Beam Spring Switch Reproduction Project

User avatar
superfoil

23 Jul 2021, 20:25

Hello all, I posted a similar thread about this elsewhere, but I figured I'd also share here since Deskthority has been an extremely helpful resource in the course of this project.

I'd like to share (perhaps a bit prematurely) the project I've spent the last year working on.

With the average price of a single Beam Spring switch somewhere in the neighborhood of $40, and seeing first hand the rust damage these switches can endure when not properly cared for, I set out to attempt a modern production.

The Project:
My primary goals were: to create a cheaper, functionally identical, replica of the original Beam Spring switch and to update the compression spring and key stem to a stainless steel variant to prevent future rust issues.

This effort is still very much a work in progress, but after hitting a major milestone (finalizing the key module mold), I figured I'd start this thread to share my progress.

Currently, I have completely and accurately created CAD models of all the components in a Beam Spring switch assembly.
I have also:
  • Created and finalized the key module injection mold
  • Produced replica beam spring components
  • Produced replica compression spring components
Going forward, I plan to:
  • Finalize production of the fly spring (currently in progress)
  • Design and produce injection mold for the key column
  • Design and produce injection mold for the fly plate
  • Begin production of the key stem
  • Procure compatible O-rings
Below are some comparison photos of original and replica components:

keyModule-fs8.png
keyModule-fs8.png (138.43 KiB) Viewed 1653 times
The key module. The replica is a sample mold, and the finish isn't final.
The finalized mold will be more consistent with the glossy surface of the original.

factoryPic.png
factoryPic.png (262.66 KiB) Viewed 1653 times
I received this picture this morning of the finalized mold, and they certainly seem more similar to the original.

compSpring-fs8.png
compSpring-fs8.png (97.54 KiB) Viewed 1653 times
The compression spring. The replica is shiner due to age and material.
The replica is made using stainless steel, while the original uses music wire (which is prone to corrosion).


beamSpring-fs8.png
beamSpring-fs8.png (340.35 KiB) Viewed 1653 times
The beam spring. This part was particularly hard to produce due to its extremely thin height.

keyColumn-fs8.png
keyColumn-fs8.png (146.19 KiB) Viewed 1653 times
The key Column. As the injection mold for this part will be quite complex,
I only currently have a 3D printed version.


assemblyView.png
assemblyView.png (23.33 KiB) Viewed 1653 times
Rendered assembly of a beam spring switch using my CAD files.

Once I have all components produced, I plan to fully release all the design files.
In the meantime, I'm more than happy to answer any questions or take any feedback.
I'll be updating this post as more progress is made on the project. Things will move a bit slow as I'm funding this 100% out of pocket and it is certainly not cheap.

Thanks! - superfoil

User avatar
Lalaland124

23 Jul 2021, 21:27

Man that's awesome!! Really impressed with the outcome so far. Do you think the sound will differ from the original due to the partly different materials you used?

Looking forward to updates :D

User avatar
superfoil

23 Jul 2021, 21:52

Lalaland124 wrote:
23 Jul 2021, 21:27
Man that's awesome!! Really impressed with the outcome so far. Do you think the sound will differ from the original due to the partly different materials you used?

Looking forward to updates :D
Thanks, I'm really pleased to hear that!

The sound is definitely one of my main concerns, so I've been careful so far as to try and preserve the original sound.

I've yet to be able to test the sound of the internal spring components, but I'm not anticipating that much change. It will likely sound a bit different (if I had to guess, perhaps a bit higher pitched?), but it shouldn't be too obvious.

I have extensively tested the affects the new key module has to sound, and it's virtually indistinguishable. It can at times sound slightly higher pitched, but the difference is so slight I'm confident you wouldn't be able to tell unless you were clacking an original and repro right next to each other with your ears pressed up against them. When all's said and done, and you have the switches mounted in a keyboard, I don't think you'd be able to tell at all. I have a few of the new key modules in my daily driver 5251 and I haven't noticed a thing (that being said, the solenoid tends to cover up most sounds).

When I have the full assembly all produced and put together I'll definitely try and post some recordings of the sound (and measurements of the feedback), but at this stage it's difficult to tell the full impact.

TL;DR: Likely expect some sound variation, leaning towards a higher pitched sound, but the anticipated difference is very minimal.

