deskthority

macmakkara wrote: ↑

Here we Go!

Got to hand it to IBM's engineers: buckling spring was truly low profile (as demanded) compared to its ancestor. That beamspring module isnt even the whole height of the old mechanism! There was still the PCB and a metal plate to put under that beast.

Meanwhile, look at the homerow on the SSK. That's what they achieved. The curved backplate raises the function row up way beyond the internal mechanism's own height.

I wonder how long IBM engineers spent designing a keyboard from drawing boards to prototypes.

Well, IBM wasn't exactly famous for being agile, from what I know. I suspect they took their time and worked through a lot of prototypes.

If only they hadn't been ordered to design any more mechanisms after the Model F! The M and M2 are both artful creations that could have been a lot worse, but the primary objective was cost cutting of course. And ultimately IBM took that to its natural conclusion: no keyboards at all!

Yeah back in the day they had more time. I'm sure Model M was much more of a financial decision in terms of production than any predecessors. But even that result brought about the "legendary" keyboard which it is to many. The SSK being our choice in that family.

Even M2 is pretty good. I suggest every IBM keyboard fan give one a try. I was fortunate and got a couple from Cindy, of course.

That reminds me: anyone got an M2 spacebar spare? My first got crushed by the AT!

Nah I'll pass on the M2. I wanna go back, three generations back to you know what... beam me up Scotty.

I have all four(?) Generations stsrting from beamspring and ending in m2 and i can tell you i love both end and between. Theyr all unique to feel and experience. M2 has its own points and i also suggest all IBM fansbto give it a try.

Just came across this thread. Marking to read later.

I think that the only way we can do this is to use a different capacitive pattern than the one used by IBM - one that allows more direct routing of the rows in order to find ways around the tenon holes (where the bolt-mod bolts go). I put some research and thought into it and i'll present some idea here to see what you guys think.

Not all capactive sensors are the same. Not all support a matrix of keys (projected capacitance) and not all that do will accept more than one "touch" at a time. I found lots of "capacitive switch design guides" out there, but few covered the "mutual capacitance" type switch that we need. Two exceptions were design guides from Atmel and STMicroelectronics. Texas Instruments opens their design guide with "Capacitive touch detection is sometimes considered more art than science. This often results in multiple design iterations before the optimum performance is achieved."

First, a look at the model F switch. The top copper consists of two rectangular pads and two small circles. The two circles and the larger rectangle are floating (not connected to anything). The smaller rectangle is connected to a "row". The bottom copper has a rectangle directly under and the same size as the large floating rectangle on the top, but this is connected to a "column". There is also a thick ground trace around all of the traces and pads. The controller sends a signal down one column and that hits all of the rectangles on that one column. Each rectangle forms a capacitive couple with the rectangle above it. and this is picked up by the rectangles connected to the rows (or "sense lines"). How much capacitance is transferred from the column to each row is dependent on if the corresponding pivot plate is up or down - but not significantly impacted by the state of any other switch on the keyboard. This property allows for NKRO. The switch came in two flavors; signal pad on the left or right. The Model M construction method needs bolt holes in the space where the Model F puts "row traces". Four Triangles
This is probably most similar to the IBM design in that it has a large signal pad on the bottom (now two connected triangles) with a same-size pad floating above it (now two triangles). The sense pad has the same shape, but is rotated 90 degrees so that the shapes intersect in the center of the switch (but on different layers). This pattern is an improvement over the IBM design in that it allows us to route the traces through the pad rather than around them. It also doubles the coupling length between the signal and sense pads (better signal-to-noise ratio). However, the IBM design allowed traces to enter and exit out of three sides whereas this design allows rows to enter and exit out of just the two sides ; columns must enter and exit the top and bottom only. Grid and Flood
This design is the most unlike the IBM design. A big flooded sensor pad like on the IBM is more sensitive, but that is not always a good thing. With high sensitivity, the signal can get lost in the noise. The IBM needed the boost in sensitivity because the sensor pad was directly above a big ground plane (the back plate) which reduces sensitivity. One way to reduce the impact of the ground plane is to use the the signal pad as a shield. This design puts the sensor directly above the signal which increases capacitance, but the sensor has less area which reduces capacitance. It has about 10 times the coupling length which means much better signal-to-noise ratio. It also completely opens routing as signal and sense lines can enter or exit from any side and angle. Streets and Canals
This combines elements of the other two designs. It has the flooded signal pad and high coupling length of the "Grid and Flood" and also floating pads like the "Four Triangles" and original IBM. Columns can enter/exit anywhere, but rows can only enter/exit from the middle of the two sides. A final design might have 16 columns and 6 rows, like this: For more info,
https://en.wikipedia.org/wiki/Capacitive_sensing
http://www.atmel.com/images/doc10752.pdf
http://www.st.com/web/en/resource/techn ... 222015.pdf

Holy schnikies! I just stumbled across this thread and the project sounds awesome. I knew there was a reason I hadn't sold my 1987-vintage SSK, even though I haven't touched it since moving on to my F-122s and Kishsaver.

