Custom Buckling Spring Membranes

[CPTBadAss]

SUCCESS!

Despite doing a truly awful job at soldering on the diodes, > 50% of the keys are working exactly as they were designed to do. The keys that don't work are the top and bottom rows (Esc, numbers; Ctrl, Alt, GUI, Space, left, down, right), and two columns (Y, H, B; close bracket, Enter, right Shift).

I tested for NKRO using AquaKey. Every combination of WASDQZ that i tried yielded correct result.

So now i disassemble and try to resolve the issues with the dead columns and rows. I'm sure my poor job with the diodes is the problem with the columns, but not sure about the rows.

Hooray!! Great job wcass!!

[wcass]

...

Can you recap what materials and production methods you're using now, there were so many iterations that I got a little lost

Membranes keyboards are cheap if produced in LARGE volume ($5 a set at 5000 units); in small volume, they are actually quite expensive ($100 a set at 5 units). They are also reliable, but the design does not lend itself to adding diodes mid-circuit for back-flow prevention. In other words, two key roll-over at best. This is an experiment trying to create custom membranes that are relatively inexpensive at low volume, and also introduce diodes into the design to enable NKRO.

A Membrane keyboard matrix consists of 3 thin layers sandwiched together; a bottom conductor layer with traces on its top side, a top conductor layer with traces on its bottom side, and a middle separator layer with no traces but holes where the keys would be. When you press on a “key” the thin top layer deforms a little over the hole in the middle layer and a conductive trace on the top layer touches a trace on the bottom layer to complete a circuit and register a key-press. 

Any “lumps” in the layer sandwich would change the way that the top layer deforms so placing diodes near the keys would probably prevent proper operation. I avoided this by making the bottom layer larger than the other layers and putting diodes on the bottom layer outside the border of the top and middle layers. By using diode arrays, I keep the chip count low and also join a group of keys into a logical column. I used 14 5x diodes for this, but other unidirectional “TVS” diode array should work.
http://www.digikey.com/product-search/en?x=0&y=0&lang=en&site=us&KeyWords=DMF05LCFLP

My bottom layer is an 8 mil (.2mm) thick PCB which is usually the thinnest substrate available from manufacturers. I made my own and assembled it myself, but do not recommend anyone else doing this except for a prototype. My middle layer is from a Model M but cut down to the right size. My top layer is 4 mil thick plastic film that I had printed at InJetFlex.
http://inkjetflex.com/site/substrate-materials-rapid-prototypin

Other parts of my M-60 project include …

• steel back plate, mat, hammers, springs, key caps from a donor Model M
• barrel frame and key caps from Unicomp
• USB panel mount , Teensy++ from PJRC
• FFC connectors from Digikey
• 20 gauge steel sheet from Home Depot

[Hellmark]

How feasible would it be to offload the diodes into a PCB that the membrane connects to?

[wcass]

How feasible would it be to offload the diodes into a PCB that the membrane connects to?

It would require one connector pin per key, so would need a connector something like this:

[Hellmark]

Something like that is completely doable, even if you ran in a couple ribbon cables.

[damorgue]

How are the sheets cut? Another option might be to put holes in the top layer at the locations where the diodes are placed if you place them next to their respective switches on the bottom layer.

You mentioned approximate prices for very large and small quantities. How much are they in medium quantities? I don't expect 5000 to be sold and 5 is few even for a prototype run.

[bcg]

Membranes keyboards are cheap if produced in LARGE volume ($5 a set at 5000 units); in small volume, they are actually quite expensive ($100 a set at 5 units). They are also reliable, but the design does not lend itself to adding diodes mid-circuit for back-flow prevention. In other words, two key roll-over at best. This is an experiment trying to create custom membranes that are relatively inexpensive at low volume, and also introduce diodes into the design to enable NKRO.

A Membrane keyboard matrix consists of 3 thin layers sandwiched together; a bottom conductor layer with traces on its top side, a top conductor layer with traces on its bottom side, and a middle separator layer with no traces but holes where the keys would be. When you press on a “key” the thin top layer deforms a little over the hole in the middle layer and a conductive trace on the top layer touches a trace on the bottom layer to complete a circuit and register a key-press. 