User avatar
dcopellino

23 Jul 2021, 22:01

Wow, my congrats with you for the awesome project that seems paving the way to www.modelbkeyboards.com for a full beamspring keyboard replica market place. It'd be worth of a venture capitalist / startup initiative.

User avatar
andreas

23 Jul 2021, 22:05

Thank you for sharing this project, such posts are very inspiring.
Your outcome looks remarkable and finely tuned.

All the best for the finish line.

User avatar
Weezer

23 Jul 2021, 22:10

I'd love to try one of these out. Are you open to mailing me one?

orihalcon

23 Jul 2021, 22:19

Beautiful project! If I had the design skills that you do, I'd likely have tried something similar myself.

I would imagine that designing a custom Beamspring PCB/custom barrel plate would be child's play compared to switch design? I've had some ideas for custom layouts using original (or these reproduction) switches.

Only other thing I could think of that may be worth doing is having a stem version that accommodates MX mount caps.

Also, how are you holding the beam spring to the stem at the bottom? Small screw perhaps? I get the preservation of the "square" cutout on the beam, but I almost think a more narrow rectangle that runs parallel to the beam is all that would be needed for alignment purposes and wouldn't have as much of a weak spot for the beam to break in the middle as it wouldn't have as thin of metal around the square.

User avatar
darkcruix

24 Jul 2021, 00:09

orihalcon wrote:
23 Jul 2021, 22:19
Beautiful project! If I had the design skills that you do, I'd likely have tried something similar myself.

I would imagine that designing a custom Beamspring PCB/custom barrel plate would be child's play compared to switch design? I've had some ideas for custom layouts using original (or these reproduction) switches.

Only other thing I could think of that may be worth doing is having a stem version that accommodates MX mount caps.

Also, how are you holding the beam spring to the stem at the bottom? Small screw perhaps? I get the preservation of the "square" cutout on the beam, but I almost think a more narrow rectangle that runs parallel to the beam is all that would be needed for alignment purposes and wouldn't have as much of a weak spot for the beam to break in the middle as it wouldn't have as thin of metal around the square.
Let me dream of a SSK and modern full size version with MX mount stems and beautiful thick double shot white keys. The Space-bar row with 7 keys (similar to the 4704) ... combined with a wrist rest like on the 3278 ....

User avatar
Kabong30

24 Jul 2021, 03:42

I'm into it. Any idea what the final product might cost? Very interested in the progress here.

inozenz

24 Jul 2021, 19:15

Would it be possible to lower the height of these switches? Other then that amazing work! Keep it up.

0x00

24 Jul 2021, 22:26

Woah, you are serious about this.
I always figured the slider would be the hardest part; how do you intend to embed the beam spring in it?

User avatar
superfoil

26 Jul 2021, 19:53

Thanks for all the kind comments! I'm happy to see other people are just as enthused about this project as I am.

Let me address some of the questions you all have posted:

On stem design:
orihalcon wrote:
23 Jul 2021, 22:19
Also, how are you holding the beam spring to the stem at the bottom? Small screw perhaps? I get the preservation of the "square" cutout on the beam, but I almost think a more narrow rectangle that runs parallel to the beam is all that would be needed for alignment purposes and wouldn't have as much of a weak spot for the beam to break in the middle as it wouldn't have as thin of metal around the square.
0x00 wrote:
24 Jul 2021, 22:26
I always figured the slider would be the hardest part; how do you intend to embed the beam spring in it?
In the original switch, the beam spring is insert molded onto the key column. Meaning, the spring is placed in the mold, and the molten resin flows through the square cutout and forms a sort of "cap" holding the spring onto the column. There's no denying that the large cutout produces some structural weakness in the spring, but I imagine the large cutout is necessary to ensure a strong flow of material through the spring to form the cap. It can be very difficult to get molten resin to properly flow in small gaps.

I really like the idea of using a small screw! Due to the logistics and tooling needed for the insert molding, the key column and beam spring assembly will likely end up being the most expensive part. However, using a screw and possibly a small washer would eliminate the need for such tooling. Using standard parts such as M3 screws could also reduce costs as you wouldn't need to produce your own screws. You could also reduce the size of the gap in the middle of the beam spring and increase its integrity as previously suggested. A side benefit would be the ability to replace the beam spring if damage occurred to it as well. The only initial downside I can imagine would be that this wouldn't be a true reproduction if I went that route, but I think the pros far outweigh the cons in this instance. I will immediately start exploring using a screw to fasten the beam spring to the column.

