Loading ...
Sorry, an error occurred while loading the content.

61264Re: [beam] Memristor

Expand Messages
  • Richard Piotter
    Jan 12, 2014
    • 0 Attachment
      I hope to have an update tomorrow. LOTS of progress! I built my first prototype boards for the new design. It's not laid out as NICELY as the DIP style PC boards I proposed will be, but it's more than sufficient for testing.

      http://richfiles.solarbotics.net/robots/AluminumBalls.jpg
      http://richfiles.solarbotics.net/robots/MemristorAluminumBalls.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_1Parts.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_2SilverPaint.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_3Balls.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_4Fit.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_5Continuity.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_6Paint.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_7Sulfur.jpg

      I decided to mask the copper traces with white enamel paint, so I only get sulfide at the actual pads. I had a lot of trouble with stray sulfide on copper where it wasn't needed on my last experimental board. Also tinned the connector pads before hand too... again, the sulfide made the old planar memristor board really stubborn to solder. Even though this is only a prototype, with a crude, hand drawn layout, I find I'm rather impressed by it so far. I had pre-assembled it to test for pad alignment between the two boards, and it looked nice. Granted, the sulfide layer is reacting as we speak, so I won't know exactly when I can assemble it, but I'll be able to assemble it in a matter of minutes once I clear the sulfur off the sulfide pads! It's literally aligning the two boards, and just snug tightening a pair of screws. We are talking about a LIGHT hand tightening, using a thread lock compound, such as Loctite 242. I could not even get my torque driver to go low enough for the task! I suspect it'll be just enough for friction to hold everything in place, and nothing more. I'll have to play with that value.

      I made one change from the concept that, quite frankly, I kinda like. I used the circuit pen that I no longer need for the planar memristors (seeing as how they don't work, since paint soaks through the sulfide), and instead put a dab on each of the top copper pads, and then stuck an aluminum ball bearing into the paint dab. I essentially welded the aluminum balls to the upper copper pads with the silver based paint. They have under 2 ohms resistance from top connector to bottom connector, through the aluminum ball between them, on a meter that routinely has 1.6 ohms in the leads! What's more, is the silver coats half the ball and the copper pad, this protecting it from reacting with any stray sulfur particles that might remain, or from oxidization.

      What the circuit pen step ultimately does for me, is it secures the aluminum balls without cutting a complex sheet to sandwich between. Instead of the foam layer, I plan to seal the unit up with a bead of RTV silicone sealant around the edge. I'll leave a pair of small openings unsealed and allow it to cure under a fan directed at it, to make sure any outgassing from the curing process doesn't linger inside. I honestly have no idea if any of the vapors released by the sealant could affect the sulfide layer, so I'd like to minimize them. I WILL have a final seal to close up the opening. If I paint both boards, I'd also seal up any vias (in the final version), making the module environment proof, without a difficult to make sheet with over two dozen tiny holes, or dealing with laser cutting. It DOES require a +$20 circuit pen, but that ain't stopping anyone from trying the holy sheet method either.

      I did pre fit everything together while the circuit paint was still tacky, as I wanted all the aluminum balls to press tight against the copper pads while the silver paint dried the balls onto the pads. That would ensure that all balls were level too, as it's possible a ball might have not bottomed out in the paint dab, against the copper pad beneath.

      All I have to do is remove the sulfur, and check the pads. If they are good, I can go ahead and screw the two boards together, and I'm done! I'll seal it at some point, but like I said, for now I'm all about testing.

      Well... That's the prototype, and I'll know soon if it works. I still have to wire up the simple circuit to turn my scope into a curve tracer. I already added a LOT of equipment to my workbench so I now have a proper function generator to feed a signal into the curve tracer. In fact, I now have 3 function generators (2 analog, one digital), 5 Bench supplies (2 dual tracing, 3 precision), and three digital counter/frequency meters. Also a DMM... Cause my old, ancient, HP DMM died on me...

      :( RIP... My HP...