IMO your design is very sensible and I think it is a solution that doesn't stray too far from the original engineering.

The solution of running individual traces to the controller (as opposed to a matrix) seems kludgy and also if you have individual lines going to the controller anyhow, it might make more sense to just use shift registers instead of those diodes.  I'm sure soldering those tiny qfn packages is a blast

This looks really good and I look forward to testing it out.  I've not etched PCB before but I'm not opposed to giving it a shot... but what is the quantity of PCBs that it makes sense to have a tested design fabricated?

Also I'm sure not an expert on this stuff but unless I'm missing something think it might be possible to do the NKRO in software... I'm going to think about it a bit more and if I haven't decided that my idea is completely stupid, I'll post what I'm thinking so that you guys can convince it is

Anyhow great job wcass!

EDIT:

BTW, how do you cut the steel plate?

[wcass]

How are the sheets cut? Another option might be to put holes in the top layer at the locations where the diodes are placed if you place them next to their respective switches on the bottom layer.

You mentioned approximate prices for very large and small quantities. How much are they in medium quantities? I don't expect 5000 to be sold and 5 is few even for a prototype run.

Sheets are cut by hand. I thought about creating holes for the diodes and even considered putting them on the back (with holes in the back plate), but this idea seemed simplest and most likely to succeed.

Traditional silver silkscreen membranes cost a lot to do the first one (make frame, pattern, tooling, etc), and the last one (break down, clean and store tooling). Every one in the middle costs $4. The copper membranes from InkJetFlex are $50 for the first two and $10 each after that.

IMO your design is very sensible and I think it is a solution that doesn't stray too far from the original engineering.

The solution of running individual traces to the controller (as opposed to a matrix) seems kludgy and also if you have individual lines going to the controller anyhow, it might make more sense to just use shift registers instead of those diodes.  I'm sure soldering those tiny qfn packages is a blast

This looks really good and I look forward to testing it out.  I've not etched PCB before but I'm not opposed to giving it a shot... but what is the quantity of PCBs that it makes sense to have a tested design fabricated?

Also I'm sure not an expert on this stuff but unless I'm missing something think it might be possible to do the NKRO in software... I'm going to think about it a bit more and if I haven't decided that my idea is completely stupid, I'll post what I'm thinking so that you guys can convince it is

Anyhow great job wcass!

EDIT:

BTW, how do you cut the steel plate?

Thanks for the praise. I wasn't 100% sure that it would work, so that's why I went DIY. Now that i know it does work, it might be time to price a group buy of PCBs with assembly. I've never done a group buy, so i would want to team up with someone who has.

I cut my steel plate using a jig saw with metal cutting blade. If this does become a group buy, you can expect me to do several videos showing and explaining every step of the assembly.

[damorgue]

Ideally we should make a run of custom BS boards later on with all the other parts as well. These are replacement membranes which fit current boards, correct?

[Melvang]

While I haven't been following this thread like I should be (since it is paving the way to custom layout BS boards) would it be feasible to do membrane top and middle layer with a PCB on the bottom layer?  This would allow full diode placement using SMD.  Just thinking out loud here.  If it isn't feasible at all I understand.

[Hellmark]

The only problem is that you'd have to have a curved PCB. M's have a curved steel plate. While most get their normal curve by having different key shapes, M has uniform key shape and curved plate.

[Melvang]

The only problem is that you'd have to have a curved PCB. M's have a curved steel plate. While most get their normal curve by having different key shapes, M has uniform key shape and curved plate.

I am guessing that having curved PCB's made would be borderline prohibitively expensive?

[Hellmark]

Yup. Copper clad boards are always done on flat sheets, unless you custom order it. Also, the materials it is made from don't take to flex well. Trying to curve it would just break it.

[Melvang]

Yup. Copper clad boards are always done on flat sheets, unless you custom order it. Also, the materials it is made from don't take to flex well. Trying to curve it would just break it.

I wonder if any online shops can do curved PCBs.  Or if it would be easier to have caps made to work with a flat pcb?  The only reason I am suggestion this would be for easier custom cases/layouts.

[regack]

Yup. Copper clad boards are always done on flat sheets, unless you custom order it. Also, the materials it is made from don't take to flex well. Trying to curve it would just break it.

I wonder if any online shops can do curved PCBs.  Or if it would be easier to have caps made to work with a flat pcb?  The only reason I am suggestion this would be for easier custom cases/layouts.

How about a series of horizontal flat PCBs that follow the curve of the plate and are joined by flexible interconnects.

[Melvang]

Yup. Copper clad boards are always done on flat sheets, unless you custom order it. Also, the materials it is made from don't take to flex well. Trying to curve it would just break it.

I wonder if any online shops can do curved PCBs.  Or if it would be easier to have caps made to work with a flat pcb?  The only reason I am suggestion this would be for easier custom cases/layouts.

How about a series of horizontal flat PCBs that follow the curve of the plate and are joined by flexible interconnects.

We can't be having any of that.  That makes to much sense.

[Hellmark]

Having a bunch of little PCBs could be a bit of a pain, as it adds a lot of places for things to go wrong, and a ton more soldering.

What I personally would suggest, and this may be something that would be of far greater interest, would be to have flat PCB, and keystems designed to take MX keycaps. That way, you could have flat boards if you wanted, since those do seem to be quite popular, and if you do want a contoured shape of the keys, then you could go get keys that are sculpted that way. It would require custom fabrication of plastic parts, and injection molds are expensive, but I think a crap ton of people would be interested in something like that.

[CPTBadAss]

Would your idea be cool Hellmark? Personally I like the keycaps and curve of a Buckling Spring board and think its part of the charm. But that doesn't mean I think its a bad idea and could be cool. Do we have enough people to support buying new keycap molds and custom plastic molding for new parts? I doubt it.

Unfortunately I don't think there's enough demand to help fund the effort...or a manufacturer that would do low volume stuff like that.

[Hellmark]

Thinking about it, it might be possible for us to make curved PCBs. We just need to find a way to get some sort of heat resistant material curved, and then copper plate it. Copper plating should be semi decent to do, and done in most cities, since it is a step in chrome plating of car parts. The big thing is finding a heat resistant material that could withstand curving.

[CarVac]

If you want to subject yourself to the hazards of phenolic resin, you can lay up the fiberglass/phenolic board on a curved surface by hand. That way you won't need to bend a premade part.

Epoxy is slightly inferior as a resin for circuit boards, but for a keyboard that doesn't need the heat or fire resistance of FR4, you could do it without needing high-temp curing.

[wcass]

While I haven't been following this thread like I should be (since it is paving the way to custom layout BS boards) would it be feasible to do membrane top and middle layer with a PCB on the bottom layer?  This would allow full diode placement using SMD.  Just thinking out loud here.  If it isn't feasible at all I understand.

That is exactly what i have done.

The PCB base matirial i use is .2mm (or 8mil) thick. This is the thinnest commonly avaiable from PCB fabs. This material is flat, but will take the gentle curve of an M backplane without breaking. I should sumerize this whole thread when i get home.

[tinco]

Did you cut holes in the membranes for the diodes? Wouldn't they be crushed between the steel plate and the plastic?

[wcass]

the keys are where the red is. blue arrows point to where the left 3 diode arrays are. these 3 chips handle 15 keys.

some pictures of the controller and how the membrane and PCB attaches



older picture, but still accurate:

[bcg]

some pictures of the controller and how the membrane and PCB attaches

hasu pointed me in the right direction to integrate Adafruit's Bluefruit controller with the tmk firmware to make a Bluetooth keyboard; not many extra components for this, just:

the Bluefruit board (http://www.adafruit.com/products/1535 - $19.95)
charge controller (http://www.adafruit.com/products/259 - $12.95)
battery (http://www.adafruit.com/products/328 - $14.95)
plus misc components like switches or buttons for pairing, LEDs for status idicators, etc

Maybe could get these components cheaper but I haven't looked that hard yet. 
My goal is to get to the point where I could plug the keyboard into USB for charging & for use with computers that don't have bluetooth, and unplug it and switch over to bluetooth for use with a laptop or another device.  Anyhow I was thinking I would want to solder everything up to a single PCB - its not obvious to me how you have the teensy attached to the bottom membrane, but since that is through-hole it would probably be easy to attach anything in the manner, right?

I have some working code already and can type with a Model M over the bluefruit controller, but I think the first version of that device doesn't support raw HID reports very well... they are sending me a replacement that is the newest version, which is nice because that also supports mouse & media controls so I should be able to easily enable those features in the firmware so that you could use hasu's mouse keys and control volume etc.  All I really have left to program is the logic to switch between USB and bluetooth to make this happen.

BTW what are the the three header pins sticking off of the teensy; are those for your lock LEDs?

[wcass]

I'll watch for your progress.
Yes, the 3 pins are for num lock and caps lock (and +5).

The controller is connected to the PCB by a 17 pin "right angle" FFC. This makes it very easy to assemble and disassemble but hard to take a picture of. Here is a picture of the top membrane fitting into a similar 5 pin "strait" FFC that connects the top membrane.


I wanted to share my method for assembling the key switch layers that solves one of the common problems when assembling a buckling spring keyboard.

The springs on BS switches are a little longer than the frame barrels are deep so that if the barrel frame top is flat against a surface then the springs will not seat properly. My previous solution for this was to use whatever junk i had on hand to wedge around the edge of the frame to prop it up. The problem with this is that a slight bump might knock one of the props loose and some of the springs will unseat.

I start with a box about the size of the keyboard and cut a hole just big enough to fit around the barrels. Pictures show it best. Also showing the assembly steps.

springs in and nicely seated

pad

top membrane with contacts face up

middle membrane is clear, so it looks the same as the picture above.
bottom membrane (OK, it is a PCB) with contacts face down

back plate

nuts on and controller attached. tape keeps LED wires from getting pinched.

[dgreekstallion]

^^^^


You are a genius. I salute to you, sir.

[bcg]

I'll watch for your progress.
Yes, the 3 pins are for num lock and caps lock (and +5).

The controller is connected to the PCB by a 17 pin "right angle" FFC. This makes it very easy to assemble and disassemble but hard to take a picture of. Here is a picture of the top membrane fitting into a similar 5 pin "strait" FFC that connects the top membrane.

Ok, I see now.  That looks like it would be easy to connect another PCB or something to mount the controller & other components.  I have my bluetooth circuit & code working well now, and it should be easy enough to integrate with anything using hasu's firmware.  I'm going to post the details to a separate thread, but its a really busy time for me so it might be a bit before I can do that.

Sort of off topic, but did you order all of your keys new from Unicomp?  Or did you harvest them from other boards?  Just curious because I'd like to get some alpha keys with the number pad printed on them like on an original SSK

[wcass]

Sort of off topic, but did you order all of your keys new from Unicomp?  Or did you harvest them from other boards?  Just curious because I'd like to get some alpha keys with the number pad printed on them like on an original SSK

most of those keys came new from Unicomp.

[joneslee85]

This is incredible piece of work wcass! You are amazing. I think you should create kick starter campaign to help funding your project and allow you to mass produce it later on.