On MX style mounts:
orihalcon wrote:
23 Jul 2021, 22:19
Only other thing I could think of that may be worth doing is having a stem version that accommodates MX mount caps.
I would certainly be interested in making a key stem that supports MX caps! In terms of design I imagine it would be fairly simple, but there will likely be some pain in manufacturing. Since the current stem is simple and flat, it can be cut from a single sheet of material. The MX mount on the other hand would be 3D and couldn't be cut from a single sheet. It would likely have to be casted into a metal mold, which can be very expensive to set up the tooling for. That being said, once it's all set up it shouldn't cost much more to produce than the original stems. I'll spend some time this week 3D printing some prototype MX stems to seem how it would all go together.

On switch height:
inozenz wrote:
24 Jul 2021, 19:15
Would it be possible to lower the height of these switches?
Theoretically I don't see any issue in reducing the height. You would need to spend a lot of time trying to get the right balance of feedback when shortening the height if you wanted a drastic reduction, though. There is some dead space in the key column and key module that could be removed for maybe a ~15mm reduction in height without needing to modify the components responsible for feedback. This is just an estimate based on first glance, though. There are likely some other issues I haven't considered.

The real issue in reducing height is how you're going to use the reduced switch. I'm not aware of many custom beam spring PCBs and cases where you could use such a switch. To my knowledge, you're currently really limited to using the original IBM beam spring units, where the height is necessary due to the case design. That being said, imagining a world where you could use a traditional beam spring without the complimentary carpal tunnel is certainly tantalizing. I wouldn't be opposed to exploring this idea down the line, but I'd like to get the original design down first (if anything, for the sake of familiarizing myself further with the switch).

On future plans:
dcopellino wrote:
23 Jul 2021, 22:01
Wow, my congrats with you for the awesome project that seems paving the way to www.modelbkeyboards.com for a full beamspring keyboard replica market place.
orihalcon wrote:
23 Jul 2021, 22:19
I would imagine that designing a custom Beamspring PCB/custom barrel plate would be child's play compared to switch design? I've had some ideas for custom layouts using original (or these reproduction) switches.
I'd be lying if I hadn't considered something along this train of thought. I figured the switches would be the most difficult part in a reproduction effort, and that if the reproduction switches were solid then the rest would follow. I do have some electrical engineer friends of mine looking over the original PCB and controller exploring the possibility of a modern version. I'm not too well versed in PCB design though, so I can't speak too much to this. However, from what I hear from them it wouldn't be too much effort. It will 100% be where my attention goes once the switches hit a point I'm comfortable with.

The rest of the assembly (barrel plate, case, etc.) would be very simple in design. The issue with these parts is the cost of prototype and production. As you can imagine, doing a test one shot metal casting of 11 pounds of solid steel isn't the cheapest thing in the world. The tooling cost for an injection molded barrel plate would also be pretty high considering the large size. For reference, the tooling for the small key module was just under $2,000. I would absolutely love to design and make these parts though, despite the cost. However, these steps would have to come later as I'm doing this all out of pocket at the moment and I don't want to bankrupt myself for a metal shell with nothing to put in it.

On cost:
Kabong30 wrote:
24 Jul 2021, 03:42
Any idea what the final product might cost?
I can't accurately speak to the parts that haven't been quoted for manufacturing yet, but I can try to paint a clearer picture. If I had to hazard an estimate based on the low quantity orders I've made so far, I'd say: $12 / switch. There's a lot of speculation in this price, though. It's also based off the precedent that one were to have manufactured only the exact number of switches needed for a single keyboard. Bulk orders would certainly drop this price; to what degree, I can't say. I can imagine it would be significant for the plastic parts, though.

On community feedback:
Weezer wrote:
23 Jul 2021, 22:10
I'd love to try one of these out. Are you open to mailing me one?
I wouldn't be opposed to sending out samples to the community to test and validate my designs. I don't have a huge quantity on hand, though. The bigger issue is that the only parts that I could send out currently are the compression springs and key modules. If people still want to sample these then feel free to PM me or something, but you won't get quite the full picture doing things piecewise.