      Oh yeah! I also confirmed you can fit a ring of 8 copper pads around a 2-56 screw and pem nut and still fit it between the two rows of a 40-pin DIP package. In other words, the DIP layout is a DEFINITE go! Even with hand drawn traces and a single sided PCB design, I was STILL able to fit a ring of 8 pads in the required space (just barely). With professionally made boards, and tiny traces, I can shrink the ring further, making an even more comfortable fit. I did make a realization however... As COOL as the 40-pin DIP package is, just for the fact that it mirrors the appearance of an actual chip very closely, I think I can create a version that uses a right angle header on the edge of the board, and would plug into a female header mounted to the board or a ribbon cable, such as an old PATA cable. It would technically take up less board space on whatever it plugged into... It would just be a LITTLE less cool than making it look like a chip! LOL

      One thing I can do with the non chip variant is simply make the board bigger... Even MUCH bigger. Going with a header and expanding the board would let me try to build a crossbar latch. If I understand the nature of a crossbar latch, it's a grid of memristors, that can be tied to a grid of rows and columns. If I'm not mistaken, I think each memristor can be tied in or left out, much like a ROM, only you get a memristor at each link vs a diode. You can do logic with a cross bar latch. A 20 pin (2x10) header could access 100 memristors, a 32 pin (2x16) header could access 256, and a 40 pin (2x20) header could access 400 memristors. even a small one might be nice, and I might make it longer in one dimension than another. if the boards are manufactured, i could add SMD resistor pads to the top, with small trace bridges between pads that can be cut with an x-acto, or repaired with a small shorting resistor or a carefully placed solder bridge. I will worry about that once I have this prototype tested.

      Wish me luck! Luck gets me 16 memristors, and an under 1.5 inch board!

