Xwhatsit/CS-Compatible IBM 3277 PCB

[emdude]

Wow, that layout is pretty clever, very TKL-like. I still have mine in its original layout; the block nav is very annoying.

Hm, I might do something like that! It's a good thing that the non-alpha block keys are uniform profile because swapping keycaps would drive me nuts otherwise.

[JP!]

Put me down for 1 66 key pcb please

[subcat]

You're a god, thanks for all the work you've done on this

[emdude]

A small follow-up: I want to thank orihalcon again for running the GB for the 66-key PCBs! He informed me that they all work fine, which is great!

Also, at his and JP's request, I've begun work on a IBM 3740-series keyboard PCB, as well as one for the 5100-series keyboard. I'll periodically update the thread with major progress. These too will go on the github repo with the 3277 CAD/EDA files when everything is said and done.

For now, I just have these pretty rough matrix mock-ups:

5100:

Spoiler:

3740:

Spoiler:

[__red__]

A small follow-up: I want to thank orihalcon again for running the GB for the 66-key PCBs! He informed me that they all work fine, which is great!

Why have I not seen this thread before? I have duplicated so much work :-/

Awesome stuff emdude@!

[The_Boom_Boy]

Wow this is a sweet project, solid work! If only I could get my hands on one of these beauties. Voted for it in the desk authority awards GL

[Wodan]

Actually made a video of my IBM 3277 PCB replacement:

[snhstq]

Removed

[Excelso]

Actually made a video of my IBM 3277 PCB replacement:

Quite a beauty. Being spanish myself, those keycaps and that keyboard are way beyond unobtanium.

[Croktopus]

I'm looking at designing a PCB of my own, but was curious about a few pretty fundamental things that I haven't seen mentioned.

1. the top side of the pcb has two pads for each switch, one appears to not be routed to anything, and the other appears to be the "row" pad. im guessing the pcb functions by the flipper essentially connecting those two pads and that changes the capacitance between the column and row, which is then read as a keypress? the main thing i want to confirm here is that, on the opposite side from the column pad, there is a completely unconnected rectangle of copper (ive seen some pictures where this pad does seem to be connected, hence my curiosity). also, all 3 of these pads should have the copper exposed? or is it ok to have solder mask over some of them

2. the trace that goes around all the column pins, is this a ground trace? if so, is there a reason why it can't simply be a ground fill? I saw some mention about concern of it being on the other side of the row pad, but is that the only reason? (can have the groundfilll go everywhere but beneath the rows)

3. has anyone experimented with doing flexible capsense pcbs, so that you get a curved profile? this is really a question about the model f, but iirc its pcb was curved, and i was thinking it might make more sense to do a flexible pcb that rests on a curved surface, so that it gets that same curvature. or is this a bad idea (never worked with flexible pcbs, so im sure there are things that i havent considered)

[Muirium]

The flipper does NOT conduct electricity between those pads. In fact, as I remember, the sense pads are all insulated. These boards are capacitive sensing, remember. No need for exposed conductors. Your PCBs job is to supply the controller a couple of reference points for each key, which the controller then measures for capacitance.

I remember Xwhatsit talking about the third pad, somewhere. IBM was very aware of signal vs. noise and it was something to do with helping that with a passive ground pad, if I recall. This was a few years ago, mind.

His manual is well worth a read. It might cover this, and certainly helps understand these boards:

http://linode.cornall.co/ibm-capsense-u ... spring.pdf

[SneakyRobb]

I'm looking at designing a PCB of my own, but was curious about a few pretty fundamental things that I haven't seen mentioned.

1. the top side of the pcb has two pads for each switch, one appears to not be routed to anything, and the other appears to be the "row" pad. im guessing the pcb functions by the flipper essentially connecting those two pads and that changes the capacitance between the column and row, which is then read as a keypress? the main thing i want to confirm here is that, on the opposite side from the column pad, there is a completely unconnected rectangle of copper (ive seen some pictures where this pad does seem to be connected, hence my curiosity). also, all 3 of these pads should have the copper exposed? or is it ok to have solder mask over some of them

2. the trace that goes around all the column pins, is this a ground trace? if so, is there a reason why it can't simply be a ground fill? I saw some mention about concern of it being on the other side of the row pad, but is that the only reason? (can have the groundfilll go everywhere but beneath the rows)

3. has anyone experimented with doing flexible capsense pcbs, so that you get a curved profile? this is really a question about the model f, but iirc its pcb was curved, and i was thinking it might make more sense to do a flexible pcb that rests on a curved surface, so that it gets that same curvature. or is this a bad idea (never worked with flexible pcbs, so im sure there are things that i havent considered)

The first ibm capacitive pcbs had both pads on the front. Later on they switched to the pads on either side. The later 2 sided pcbs work just fine. I have made pcbs with both pads on the front but I did it for fun and to be kind of "retro."