I hope that's answered most of your questions! If I've missed something or people come up with something else please let me know. Thanks again! - superfoil

User avatar
shine

26 Jul 2021, 23:17

very interesting, i'm in for whatever you need. do you plan on making keyboards? i would love to have an unsaver beamspring, or a kishaver beamspring, or a ton of beamships :)

Johnbo

27 Jul 2021, 00:19

This is super cool, I've been hoping for a project like this to come along. I'll definitely be following your progress!

User avatar
DMA

02 Aug 2021, 12:09

Controller is solved in at least two different ways already (one by xwhatsit, reportedly improved to "good enough" by pandrew, one by yours truly), don't waste your friends' time on that.

PCB is trivial - no magic there whatsoever. Just divide the area under flipper into two equal-area pieces, stitch matrix from those, you're done. Also already done and verified working: beamspring produces a strong, clear, consistent signal.
The only thing is to not forget to provide enough conductive plane under the PCB - but since the subassembly holding keys is usually metal it solves itself.
wcass and emdude have experience in making those.

The main problem of the beamspring is debris ingress - various shit falls inside the switch and interferes with normal operation. That's why all these saran wraps covering the top plate. If you can solve that problem - you'll be a hero.

I would probably widen the beam around the center - I don't trust the thickness around the hole, and that part has minimal impact on anything, as it's mostly stationary. Another variant would be to hold the beam by a slot in the slider - things _should_ be mostly self-centering anyway, may be you don't even need additional centering. That slot tho is an additional (likely removable) mold element.

TBH, pay no attention to sound - the beamspring's "thing" is the kickback as plate rams into the slider. It is not visible in HaaTa's charts, but it's definitely there - just try slowly pressing the key and it will eventually drop faster than your finger and immediately return with vengeance (and then oscillate a bit).

MX stems would be nice - probably as just a cruciform column molded with the slider - but the problem is enormously deep travel of the beamspring. Stabilizers would need to endure what, 4mm of travel compared to standard 2.5? Of course there can be stabilizer modules, but still, the travel is larger than everything you've seen before.

Re: making them shorter - those 15mm are not dead space, they keep stem parallel to it's tunnel. Without it, there's a likelyhood the switch will be _extremely_ wobbly. Could be solved, I'm sure - but don't expect it to be easy.

But your main enemy is still dust ingress. Current beamspring is more maintenance-expensive than a rifle. Not a sustailable solution.
"The metal shell" is just a small metal plate with 9 holes in it, another one - solid with bolt holes for standoffs. Cheap.

PS: Kudos for dropping money for IM tooling - 3d-printed versions would totally suffice at this stage.

User avatar
DMA

02 Aug 2021, 12:15

..and please try to make them cheap. There will be larger order quantities - may be 10k switches (I don't really believe in 100k - even at $5/switch it's still $500 per keyboard just in switches, and this is just crazy.) - which should help with molded parts as material is seriously cheap, but not with everything.

So. No fancy materials, cheaper manufacturing process, etc etc (like, no "bending the PCBs" shit - it's completely useless and requires a custom jig, aggravated by the large processing time, which should make it _really_ expensive op.).
If something turns out to be too cheap - it will likely snap in one place in a single component, so a) repairable. and b) you can beef up _that_ component alone for the patched version, and call it a day.

pandrew

03 Aug 2021, 00:47

Amazing work!

If you get to mass-producing full switches I'll be in for about a keyboard's worth, say 120 to make a custom keyboard. I would prefer higher long-term reliability, rather then completely reproducing the design just for the sake of authenticity. If there's any risk that making the beamspring the exact shape IBM had would be less reliable (maybe due to different factory process, or due to different material), and if having a smaller hole and screw would make it more reliable, then I'd prefer the screw method. I think I'd be more interested in the MX-compatible caps solution.

