close to the sense amp with two
    leads from the core memory to the sense amp.
    If closer to the memory, would it make sense to put more turns in the secondary to increase the voltage of the pulse?

    Here is a re-hosted image of the hand would toroid, and you can see the single core bead (same as mine) to the right:
    [​IMG]

    Secondly, I cannot find the same part Wayne used for the toroid core. Is either of these suitable?
    [​IMG]
    One is L8 ferrite material, roughly (or even maybe precisely) twice all of the dimensions of the ferrite Wayne used,
    and is supposed to be for wide range transformers in RF, etc.

    Any advice here would be appreciated... or if you see anything wrong with my plans so far.
    Cheers, Art.
     
  19. ian field

    Distinguished Member

    Oct 27, 2012
    4,413
    782
    Years ago I recovered a couple of panes from a scrap computer, I hid them behind the setee hoping they might be worth something someday as a collectible relic - ashamed to say, I'm pretty sure they got thrown out at some point.

    The core driver boards (or what's left of them) might still be in one of the tea chests at the back of the garage. One of the boards had neat rows of the old style 2N2907 in TO18 cans - they all got harvested and used in numerous projects. There were loads of TO5 transistors with unidentifiable part numbers, they're still on the board wherever it is.
     
  20. RichardO

    Well-Known Member

    May 4, 2013
    1,228
    382
    Wow! Your cores are smaller than I expected. I need to measure the cores that I have. I doubt the are much smaller than yours...
    So, the wire you used is 30 gauge? The 44 gauge wire I have would definitely be a lot better.


    Hardcore? Of course. All ferrite cores are hard and brittle. ;)
     
Loading...