Let's make our own beam-spring switches

[enthdegree]

Looks like IBM beam spring switches have the following parts:

• Housing
• Plunger
• Stem
• Top spring
• Beam spring, riveted to plunger
• Fly plate (two parts: the pad and the beamspring mount

I'm going to try my hand at making one. Here is my plan:

• Housing, plunger, fly pad can be 3D printed, sanded down by hand so that they fit together nicely. Add a Cherry mount to the top half of the plunger.
• Top spring can probably be sourced from somewhere
• Beam spring and fly can be etched from a sheet of spring steel (something like Amazon item # B00065V03A)

I think the most difficult part is going to be making the fly plate: Bending the fly into the right shape, getting it to mount securely on its pad. It's also made of a slightly thicker steel than the spring. Getting the beamspring securely onto the plunger might also be hard, but I might be able to mount it on mechanically or worst case just rivet it on.

[xueyao]

I think it's definitely doable but to achieve a precision and quality of the original requires specialised tooling and machinery which will be  extremely costly

[Leslieann]

Sanding 200 parts for a snug fit is not my idea of fun, especially when I know that pressing it slightly wrong would break it.
There are some resin printers that could do it, nearly at the same quality and strength as injection molded parts,  but you're going to pay through the nose for it.

[xueyao]

Sanding 200 parts for a snug fit is not my idea of fun, especially when I know that pressing it slightly wrong would break it.
There are some resin printers that could do it, nearly at the same quality and strength as injection molded parts,  but you're going to pay through the nose for it.

Agreed, but I think if OP wanted 1 or 2 replacement switches, it's definitely doable. To mass produce these would take an extraordinary amount of effort and budget put together.

[enthdegree]

People take apart, clean, lube and reassemble switches all the time. I cannot imagine hand-finishing each plastic piece is on a different level of involvement than that.

[Phenix]

If 'its worth the trouble' I guess hand-polishing all the parts aint be the problem and would be happily done for a beamspring build.

but beside switches we have to either design the beamspring 2. 0 as a cherry mx compatible switch or creating new pcbs (and maybe cases as well). Former is imho easier - IF the beamspring can be pushed inside a MX switch case.

Somehow i feel this is not easy. Is it possible?

[enthdegree]

but beside switches we have to either design the beamspring 2. 0 as a cherry mx compatible switch or creating new pcbs (and maybe cases as well). Former is imho easier - IF the beamspring can be pushed inside a MX switch case.

Somehow i feel this is not easy. Is it possible?

Some form beam-spring-type switch in a cherry-sized package should be possible since Alps did it (google "alps plate spring") but not with IBM's design which is as thick as an entire modern keyboard.

I am not a mechanical engineer and don't know how to determine what the weak points of a design are, so I'm just going to copy IBM's for the first prototype. I think I will just try to 3D scan the plastic parts, then edit the plunger to have a cherry stem.

[xueyao]

but beside switches we have to either design the beamspring 2. 0 as a cherry mx compatible switch or creating new pcbs (and maybe cases as well). Former is imho easier - IF the beamspring can be pushed inside a MX switch case.

Somehow i feel this is not easy. Is it possible?

Some form beam-spring-type switch in a cherry-sized package should be possible since Alps did it (google "alps plate spring") but not with IBM's design which is as thick as an entire modern keyboard.

I am not a mechanical engineer and don't know how to determine what the weak points of a design are, so I'm just going to copy IBM's for the first prototype. I think I will just try to 3D scan the plastic parts, then edit the plunger to have a cherry stem.

Does the beamspring support the mx in terms of spacing and profile of keycaps?

[enthdegree]

The switch's footprint is 19mm x 19mm which means the fit will be tight at 1mm for cherry spacing. But it is possible if the switches are mounted in a tray over the sensing mechanism, like in a real beamspring board.

[0100010]

I'd be all over new production beamspring switches, along with custom layouts; 62 key HHKB-ish to start.

[xueyao]

I'd be all over new production beamspring switches, along with custom layouts; 62 key HHKB-ish to start.

If the reproduction quality can reach IBM quality, I suspect it'll probably cost upwards of 2k+

[0100010]

I doubt that much.  Look at what Ellipse has done with Model F remakes.

[Phenix]

The F is imho a way easier design.

F is
PCB , flipper/spring, barrel frame, keycap

beamspring switches are afaik way more difficult - more internal parts. And those parts have to be machined more precisely..

[Leslieann]

I'd be all over new production beamspring switches, along with custom layouts; 62 key HHKB-ish to start.

If the reproduction quality can reach IBM quality, I suspect it'll probably cost upwards of 2k+

To match IBM quality, $2k is the cost of one half of the final mold.
By the time you had your molds in the machine, ready to press the button, I wouldn't be surprised if you have $40k invested if you have never done anything like this before. Half that if you have.

People take apart, clean, lube and reassemble switches all the time. I cannot imagine hand-finishing each plastic piece is on a different level of involvement than that.

Cleaning a 3d print, also called post processing  is not the same as wiping off some dust or lube.
On something this precise, I could easily see spending 30-120 minutes per switch with a jewelers file in hand looking through a microscope, making these fit like they should.

Cleaning a 3d print involves cutting off any support material, stringing, sanding off any blobs or ridges, and then there is the fact that 3d printing can't produce precise numbers. You cannot make a wall thinner than the nozzle is wide, you cannot make a corner sharper than the radius of the nozzle, anytime you start and stop, the nozzle can leave a small imperfection, which is fine on most parts, but not two internal parts that needs to slide together. Then there is the layers, which do not slide across each other smoothly.

Resin is better, but that too involves a lot of cleanup work after, you need to wash it, then let it cure, and hope it doesn't distort during this time. You still also have layers that don't slide past each other and support structures to remove.

[enthdegree]

Alright I  put this project on hold after I read that beam-springs were meant to feel like a Selectric which is supposed to feel the very best.
As a new owner of a working Selectric I kind of disagree. I feel like people make claims that Selectrics feel better just for the elitism of saying they prefer the most archaic, unsupportable and unobtainable gear possible. Selectric keys feel practically linear until actuation, and most of the tactility comes from mechanism jolting the machine around. After a little while it feels not far removed from pushing against flat plastic which is uncomfortable. The keys are also way more wobbly than any computer keyboard I have ever tried.

So now I'm back to trying to build beamsprings.

[__red__]

Thread, ARISE!

I might have a resin printer.
I might have working parts for everything except the actual metal pieces.

Anyone here know about metal tooling?

[senso]

Dies for pressing are more or less the same price, be it metal or plastic, plastic injection molds might be a tad more expensive, because you dont usually hit the nail on the first, nor the second try, but that being said, expect some 10k+ to order all the tooling.