Regarding the debris problem that DMA mentioned, I have a few ideas:
  • There are two holes on top of the case, which I see you replicated. I wonder what its purpose is, and how important it is. We could experiment with plugging them, or changing the shape to a finer grille-something, or changing its position, cause I suspect that could be one major place of entry. The contamination shield is never gonna protect you from the decomposing mat/foam falling through those holes. Correct me if I'm wrong, but I'm pretty sure the mat/foam, is under the contamination shield, and is right up to and touching the top surface of the key module, so touching those holes too.
  • I have a suspicion that not all problems that we blame on dust/debris is caused by dust/debris. Sometimes the signal level change is just too high, and it's right after someone swears they cleaned the keyboard. I think in some cases the flyplate could be getting stuck somewhere, either on the side of the key module case which I think is less likely, or maybe more likely, the flyplate is not completely parallel with the PCB, and when the flyplate's sharp corner touches the pcb maybe it forces it to stay half-up-in-the-air because of slight misalignments. If it sticks once, it's more likely to stick again, because the corner digs a little bit into the solder mask. I don't have any proof of this theory yet. I would love if we could just somehow look into a switch when it gets into a stuck state, without opening the whole assembly, and releasing the pressure. Or try to reproduce it with a transparent switch module. Unfortunately my 5251 doesn't show this problem at all. But if I'm right about this failure mechanism, we could mitigate this by:
    • Smoother solder-mask on the PCB
    • We should check if the shape of the sense pads could potentially contribute to the flyplate catching, since the pcb surface is taller where there's copper underneath.
    • We could make the flyplate edges/corners beveled/rounded so it's less likely to catch. Original ones are just too sharp in my opinion.
    • We could apply some kind of super thin lube on the pcb. Maybe some material exists that isn't completely crazy to do this with. Do VCIs(Vapor Corrosion Inhibitors) also lubricate, maybe?
On a slightly different note, are you aware of SneakyRobb's work on this topic? Which I also find amazing, see here: viewtopic.php?t=21851

User avatar
superfoil

03 Aug 2021, 07:16

Thanks for all of the feedback! Lots of things to address here, hopefully I got to everything in this reply:

On controller design:
DMA wrote:
02 Aug 2021, 12:09
Controller is solved in at least two different ways already.
Fair enough. No work has started on the controller specifically yet anyway, so going with the existing designs is definitely the correct move.

On beam spring width:
So, this last week I experimented with using a screw to fasten the spring to the key column. I no longer think this is a valid solution, for a couple of reasons; Primarily, molding threading is a very difficult task even with large threads. The precision needed to mold the threading for the size of screw the beam spring needs simply isn't feasible (this is often why in laptops you'll see the over molded metal nuts in the plastic shells of the assembly). I really didn't want to custom tool a screw, so I looked at existing screws to use for this purpose. I settled with a wafer head m2.3 screw as it's readily available, and super small and low profile. The issue is, the screw barely even fits through the already large hole in the middle of the beam spring, so reducing the size of the hole wouldn't be possible anyway.

With this in mind, I will be exploring other solutions. I think DMA's suggestion of using a removable mold element to slot the spring in is very good. I will be experimenting with a friction fit design for a removable "pin" that secures the spring to the key column.

On dust ingress:
pandrew wrote:
03 Aug 2021, 00:47
There are two holes on top of the case, which I see you replicated. I wonder what its purpose is, and how important it is. We could experiment with plugging them, or changing the shape to a finer grille-something, or changing its position, cause I suspect that could be one major place of entry.
Mechanically, I'm sure these holes don't have any purpose. The only thing I think they may be for is ejection of the part from the mold. Their alignment with the key module's overhanging clips makes me wonder if the negative space is necessary to eject the key module without having it get caught in the mold. I definitely agree that this is the major ingress point for dust in the switches. While dust can certainly enter via the key column's hole, the spring and O ring make it a bit more difficult than the two open holes on the module. I'll experiment with some 3D prints of the key module modified to close these holes off. I'll also talk with the injection mold factory regarding the revised design to confirm or deny my theory on their use in terms of ejection. I think finding a way to close these off will certainly assist in the mitigation of dust ingress.
I would love if we could just somehow look into a switch when it gets into a stuck state, without opening the whole assembly, and releasing the pressure. Or try to reproduce it with a transparent switch module.
I think your theory could very well be true. I have experienced what I believed to be "sticky" fly plates causing issues in operation. I actually have on hand tons of clear resin for 3D printing. I'll print off a couple clear key modules that one could slot in existing switch components into to test this theory. If people are interested in testing this as well shoot me a PM and I can send the clear key modules your way (I just need an excuse to use this clear resin). I'd like to see exactly what causes this issue before exploring possible fixes (but I think rounding the fly plate isn't a bad idea, as it really seems to like to catch. My money is on this being a major contributor to the issue).

