20181214T22000Z: Learn how to make Model F / Beamspring PCBs

[__red__]

I'll be doing a Google Hangout, Yourtube live-stream or something as discussed in other threads to teach people how to design & fab their own Model F / Beamspring PCBs.

Timewise that's:

23:00 CEST
22:00 UTC
17:00 EST
16:00 CST
15:00 MST
14:00 PST

If that time doesn't work out for some post down some better times and I'll do a re-do.

Thanks,



Red

[hansichen]

Sounds interesting. I hope you'll record the vod so that people can always catch up and have a source to get some experience in pcb design.

[John Doe]

interesting

[andrea-i]

Please __red__ record this!!! my newborn baby might as well decide I am not supposed to learn pcb design but take her for a stroll instead : D

[__red__]

Please __red__ record this!!! my newborn baby might as well decide I am not supposed to learn pcb design but take her for a stroll instead : D

Hey Congratulations! Is this your first babe?

[andrea-i]

Please __red__ record this!!! my newborn baby might as well decide I am not supposed to learn pcb design but take her for a stroll instead : D

Hey Congratulations! Is this your first babe?

First one <3
Looking forward to see the recording if you'll upload it later on!

[DiodeHead]

yes please, post your youtube channel and record it T_T so we can watch it anytime.

[snacksthecat]

I’ll be tuning in for sure!

[JP!]

I would be interested also. Have you decided on a platform for streaming?

[tron]

I'm interested. Please record it so we can play it back later.

[JP!]

I'm interested. Please record it so we can play it back later.

Yes and also maybe add it to the wiki

[__red__]

It will be on youtube.
I will start streaming 5 minutes early.

Feel free to join the chat and ask questions.

[__red__]

The channel -=- Look for the "live"... once it's live I'll post the specific video.

https://www.youtube.com/channel/UC3kkrn ... VW-n5emaXQ

[__red__]

Live (ish)

[listofoptions]

Thanks for the vid! Ill probably try and use your diptrace stuff for KiCad (as that's my preferred system) in a branch of your github. definitely inspiring!

[DiodeHead]

thanks for sharing

I have a question, could the outline that you make in the second layer just be a copper pour?? wouldn't that be much less work?? or is made this way because a copper pour would interfere with the capacitance of the switches.

I liked a lot the 0.6 mm tip on PCB material so it's bendy, that is something that I would like to test for sure, please if you have any project with a curbed PCB post a link.

Also would love to see more about how you cast the model F pieces that you show at the beginning of the video.

thanks.

[__red__]

I have a question, could the outline that you make in the second layer just be a copper pour?? wouldn't that be much less work?? or is made this way because a copper pour would interfere with the capacitance of the switches.

That's an excellent question - I don't honestly know. Like I mentioned in the video, the history of the making of these PCBs really has been a history of cargo-cult. If you actually look all the way back to the really, REALLY old beamsprings (like the beamdesk module for example) you see that IBM used a different pad design again.

I liked a lot the 0.6 mm tip on PCB material so it's bendy, that is something that I would like to test for sure, please if you have any project with a curbed PCB post a link.

I don't - but I have made "PCBs" with a vinyl cutter, acetate sheet, and copper tape and the capacitive swing is insane.

Remember, the thinner the board, the closer the pads are which alters the value of the second capacitor.

In all seriousness - it might be a good idea to do a series of capacitance test boards.

Say we fabricate 10-15 different designs to test all of these different variables and try and really get a handle on the answers to all of these questions.

Hypothetical: If I were to do say 20 different board designs to test things like board thickness, ground pours, single vs double-sided pads, etc... and committed to sending 1 of each design to our community's main F developers - I wonder if people would either chip in a few $$ to help offset the costs or just want to buy a set to experiment with?

Also would love to see more about how you cast the model F pieces that you show at the beginning of the video.

All my gear for doing that is in storage so unlikely to re-appear in a timely manner. Nothing special at all about what I was doing - just make sure you invest in both a vacuum chamber and a pressure pot if you go that route.

[matt3o]

I would be totally interested in sponsoring such a project... honestly what scares me is not actually the PCB but everything else around it...

[DiodeHead]

no, the bending stuff I meant for a normal switch PCB and test curvature to see if more comfortable or not, like maltron keyboards but I think they are much more complicated than a simple curvature.

is that capacitive design the same of touch-capacitive??
http://www.ti.com/lit/an/slaa576a/slaa576a.pdf

I guess we need also a deep guide from DMA on his knowledge making common sense controller

[__red__]

no, the bending stuff I meant for a normal switch PCB and test curvature to see if more comfortable or not, like maltron keyboards but I think they are much more complicated than a simple curvature.

For that much curvature you're going to want to try and use FPCs (which is even more expensive). Again, something we could test in theory.

is that capacitive design the same of touch-capacitive??
http://www.ti.com/lit/an/slaa576a/slaa576a.pdf

No, but I guess in principle it could be if you're willing to drop enough capsense pins. The design-docs you link don't seem to cover capsense matrix design (which is what the ModelF/Beamsprings are).

I guess we need also a deep guide from DMA on his knowledge making common sense controller

Paging @DMA

[__red__]

I would be totally interested in sponsoring such a project... honestly what scares me is not actually the PCB but everything else around it...

Which is why I suggested sending one of each to the usual crew.

If we keep the size/format of the boards consistent then we can just build a testing rig which we can insert and remove the test boards as required.

The way I envisioned it working would be in much the same way that we see the switch force curves published here.

@DMA's Common Sense controller provides actual values for each location as opposed to 'up/down'. In principle we should be able to then use this to publish the up/down/delta for each of the designs and hopefully learn which of things we're testing maximizes the delta which minimizing the interference from other keys.

[DiodeHead]

I don't - but I have made "PCBs" with a vinyl cutter, acetate sheet, and copper tape and the capacitive swing is insane.
Remember, the thinner the board, the closer the pads are which alters the value of the second capacitor.

So from everything said, I understand that this works more or less like tuning air capacitors in old radios, but here you want fixed values instead of variable ones. Then the design rules would be?? :

-- you want your PCB to have the less movement possible so capacitance values won't change.
-- you want the same amount of copper surface per row and column, that way parasitic capacitance should be equal if the amount of surface is equal??, a tool like the differential pairs in kicad would help with that, what I'm not sure is if you can configure it to work only in thickness and don't make the wavy lines, since wavy lines would mean very small capacitors wouldn't it?

[__red__]

-- you want your PCB to have the less movement possible so capacitance values won't change.

Yup - and typically the PCB is smooshed between two metal plates to also protect against wandering hands etc which of course would change the impedience seen at the sample frequency (since we just went into the radio domain) hihi

-- you want the same amount of copper surface per row and column, that way parasitic capacitance should be equal if the amount of surface is equal??, a tool like the differential pairs in kicad would help with that, what I'm not sure is if you can configure it to work only in thickness and don't make the wavy lines, since wavy lines would mean very small capacitors wouldn't it?

Depends which axis you're talking about:

Code: Select all

                   A
        +--------+  +--------+
        |        | B|        |
--------+--------+--+--------+--------
                         C
--------------------+--------+--------
                    |        |--------
                    +--------+

A is determined by the foot.
B is determined by the amount of Cu they used.
C is determined by the thickness of the PCB.

B&C should be constant for any PCB - BUT, unless you spend the spendy on impedance controlled PCB processing they're not going to be constant over multiple batches.

Also - these are not capacitive relative to ground (and if they were we wouldn't care as long as they were constant); they're capacitive relative to neighboring pads that's our issue.

It's a matrix. If we wiggle row 1 and sense on row 5 - there will be *some* change seen if someone is pressing row 0, column 5 for two reasons:

0. The signal line for row 0 and row 1 run next to each other for a long way.
1. We strobed row 0 before row 1. If our duty cycle is fast enough then it's still possible that the pads on row 0 are holding sufficient charge to cause us a headache.

Take some time and find the theory threads on this forum - especially the one written by DMA on how and why he decided to sample the way he does. It has oscilloscope traces you can see this kind of behavior in.

[wcass]

In all seriousness - it might be a good idea to do a series of capacitance test boards.

Say we fabricate 10-15 different designs to test all of these different variables and try and really get a handle on the answers to all of these questions.

I had a few "pad design test" boards made a while back. DMA did the testing on them as i don't have the equipment to do that. Let me see if i can track down the thread ....

Fond it!
workshop-f7/alternative-capsense-patter ... 19166.html

But, all of those used 2 layers on 0.8 mm PCB. My newest custom uses a 4 layer PCB. The advantage of 4 layers is that the top two layers are super close (big capacitance change) and you can use 3rd layer as shield which allows you to route on layer 4 ignoring everything above on layers 1 and 2.