      Richard Piotter
      richfiles


      Begin forwarded message:
      > From: Richard Piotter <richfiles1@...>
      > Subject: Re: [beam] Memristor
      > Date: January 9, 2014 5:12:29 PM CST
      > To: beam@yahoogroups.com
      > Reply-To: beam@yahoogroups.com
      >
      > Oh wow, the aluminum balls are so tiny!
      > You don't really realize just HOW small they are till you see them!
      > It also makes you think... Dang... I paid $11.20 for just this! XD
      >
      > The parts arrived today! Gonna do a quick prototype board this weekend and see how the hexagonal/octagonal ball pattern situated around a screw works, and if it functions at all as a memristor! Can't wait! I'll also spend some time watching video tutorials on how to use Eaglesoft CAD, so I can get the PC boards made.
      >
      > I'm really hopeful, and kinda excited!
      >
      > Richard Piotter
      >
      > Begin forwarded message:
      > > From: Martin McKee <martinjaymckee@...>
      > > Subject: Re: [beam] Memristor
      > > Date: January 7, 2014 6:27:17 PM CST
      > > To: beam@yahoogroups.com
      > > Reply-To: beam@yahoogroups.com
      > >
      > > Yes, that makes it quite clear, and it sounds like a nice robust system. Hadn't thought about soldering the connector to the upper board and mounting the stack that way. It's a very interesting project -- makes me want to look at circuits to interface to the module. I've always been interested, especially, in the AI portions of robotics and the possibilities that memristors represent are attractive.
      > >
      > > Ah dear, where has my time gone?
      > >
      > > Martin Jay McKee
      > >
      > >
      > > On Tue, Jan 7, 2014 at 1:27 PM, Richard Piotter <richfiles1@...> wrote:
      > > The upper connection will mimic the bench made memristors that have been seen so far... in that it is purely a mechanical connection. It will be held together by compression between the upper and lower PC boards, as applied by the torqued screws. Lower board gets a pem nut soldered or pressed into the bottom, while the top gets a screw going through to the lower board.
      > >
      > > Between the two boards will be a sheet of compressible material, such as that type of foam packing sheet you often see used for protecting new electronics from scratches, sandwiched between the two PC board layers. The foam will have holes for both the screws and the memristors. I will also be likely to incorporate an alignment key, if it proves necessary. Basically two headers on opposite corners of the board electrically attached to nothing, meant to simply pierce the foam and hold it stable. Materials and assembly are as follows:
      > >
      > > BOM:
      > >
      > > 1x lower PC Board
      > > 1x Upper PC Board
      > > 4x 2-56 Screw (approximately 3/16 to 1/4 inch length for the balls I'll be using)
      > > 4x 2-56 pem nut (solderable preferred)
      > > 4x 20 pin single inline header (male) (or 2x 20-pin, or 8x 5 pin, etc.)
      > > **PC board side of header leads should be long enough to pass through both PC boards.
      > > 20x aluminum BBs (I am getting mine from McMaster Carr)
      > > 1x cut foam sheet
      > > 2x (optional) single header pin for use as foam sheet alignment key
      > > Solder
      > > (Optional) Sealant or conformal coat
      > >
      > > Assembly:
      > >
      > > 1: Press or solder pen nuts to lower board.
      > > 2: Solder header to lower (sulfide) PC board. Solder key pins if present.
      > > 3: (Optional) Seal the vias and solder joints for the headers, and mask the pem nut holes.
      > > 4: Generate copper sulfide by applying sulfur to the lower pads and allowing it to react
      > > 5: Removing the sulfur
      > > 6: Inspect the copper sulfide pads, and redo step 4 if need be. Otherwise...
      > > 7: Make or have foam sheets made that have a hole pattern matching pads, screws, and keyways.
      > > *I'm going to ask a guy with a laser cutter to do my foam, since I plan to build small
      > > 8: Place foam sheet on first PC board, aligning any key holes to key pins (if present).
      > > 9: Place an aluminum ball bearing inside each hole in the sheet that corresponds to a pad.
      > > 10: Line the upper board with the lower board and start the screws in.
      > > 11: Using thread lock compound, such as Loctite, Torque the screws (spec TBD).
      > > 12: Insert header for upper board through lower board and solder to upper board.
      > > 13: Test memristors electrically.
      > > **If any fail, you may attempt to go back and repair, or mark it's pins as defective and amply not use it. There are 20 memristors on the PC board, I have no idea what the consistency from part to part will be, nor do I know what kind of yields this will result in.
      > > 14: (Recommended) If device is satisfactory, seal the screws so they can not be easily turned.