I suspect that the reason they changed to front and back pads for the model f, is that they thought the bending might damage the vias that are necessary when you have both pads on the front.

So if you are making a flexible model f pcb, I would recommend trying to make pcbs where the columns and rows are on different sides like the emdude 3277 pcb in this thread or most common model f ones you see.



For the front unconnected pad.
I'm not aware of official IBM documentation about this but the common reasoning in PCB design is that "unconnected" copperfill can act as an antenna or capacitive coupler(?) and cause electrical interference between traces. In this case this is effect is desirable.
I think the IBM engineers must have thought that the signal between 2 pads on different sides needed a bit of help. They must have put that unconnected pad there to increase the signal.


I believe flexible pcbs work just fine for model f style keyboards. When you see the occasional custom model f, they are usually flat pcbs. This is not because the pcb cant bend, it is because it is difficult source custom curved metal to hold it all together. So people just make them flat.

[manisteinn]

3. has anyone experimented with doing flexible capsense pcbs, so that you get a curved profile? this is really a question about the model f, but iirc its pcb was curved, and i was thinking it might make more sense to do a flexible pcb that rests on a curved surface, so that it gets that same curvature. or is this a bad idea (never worked with flexible pcbs, so im sure there are things that i havent considered)

I made a Model-F numpad PCB for testing last year, the board is tightly sandwiched in a 2-part 3d-printed curved case and screwed in place through holes at the edges of each row. My PCB top surface bend radius is 300mm and I used the original's thickness of 0.8mm without any issues (0.6mm is the same price at JLCPCB). I shelved the project but I'm planning to share the CAD data soon.

[DMA]

Crocktopus,
Floating pad on the top side is mostly because it makes the pad same height. Same reason for the round dots on the top side of the key footprint - flipper legs are standing on those.
Also it boosts signal somewhat - but you know what boosts signal even more? Direct electrical connection. My experiments shown 4x boost if memory serves.
Soldermask them all over, prevents corrosion. Or if you're planning to expose all that gold for an additional bling factor - go ahead, but don't have any solder mask anywhere within the key area, to keep things level. And also don't do that if it's not ENIG - things will rust thru pretty quickly.

The outside trace is ground - you can do the ground fill on all free space you see, it won't affect the signal, but why so much obsession with ground fills around this forum? What are you guys trying to achieve with those? It's not RF, it's not 100A. It just makes things look uglier. The only exception - if you plan to make a keyboard without a conductive plate under the PCB. You'll need that ground plane in this case.

Flexible as in really flexible PCBs (polyimide, mylar printing) are super-expensive - some of us got quoted like $700/full size keyboard.

Model F PCBs were NOT curved, they were normal PCBs, the only person who ever did PCB bending is Ellipse. They're just 0.8mm, not usual 1.6, so they do bend without breaking.
You can go 0.6mm or even 0.4, it will make things easier to keep together as PCB won't produce as much force on your fasteners trying to get flat again.

[ro0llo]

Are those still being sold or will there be another gb in the future? Since yesterday finally got a beamspring!!

[kmnov2017]

Are those still being sold or will there be another gb in the future? Since yesterday finally got a beamspring!!

You can order one on your own from JLCPCB.com
Find the Gerber files on this thread, upload to JLCPCB and pay.

[ro0llo]

Are those still being sold or will there be another gb in the future? Since yesterday finally got a beamspring!!

You can order one on your own from JLCPCB.com
Find the Gerber files on this thread, upload to JLCPCB and pay.

Oh thats website is gold and a lot cheaper then I expected, thanks!

[TheNeonKnight]

Anyone happen to know which xwhatsit's work with this pcb?