Somebody with a spare SSK needs to sell it to wcass for a reasonable price so we can make this project a reality.

If i had a spare SSK i would trade let him have it in exchange for a set of FSK bits when the time came

Unfortunately i only have the one.

If I can find a few volunteers, i might not need to do this myself.

I would need some volunteers to print out a PDF on 11x17" paper ($0.25 at the office supply store) and use this to verify the locations of the holes on the back plate and membranes. PCB production and shipping to US will run about $180 for 10 copies of the PCB, so if there are 10 folks reading this that would donate $18 each (plus shipping) - you could each have a PCB that may - or may not work. This is an untested pad design - there is NO GUARANTEE THAT THIS WILL WORK. But the change is 100% reversible and your $ investment would be pretty small. You would need about 85 model F pivot plates (an XT has 83) and an Xwhatsit controller too. Actually, i don't know if Xwhatsit controller will work with only 6 of the 8 rows connected up, so there is that too.

Steel bolts might be fine, but you might also look into taps and plastic screws to replace the tenons. McMaster has 8mm M2 pan head in white or black nylon at under $8 per hundred. M2 taps can be found for under $2 delivered on eBay. I would tap the hole and screw up from the back plate using no nut.

Also, if we know someone in the Cleveland, OH area - there is an electronics surplus supplier in Mentor, OH that is selling .156 pitch ribbon cable by the foot. Minimum order and shipping is a bit high, so a walk in buyer would be preferred. The whole group would need 10' total, so $3.50
http://www.electronicsurplus.com/molex- ... uctor-gray

Sign me up. I have two SSKs here.

I would need some volunteers to print out a PDF on 11x17" paper

I can do this for free, our printer at work has this size. I use it for schematics all the time.

You would need about 85 model F pivot plates (an XT has 83)

I have enough Bigfoots and Blue switch Fs here that I can borrow them for sure.

and an Xwhatsit controller too

I have at least one spare of these, possibly two.

i don't know if Xwhatsit controller will work with only 6 of the 8 rows connected up

It will. The Xwhatsit utility app allows (and encourages) ignoring entire rows or columns.

I would tap the hole and screw up from the back plate using no nut.

I've done this using a wood screw. From boltdepot.com I've used 4249 Wood screws, Slotted flat head, Brass, #1 x 1/4", and 5563 Flat washers, Stainless steel 18-8, #4.

OK, after you print this, measure the line across the top - it should be VERY close to 12.00"
If it is not, adjust the print scale to 101% or 99% or some fraction to get it as close as possible.

The next step sorta needs your FSSK disassembled. Lay one of the membranes (doesn't matter which one) on top of the drawing. Line up the hole at the bottom left of the Delete key with the corresponding hole in the membrane. If i measured correctly, all of the other holes should line up at the same time. Let me know if any are off and by how much. So long as my circles are bigger than the membrane's holes - it is OK.

Then take the drawing and lay it in the curve of the back plate. Again, make sure that one hole at the bottom left of the Delete key is aligned with the drawing, and check that the other holes line up too. I am sure that the one hole above the F1 key is NOT correct, so will need your input to adjust that. I would like to use that hole as a ground for the PCB.

Thanks

Maybe we could print a test PCB with various versions of the pads and test them with xwhatsit's controller? A 10x5cm PCB should fit enough pads to test a couple designs and the variable threshold on the controller will come in handy.
Also a set of calibration pads in the final design would be great for the controller to auto-set the threshold.
I'm really interested in funding the prototype PCB, but right now we're having trouble with payments outside Greece so I can't really commit.

I'd be in for $18 plus shipping for a prototype PCB, plus whatever the cost is for the ribbon cable. Actually, I'd like to use this opportunity to buy an extra foot or two of that cable just to have on hand for future projects or botched soldering jobs, if I can. Might as well get that if we're group buying now.

I saw some varying speculation earlier in the thread about barrels. Would this still use the stock M barrels and plastic top plate? I have an extra xwhatsit F controller I bought directly from him (I knew that would come in handy) and lots of F flippies laying around (from an XT and F-122s). I could also use XT or F-122 barrels if necessary, if a future revision of the design used a traditional metal plate.

Also, for anyone who's interested in doing this but doesn't have a source of F flippies, Eclipse has said he will have extra brand new F flippies for sale from his F77/F62 revival project. I think he'll have extras barrels available for sale too, if we end up needing those. Best to let him know ahead of time if you want to do that, so he orders enough. With a $325 apiece estimated cost, that might end up being the kind of project that only happens once.

Thanks for working on this, wcass! You are a real stand-up guy.