      > > 15: (Optional) You may tack the corner solder joints of the lower board header to the upper board holes (there will be no electrical connection, merely a mechanical connection). Alignment key headers could theoretically also be tacked to the upper board for additional strength.
      > > 16: (Optional) Seal the sides of the device to fully encase the internal components.
      > > ** Any outgassing by the material of the sealant must be confirmed to NOT react with the copper sulfide, or either the copper or aluminum metals.
      > >
      > > Step 17: Use it!
      > >
      > > The entire device is meant to be mechanically secured. Torquing the screw down at the center of each ring of memristors is intended to create a compression fit between the copper pad above, the aluminum ball, and the copper sulfide pad below. I just ordered some $32 in balls, screws, and nuts from McMaster Carr. I could have got cheaper hardware, but I am dead set on solderable pem nuts, and with minimums, I'll have parts to build five right off the bat I went with the 1/32 (0.77mm) aluminum balls. I just really like the compactness they offer, and if I get issues with them, I can always just reorder the 5/64 (1.98mm) balls and try them.
      > >
      > > The biggest issue I can foresee, if these work, is that repeated insertion or removal from a socket could be detrimental to the internal structure if twisting is involved. I would personally solder it to a PC board, or socket it once, if it's tested good, and then do modifications on what it is attached to. mechanical stresses could alter the state of the memristor, considering the non volatile state of it's resistance. Under normal motion and operation, it should remain intact and secure. If I were to make it bigger, I could have more screws. That alone would make it more rigid, but I wanted to maintain the industry standard footprint. I may just try to get thicker PC board material for the lower board, to simply give it a firm foundation, and let the upper board be normal thickness.
      > >
      > > I'm going to do a PC board layout and the PC boards would then be an item that could simply be ordered from a board manufacturer from the files. Even just having the full solder mask to protect the traces from any lingering sulfur alone is worth it to buy the boards professionally manufactured. It also ensures pad alignment.
      > >
      > > And while I'm on professional PC board manufacturers... I love how this one board house I looked at puts "Live" salesperson in quotes, just like that... What? Do they have a few undead sales personel on staff?! LOL XD
      > >
      > > I want to ask my boss at one of my jobs what material we use for a gasket in one of our products. If i ca get it in sheets, and then get some cutting dies for the holes, I could make some PRIMO pieces for the middle layer! I couldn't afford to have it tooled up though, not without something like a kickstarter. I'd have no chance. For the time being the middle layer of material will have to be hand cut or laser cut.
      > >
      > > I hope that explains things clearly enough to understand.
      > >
      > > And in reply to Connor, I'd say that assembly is fairly straight forward for now. The boards will be ordered from a professional board manufacturer. Cutting the foam/rubber/silicone gasket that sits between the boards will be the most complex task, and if I can get a bunch laser cut, then the hardest task becomes soldering the 40 pin headers to a pair of PC boards, and torquing 4 screws to the same torque. Not terribly bad by any means. Nothing a little time with a soldering iron and a screwdriver won't assemble. I do suspect torque pressure to be light. Too much torque and you risk damage to the copper sulfide pads, I think.
      > >
      > > Richard Piotter
      > > richfiles
      > >
      > >
      > >
      > > Begin forwarded message:
      > >
      > > > From: Martin McKee <martinjaymckee@...>
      > > > Subject: Re: [beam] Memristor
      > > > Date: January 7, 2014 12:05:20 PM CST
      > > > To: beam@yahoogroups.com
      > > > Reply-To: beam@yahoogroups.com
      > > >
      > > > The layout looks good to me. What are your plans for doing the upper connection without a rigid solder joint?
      > > >
      > > > Martin Jay McKee
      > > >
      > > >
      > > > On Mon, Jan 6, 2014 at 9:32 PM, Richard Piotter <richfiles1@...> wrote:
      > > > This is a quick layout of the new memristor concept. I am just learning this program (Eagle), so I have vias simply representing the memristor pad locations (and I for the life of me can't figure out how to change layers). Dimensions are not final, but the general layout should be clear. Four screws and nuts (likely pem nuts, preferably solderable pem nuts if I can find them)
      > > >
      > > > The hexagonal pattern provides both symmetry around the screw (for even torquing pressure between each memristor) and an even count of 20 memristors. The pinout will connect the memristors across each pair of pins across the width, so Pin 1 - Header 1 and Pin 1 - Header 2 is one memristor. Pin 20 - Header 1 and Pin 20 - Header 2 is the last memristor.
      > > >
      > > > So, what does everyone think of this layout?
      > > > (A reminder, the vias only represent the actual memristor ball locations between the boards... I still gotta learn the program to really use it! LOL)
      > > >
      > > >
      > > >
      > > > Richard Piotter
      > > > richfiles
      > > >
      > > > Begin forwarded message:
      > > >
      > > >> Made a little design progress. I have to leave soon, so I'll make this real quick...
      > > >>
      > > >> I think I have a viable PC board design that places 20 memristors in a module that would mimic a 40 pin DIP IC, both in size and pin configuration. It would fit PC boards, 40-pin sockets, and breadboards. I still need to do testing, and order parts, but it's looking good so far, as long as my actual experiments pass their tests.
      > > >>
      > > >> Wish me luck!
      > > >>
      > > >> Richard Piotter
      > > >> richfiles
      > > >>
      > > >> Begin forwarded message:
      > > >>> From: Richard Piotter <richfiles1@...>
      > > >>> Subject: Re: [beam] Memristor
      > > >>> Date: January 4, 2014 1:58:13 AM CST
      > > >>> To: beam@yahoogroups.com
      > > >>> Reply-To: beam@yahoogroups.com
      > > >>>
      > > >>> I found these two parts at McMaster Carr. $11.20 for 100 balls. I can go with 1/32 balls (0.79mm) or 5/64 balls (1.98mm). I think I will get the larger ones to start with, but I am seriously debating on the smaller ones. At less than half the diameter of the bigger one, there would be a significant reduction in size, potentially allowing for even greater density. The problem is that I don't know how the sulfide may affect things over time, or if it's wise for the pads to be so close.
      > > >>>
      > > >>> http://www.mcmaster.com/#34665k21/=q3lxcc 1/32" balls
      > > >>> http://www.mcmaster.com/#34665k25/=q3lxct 5/54" balls
      > > >>>
      > > >>> The other concern, is these are 2017 alloy, with a touch of copper. I don't know if that is good or bad for memristor applications, as no one to date has actually IDed the alloy of aluminum used in their experiments. I don't know if 1100 (over 99% aluminum) was used, or if an altogether different alloy was used.
      > > >>>
      > > >>> So, opinions... does the ultra tiny 1/32 balls make sense, or should I restrict myself to the larger 5/64 balls for experimentation, for now. As for electrical connection, I am still considering headers, though now i am kinda looking at 40 pin PATA-IDE style connectors, as the cables are pretty much available for free everywhere, thanks to SATA!
      > > >>>
      > > >>> I also considered a balancing network for each memristor. A resistor pad would be be across each memristor to create a limited maximum resistance. A resistor pad in series would set a minimum resistance. Not certain how necessary this will be. But I have certainly considered it. I have also considered a board with memristors tied to a common rail, which would very nearly double the memristor count compared to the pin count. A common system needs a split +/- voltage rail to work, or a virtual ground set between + and real ground to achieve the proper memristance effect. with both pins of a memristor brought out, you can achieve the memristance effect by simply reversing analog or logic states on each side.
      > > >>>
      > > >>> I am planning to make an actual PC board layout for this. It'll simplify assembly, ensure alignment, and mean It's repeatable by simply ordering more parts. What configuration would anyone else be interested in if I made a PC board layout that anyone could order from a PC board maker? Common connected memristors, or both leads brought out. Common doubles the memristor count, at the expense of needing +/- voltage rails (or to make ground half the total voltage and use the REAL ground as a negative voltage in relation to the other two voltages). Individual I/O results in fewer memristors or much more physical connectors, but is more flexible. I am personally opting for the individual lines option, but if there was demand for common lines, I might make a second board layout for it.
      > > >>>
      > > >>> Richard Piotter
      > > >>> richfiles
      > > >>>
      > > >>>
      > > >>>
      > > >>> Begin forwarded message:
      > > >>>> From: Richard Piotter <richfiles1@...>
      > > >>>> Subject: Re: [beam] Memristor
      > > >>>> Date: December 27, 2013 12:08:32 AM CST
      > > >>>> To: beam@yahoogroups.com
      > > >>>> Reply-To: beam@yahoogroups.com
      > > >>>>
      > > >>>> Glad to hear someone else sees the potential. Now I need to order parts... The rubber pad, the aluminum balls, better PC board material (the radio shack stuff is brittle... I want stuff that has strength and will flex instead of crack).
      > > >>>>
      > > >>>> Looks like the airsoft BBs are 6mm in diameter... A little bigger than I would like. I'll see if I can look at that welding supply source. If they can't help, then 6mm gap it is. I was hoping for about 2-3mm.
