Teardown - IBM beamsprings

[Chyros]

Another instalment in the Teardown series, today we look at what's arguably the best keyboard switch in all of history. Hope you enjoy the video!

[__red__]

Nice video as always however, I'm not sure your description as to how the sensing works is accurate.

Arguably, that only matters to those of us building replacement PCBs or capsense circuits.

[vometia]

Very interesting video, and that's speaking as someone who very rarely manages to sit through a video. A bit of a revelation to me too, given the beam spring's reputation, I'd imagined the switch to be hideously complicated and made of the finest jewels: curious to see it's only a moderately complex switch and for that reason I'm wondering why nobody else has made them considering it's a sort of holy grail. I mean assuming it's not vintage patents still standing in the way.

I guess "big and tall" is a reason when everybody likes low-profile keyboards, it seems, but I could probably live with e.g. my Model M being half an inch or so taller. There must be a market for these. With nice, spherical caps with centred legends. Come on, keyboard manufacturers, somebody must want to resurrect the legendary beam springs.

[Chyros]

Nice video as always however, I'm not sure your description as to how the sensing works is accurate.

Arguably, that only matters to those of us building replacement PCBs or capsense circuits.

What do you think is inaccurate about it?

I delve more deeply into it in my F122 video, btw.

[Scarpia]

Thanks for this video, it was really interesting and easy to follow (sometimes when I watch more complicated switch teardowns I walk away still not understanding how the heck it all fit together).

[TuxKey]

having worked for years @ ibm ...i associate this switch with work ..for years that was ok and joyful....these days i don't care for it hahaha don't remember typing on one to be honest.. usually i was standing in a loud computer room .. thinking back i think it was ok a bit loud don't know what all the fuss is about to be honnest.. perhaps it's bin to long for me..
ahh i'm happy with my topre board.. only switchs on my wish list are the hako's and the new zilent switches from zeal..
perhaps i'l find one of these boards in at the next dutch meetup..

[Mr.JAM]

@chyros
https://youtu.be/p_JTZo2rKmw?t=7m31s
Your explanation is not wrong, but I think it would be better to explain about the charge and drain of capacitor with this way. http://downloads.cornall.co/ibm-capsens ... x1-90003.1 it's not have to be AC, and it could be done with PULSE DC. So, actually it should be the voltage difference sensing, not capsensing.

[__red__]

What do you think is inaccurate about it?
I delve more deeply into it in my F122 video, btw.

Looks like my phone ate my last reply.

Basically, you said that it worked by comparing the change in capacitance against a third pad. As far as I understand it, that's not quite how it works. I'm prepared to be wrong - please tell me if I'm mistaken!


The earliest boards just had two pads on the top and a ground-plane underneath them. The plate would change the capacitance between those two pads which was then detected.

When we move to the three pad model, there are two capacitors in series:

Incoming bad ascii-arts:

Code: Select all

    |    
XXXXXXXX XXXXXXXX
PCBPCBPCBPCBPCBP
         XXXXXXXX
             |

The top two pads are the capacitor that can vary because of the landing of the plate.
The right-pad and pad below form another (static value) capacitor.

Code: Select all

    C1         C2
----||----+----||----

Since 1/C = 1/C1 + 1/C2 (Capacitors in Series) it seems to me that the net effect would be to reduce "swing", but also the amount of noise. If I have modeled the math correctly, even though the swing is lower - the S/N ration should be better.

But then again, I could be completely wrong

[__red__]

So, actually it should be the voltage difference sensing, not capsensing.

I think that's a distinction without a difference since that is how capsense typically works.
(Although strictly speaking, the term 'capsense' is a Cypress trademark as opposed to a technical method)