This would be a direct replacement for just the membranes. The pivot plates and controller would also need to be replaced, but barrel frame would not change. The thickness of the PCB would be 0.6mm - very close to the same thickness as the 3 membranes plus rubber mat. It could be put in a vintage SSK without damaging any existing parts (other than requiring a bolt mod). It should be possible to modify a Unicomp Classic or Ultra Classic and get two or three more keys than the original SSK (this PCB has the extra pads on the bottom row) but the case, barrel frame, and back plate would need to be cut down.

As for starting with a small test PCB, it would actually cost more (would need new back plate, case, barrel frame) and would you use that much after the test? I agree that calibration pads would be nice, but they are supposed to have close to the same capacitance as a "down" switch. It would be hard to match that without knowing what it is.

Using as much of the SSK as possible seems like a great idea, and I like that it'll be fully reversible, just in case. I agree that a specialized test PCB shouldn't be needed, especially if the full size version is only going to cost maybe $25 each shipped. Even going through 2-3 iterations, if necessary, shouldn't be a huge burden on people like us who spend upwards of $100 on a set of plastic keys caps!

I would be totally in for one PCB! $18 sounds good to me, just give me a FSSK already.

yeah me too at that price.

wcass wrote: ↑This would be a direct replacement for just the membranes. The pivot plates and controller would also need to be replaced, but barrel frame would not change. The thickness of the PCB would be 0.6mm - very close to the same thickness as the 3 membranes plus rubber mat. It could be put in a vintage SSK without damaging any existing parts (other than requiring a bolt mod). It should be possible to modify a Unicomp Classic or Ultra Classic and get two or three more keys than the original SSK (this PCB has the extra pads on the bottom row) but the case, barrel frame, and back plate would need to be cut down.

As for starting with a small test PCB, it would actually cost more (would need new back plate, case, barrel frame) and would you use that much after the test? I agree that calibration pads would be nice, but they are supposed to have close to the same capacitance as a "down" switch. It would be hard to match that without knowing what it is.

wcass: thanks for the progress ! I have not connected for a week and look what I found when I came back to the club. This is great.

I will print the pdf next week and give it a try on paper on the SSK that I already have with F flippers.
I have also a xwhatsit controller to test if there is enough space in the keyboard for all parts. We could also think of integrating the xwhatsit controller in the PCB. This will make the assembly simpler.

If this works, I could order a prototype and give it a try.

We are getting there

It would be nice to put the controller on the PCB, but i worry about bending the PCB under the surface mount components. My plan for attaching the controller is to use ribbon cable soldered to to surface mount pads on the top of the matrix PCB and route it similar to the tails on the membrane - over the back plate and down to the controller in the base of the case.

I am interested in this project! I assume that the goal is to create a true capacitive buckling spring (Model F) keyboard housed in an IBM Model M SSK case.

The idea is to take an SSK, replace the springs and pivot plates with those from a Model F, replace the membrane with a newly created curved PCB, solder on an xwhatsit controller, bolt mod everything back together, and hopefully end up with a TKL keyboard that has a (mostly) Model F typing feel.

You may need an older spec SSK for this to work, but that's not known for sure yet. It sounds like later models of SSK would at least need some modifications done to the barrel plate so the larger Model F pivot plates will fit..

I really hope this project will work, as i will have a shitload of F flippers in a few weeks...
I'm glad both of my SSK's are the old ones because they are no problems with them with the flippers.

I should also have enough xwhatsit controller for this project.

I've got an SSK barrel frame from Unicomp. Reckon that would work with this? I don't have a backplate to match it, but I could get one ordered up. I'm already in a little plate GB for the XTant.

You see, I've a few SSKs already and this would make a nice experiment in one keyboard's case. While I twiddle my thumbs and fear the bolt mod that SSK really needs!

What are we doing for Xwhatsit controllers? Teaming up with Ellipse?

How can we determine if our SSK is a correct version for this transformation? Is there a list of serial numbers or dates of suitable models?

Muirium wrote: ↑I've got an SSK barrel frame from Unicomp. Reckon that would work with this? I don't have a backplate to match it, but I could get one ordered up. I'm already in a little plate GB for the XTant.

You see, I've a few SSKs already and this would make a nice experiment in one keyboard's case. While I twiddle my thumbs and fear the bolt mod that SSK really needs!

Let us test first with our units. Once this is done, with the complete information on the table, it will be easier to make the right decision. Patience

What are we doing for Xwhatsit controllers? Teaming up with Ellipse?

Nothing still. I have a spare to test. Once the prototype is tested we will decide if it will be integrated in the PCB or not.
This is secondary at present.

Hypersphere wrote: ↑How can we determine if our SSK is a correct version for this transformation? Is there a list of serial numbers or dates of suitable models?

It depends on the barrel plate. The F flippers should fit. I do not know the relation with the serial / model numbers.
This investigation will come later.

OK, I printed the PCB pdf file with the following result: