deskthority

Well, I don't know how to start to talk about this, but I'll do my best.
Three days ago I got the idea of design my own switch, considering what I think may result in an S-tier switch. I came up with this thing, the model was ended today:


PFH magnetic collar switch


Naked view

As you can see the switch is a bit too complex and somehow over engineered. I'll try to explain part by part how this thing operates, clicks and all the important stuff.
In summary, what is this thing? This is an SPST reed switch switch operated by a magnetic clicker, actuating at 2mm with an overrun of 1.5mm.
Now that we have polished the general concept, we go by pieces:

Housing upper part: Maintains the switch closed using screws on the top. Not much to say about this part.
Housing middle part: Similar to the upper part, not much to say. It has a piece of plastic in the middle to retain the orange piece called ferromagnetic plate and also prevents the slider from having a too large overtravel. Because of this plastic neck the housing is divided in three parts.
Slider: Standard MX cross mount with a flat H shapped base in order to have a very high stability and avoid wobble (this shape may cause some friction problems). The spring under it is designed to be screwed/moulded inside the plastic piece, but this may not be necessary. It has an extension inside the spring to push the magnetic plate after the 2mm pretravel.
Ferromagnetic plate: A piece of metal used to attract the magnetic plate, goes glued under the plastic collar of the middle part of the housing.
Magnetic plate: Two magnets moulded (or glued) in a plastic piece. The one with a ring shape is used as a clicker, resisting being separated from the ferromagnetic plate. The flat one is used to activate the reed switch. It has a sort of engraving to hold the spring in place.
Housing lower part: Holds in place the reed switch and has two plastic fingers to prevent the magnetic plate from hitting it. I still can't decide if the screws just "end" here or use them as a way to hold the switch to the PCB instead of a standard plate.

This thing isn't ended, obviously. Some measurements aren't the ideal ones and the concept is not even proven. I would like to be able to promise that one day they will see the physical version of this thing, but I don't think I can afford it, I just want to make a logistical contribution about how great the magnets can be. I will continue updating this thread if necessary.

very interesting that you come up with a mx version of a design that has already been done by Siemens in the 1970s viewtopic.php?f=62&t=9239 I bought the white capped version of the boared posted at the bottom of Haata's thread, it is a pleasure to type on.

Anakey wrote: 15 Jul 2019, 09:26 very interesting that you come up with a mx version of a design that has already been done by Siemens in the 1970s viewtopic.php?f=62&t=9239 I bought the white capped version of the boared posted at the bottom of Haata's thread, it is a pleasure to type on.

I didn't understand what was that switch until now, thanks! I guess it's called "the german beamspring" for some good reason.

it is as good as a beamspring, though it is too rare to be well known enough to have a value attached to it. Other then mine and Haata's there was only one other example using the style of switches that i am aware of.

That is really cool. I hope to see further updates. Good job!

I'm working on a new version of the switch, which will be mounted on the PCB like something between Keytronic magnetic reed and Fujitsu leafspring. I'm having some problems, first one is find the proper magnets to work with, I think the neodymium ones will be to much strong. Also I don't know in what kind of spring work. Something similar in dimensions to the ones in the beamspring switches will be fine.
I'm also looking for a cheap prototyping method. Any ideas will be well received.

3d printed plastic pieces can be quite rough, but for rough prototypes they can be very useful. You can print in abs with finer resolutions to smooth them out and use the acetone/filler/sanding methods to get them less rough. For longer term use I would use the printer to make the model and then smooth it out with acetone/fillers etc. Then you can use the casting resin to make nice copies for further testing.

I have no idea about magnet sourcing. You might have to find a reasonably priced and sized magnet first and then size the rest of the switch around that. You may get lucky and find the perfect magnet, or you may have to compromise.

For beamsprings I have made some semi-okayish ones by hand. I still need to find a better way to make them.

For return springs I have on order some of these. (not sure they will work perfectly tbh) https://us.misumi-ec.com/vona2/detail/ ... de=UY13-15

They were kind of expensive to ship. I would recommend the thread called "low profile beamspring" where he uses return springs from existing modern keyswitches. This is probably a better long term strategy.

Hi. Sorry for the slow updating. The resin idea is quite good for test switches (the idea is use POM to achieve a good smoothness). I also found a standard type of magnet. When I was talking about beamspring springs it was because of the compressión springs they have, but this part is not too critical because custom springs are not something ultra expensive.
Anyway, you gave me the idea of use the same stencil material from your homemade beamspring switches as a ferromagnetic attractor.

After some problems with my hard drive, I managed to design two variants of the same switch:



Sorry for the informal update and the lack of a spring, I'm a bit burned about this project right now. The green one is my "all purpose switch", with no hysteresis and a tactile event similar to the one in the Cherry MX green. The blue one is more like a beamspring. I'll explain it better late.

I think you need to see this: http://lcamtuf.coredump.cx/gcnc/ch1/

spongebob1981 wrote: 14 Aug 2019, 14:36 I think you need to see this: http://lcamtuf.coredump.cx/gcnc/ch1/

Thanks dude. I was checking that guide in general terms but I intend something more rudimentary.

Another update:

Here is an assembly of the modules mounted on the PCB. I didn't bothered in the reed switches, I may end replacing them with a more sofisticated hall effect sensing. Here you can appreciate the lack of a metal plate, which will result in a cost reduction if I ever build the keyboard. The integrity of the whole assembly will be sustained by two horizontal metal bars under the PCB.


I'm also working on a keycap profile for it. I don't know anything about the ergonomic theory behind the keycaps, so correct me if this thing is a pile of shit. Instead of an orthodox format I decided to make something like "Siemens keycaps on steroids" with that kind of dots instead of a full spherical surface. Here they are:

I also noticed some problems in the blue switch:

A lot of time passed, but here is the MAGSEP 3.0 in all his glory;


I 3D printed the last version, but had several problems like binding (beyond the quality of 3D prints), wrong measurements, very high profile, etc. This version has a better stabilization, lower profile, a nice travel, standard MX springs and is capacitive. I also changed the moving disk magnet for a fixed ring magnet and a ferromagnetic piece in the moving part. If someone wants the files to experiment with just tell me.
I'm working in this project with Abrahamstechnology as I help him with his blue Alps replica, so, thanks mate.