On price:
DMA wrote:
02 Aug 2021, 12:15
..and please try to make them cheap.
It really is far too earlier for me to accurately guess what the final cost will be, there are just so many variables to account for at this stage. I only have low order sample quantities to estimate off of, and all tooling so far has been done for low batch runs. If for instance I were to need to manufacture 10k switches, the machinery would need to be re-tooled for the high quantity (which would result in cheaper per piece cost), and there would of course be bulk discounts on material as well. I can try and probe suppliers for specifics on these numbers, but at the end of the day it's all conjecture until every part is finalized and tooled, and we have an accurate production quantity estimate. I'm certainly not trying to use any fancy materials where not required, but at least for the springs I am going to continue to use stainless steel as I have seen first hand how rust can destroy these components.
Last edited by superfoil on 03 Aug 2021, 07:23, edited 2 times in total.

User avatar
superfoil

03 Aug 2021, 07:17

On a slightly different note, are you aware of SneakyRobb's work on this topic?
I actually haven't seen this, I'll have to take some time to read it all through. Thanks for the link!

User avatar
DMA

03 Aug 2021, 07:51

There are people on this forum with injection molding experience. But my current rough understanding is "mold is good for probably 100k to 1M impressions (depending heavily on plastic, mold material, temperature and pressure used), ABS is like $3/kg, everything else in total is on the order of plastic costs."

I would be careful with friction-fitting metal into a slot in plastic - that would create a permanent tension point, which can later lead to cracking. In my experience, plastics are generally better on withstanding compression than tension.
Not a medical advice though :D

I had an impression that metal inserts are in plastic chiefly to distribute the load to a larger contact area. There are lot of other options though - thread-forming screws, retaining rings.. they should all be matched to the plastic you're going to use though.

Ellipse

07 Aug 2021, 07:37

Based on the posting of this great project I have publicly posted the Brand New Beam Spring Keyboards reproduction project that I've been working on for a few years now and sitting patiently until the Model F project is finished. viewtopic.php?f=50&t=26174&p=491823#p491823

I think there is a demand for replacing some of the old IBM beam spring modules on the existing keyboards, especially now that these keyboards have often sold for more than $1,000 each in recent months. The thin metal pieces in the module are easy to break - I've had a few broken in my own two beam springs and had to get extras from orihalcon in years past.

As noted there, I will reiterate that my new beam springs did not delay the new Model F project and that they will not ship and are not a focus until I've sent out the new Model F Keyboards. Additional details are in the above link.

My project is not in direct competition to your project as my modules are not backwards compatible with the original IBM beam spring keyboards. They have been redesigned to be used with Cherry MX type keys natively and to remove all of the wasted space in the modules and in the beam spring cases, while maintaining 100% original key travel and the same exact specifications for the fly plates, beam module footprint, etc.

Also you noted a production cost of about $12 per module, possibly less with more quantities. My new Beam Spring keyboards will be selling for about $4 a module (obviously plus the cost of the case, PCB, etc.). A keyboard with a cost of $12 per module would have to sell for more than $1,000.

I will be following this project with interest and hope that you will be able to continue pursuing it. I would be happy to discuss offering my capacitive beam flippers, metal fly plates, and the like if you so need them as I've made some extras and have all of the tooling already made. The most important part you can make is the reproduction white inserts in the modules, with the new metal part attached as those are the most fragile in my experience.
2021-08-06_23-38-40.jpg
2021-08-06_23-38-40.jpg (677.92 KiB) Viewed 835 times
2021-08-06_23-42-10.jpg
2021-08-06_23-42-10.jpg (731.74 KiB) Viewed 835 times

User avatar
DMA

09 Aug 2021, 05:45

$4 per mass-produced switch. Wow, that's a whole new level of grift - beating even the now-legendary $4000 NIB SSK.

Post Reply

Return to “Workshop”