      > > >>>>
      > > >>>> I got a lot to do!
      > > >>>>
      > > >>>> Richard Piotter
      > > >>>> richfiles
      > > >>>>
      > > >>>>
      > > >>>> Begin forwarded message:
      > > >>>>> From: Martin McKee <martinjaymckee@...>
      > > >>>>> Subject: Re: [beam] Memristor
      > > >>>>> Date: December 26, 2013 11:45:23 PM CST
      > > >>>>> To: beam@yahoogroups.com
      > > >>>>> Reply-To: beam@yahoogroups.com
      > > >>>>>
      > > >>>>> It sounds good to me. As you say, it should be pretty repeatable. With that form of mechanical construction, the biggest area of uncertainty should, likely, be the creation of the sulfide layer which could, very likely, very within a fairly wide range given level of clean on the board, temperature, coating density, etc. But, it could definitely be a major improvement in uniformity; and, it's likely to be very strong. That would make it much easier to experiment with.
      > > >>>>>
      > > >>>>> Interesting,
      > > >>>>> Martin Jay McKee
      > > >>>>>
      > > >>>>>
      > > >>>>> On Thu, Dec 26, 2013 at 9:47 PM, Richard Piotter <richfiles1@...> wrote:
      > > >>>>>
      > > >>>>> So my previous assessment of the liquid conductive paint pen "seeping" through imperfections of the sulfide layer and bonding directly to the copper pad beneath seems to be confirmed after some dissection. I actually had a "dot" of silver paint that resists separation from the copper unless picked at with some effort.
      > > >>>>>
      > > >>>>> Long story short... Paint pen is a bust... :(
      > > >>>>>
      > > >>>>> Good news, is the most annoying motor line we produced at the old company I used to work at gave me a BRILLIANT idea! Back then, we made a motor driver that had triplets of MOSFETs held against a sil pad on a heat sink by a cross bar with rubber pads and screws threaded ingot eh bar from outside the heatsink. The idea I conceived falls back to the idea of using mechanical point contacts, but eliminates so many of the issues of the cobbled together bench top experiments that prevent hobby memristors from being actually... you know, useable for REAL world applications!
      > > >>>>>
      > > >>>>> http://richfiles.solarbotics.net/robots/MemristorAlBall.jpg
      > > >>>>>
      > > >>>>> To explain, I discovered aluminum balls are commonly available as airsoft BBs, as well as for use in welding applications. Ideally, one would construct a pair of PC boards, with pads that mirror one another in alignment corresponding to each memristor. You would drill several screw holes to evenly distribute tension across the boards. you would use a sheet of foam or rubber like material that is soft and compressible between the two boards. A hole would be punched in the foam corresponding to each memristor and each screw hole. The aluminum balls would be dropped into each hole corresponding to a memristor, and the top PCB would be screwed down onto the bottom PCB, sandwiching the aluminum balls and the foam/rubber layer between them. As the foam/rubber type material compresses, it will simultaneously secure the ball from vibration and shield the memristor surfaces front he environment. As the boar dis tightened, the ball should make a mechanical contact with the copper pad of the top board, and the Copper Sulfide layer of the bottom plate. This ought to form a stable AND protected Copper - Copper Sulfide - Aluminum Oxide - Aluminum junction... Our memristor.
      > > >>>>>
      > > >>>>> The technique remains 100% mechanical, but secures, protects, and maintains constant tension on all the junctions. By using a torque driver, one can even guarantee all screws have the same tension. This in combination with the uniform junction size, as defined by the small point of contact defined by the aluminum ball's actual point of contact, potentially leaves us with a UNIFORM manner of memristor construction, and possibly a uniform characteristic!
      > > >>>>>
      > > >>>>> So... Thoughts???
      > > >>>>>
      > > >>>>>
      > > >>>>>
      > > >>>>> Richard Piotter
      > > >>>>> richfiles
      > > >>>>>
      > > >>>>> Begin forwarded message:
      > > >>>>>> From: Richard Piotter <richfiles1@...>
      > > >>>>>> Subject: Re: [beam] Memristor
      > > >>>>>> Date: December 23, 2013 6:48:40 AM CST
      > > >>>>>> To: beam@yahoogroups.com
      > > >>>>>> Reply-To: beam@yahoogroups.com
      > > >>>>>>
      > > >>>>>> I think the focus so far has been using sulfur powder, seeing as it's readily available to hobbyists with no restrictions. There is less danger in it as well, considering the toxicity of said gas.
      > > >>>>>>
      > > >>>>>> I suspect my liquid circuit pen paint managed to find imperfections in the sulfide layer and seeped through them to the copper, thus explaining my approximate 50% short ratio. The ones that didn't, I'll have to experiment on, once I get the rest of my lab set up.
      > > >>>>>>
      > > >>>>>> Richard Piotter
      > > >>>>>> richfiles
    • Show all 64 messages in this topic