[bcg]

wcass

Let me know what you think about this product (important part starts at 18:18):

What I'm thinking is to make both sides of the membrane out of the printed sheets, without any PCB etching... then just stick the standard plastic barrier from the Model M in between... at the top of the printed circuit you could have printed terminals and you could attach surface mount I/O expanders or Parallel to Serial Shift Registers ... if you use 8 or 16 bit shift registers you wouldn't need a teensy++ anymore, could use something smaller/cheaper so you could save some cash, also hasu uses 8 bit rows in his matrix array so would be super easy to scan the matrix... plus no cost for the copper clad or the chemicals... just stick the ICs on with the z-tape and then run two flex "cables" to the microcontroller, basically like you're doing now, just without having to solder those tiny QFN diodes

If its durable enough I think it could make this project really feasible.  Could even design the printed sheets as a full 101/103/104 layout and make it so that you could optionally cut off the number pad, arrow keys, function keys, etc to make smaller layouts if you want.

What do you think?  I'm up for splitting the cost of a test run if you're interested...

[wcass]

I looked into Z tape before deciding to do it this way. The diode array i use are really small (1.6mm x 1.6mm); each pad is only .08 mm² and has just .25mm between pads. From the datasheet, the tape needs 3.2 mm² pads minimum and at least .4mm between pads.

But an 8 by X matrix is definitely doable.

[mkawa]

is there a datasheet on the z-tape anywhere? the PSAs with really high tear strength and vibration resistance are ridiculously expensive. i have my doubts about tqfn joints lasting at a curvature under constant vibration.

and then there's the geometric requirements due to resistance of the material. it's a doped dieletric, most likely, and those tend to have really high resistance unless you give them oddly shaped traces.

[bcg]

Yeah, sorry I should have been more clear... I was talking about not using the QFN diodes at all, and using something larger like this:

http://www.digikey.com/product-detail/en/74HC597D,653/568-8161-1-ND/2762851

although if you need 3.2mm that's not really going to work regardless... guess I'll just have to get the photoresist gear to try this out and do as you did wcass

EDIT:  This is what made me think that this would be possible for this project... maybe they have the ICs attached to the flexible circuits via a different method though and the z-tape is just for the sticker part - http://www.crowdsupply.com/chibitronics/circuit-stickers

[bcg]

is there a datasheet on the z-tape anywhere? the PSAs with really high tear strength and vibration resistance are ridiculously expensive. i have my doubts about tqfn joints lasting at a curvature under constant vibration.

and then there's the geometric requirements due to resistance of the material. it's a doped dieletric, most likely, and those tend to have really high resistance unless you give them oddly shaped traces.

Can you define what PSA means?  Google says its a prostate test and to go to the second page of search results would be very 1990's of me...

datasheet: http://www.adafruit.com/datasheets/3m%209703.pdf

[damorgue]

Can you define what PSA means?

Probably Pressure Sensitive Adhesive.

I doubt it is public service announcement

[mkawa]

pressure sensitive adhesive is adhesive that cures when you put compressive force on it. usually they're modified thermosets.

[mkawa]

also i suspect that what's going on with this z-tape is that it's just conductive enough to conduct with the assistance of a large nearby conductor given enough free electrons for current, kind of like a semiconductor but maybe mixed into a siloxane copolymer substrate or something. what i would be afraid of with a model F is that the AC impedance between two geometric points on the tape that are both in contact with different pieces of metal conductor will be non-zero. that is, DC resistance may be infinite, but there will be weird capacitive effects that will screw up the capacitive switches and matrix

[idollar]

impressive work. The reporting was even more impressive.
At the end of this post https://geekhack.org/index.php?topic=40111.msg1102755#msg1102755  there is a picture of the keyboard.
Is this the final keyboard or did you build a different case for it ?

Cheers

[wcass]

Yep, that's it. Very primitive.

I got a Kishsaver, so have not done anything with this since. The next logical step would be to get a proper professionally made PCB and case. Might build a controller onto the bottom membrane.

[ccc24]

is there any change this will be going into production or a GB?

[wcass]

I don't think the cost and quality is good enough to go into production. I did make an attempt to take it to the next level ... custom layout in custom case, professionally produced PCB and membrane, controller built onto the PCB. It was at the 2015 KeyCon.


The membrane + PCB was more than 2x the cost of one PCB and the controller much more difficult to program than xwhatsit's - so why not go model F style?


If you have not seen Ellipse's F62 reproduction thread, check it out.