READY - $10 - DIY Through Hole Universal Model F and Beamspring Controller

[kralcifer]

kicad is still better than diptrace or FRAGGIN eagle. altium? well no mr moneybags its not, but at that price point literally nothing is good enough

I'm new to all the helpful software but I'm going to bet on Fusion 360's January addition of Electronics:

https://new.engineering.com/story/fusio ... cal-design

I've been pleasantly surprised with their parametric 3D modeling. One of their stated guiding philosophies is "to integrate as many tools as possible across a single workspace". I'd say I'm rooting for them because I want that. I want to 3D model my DEC keyboard's plastic assembly to have a rail that will slide in a row of Amoeba single switch. I'm hoping I can bring in mtl's schematics and layout. That article mentions a lot of great things about knowing the size of both and telling you when there's a problem. I'll let you know if any of it works out.

Not sure why it's FRAGGIN eagle to you but I'm hoping with the integration work they've fixed some of what you didn't like. And it's free to hobbyists.

[kmnov2017]

This project is now ready and a go. No more expensive xwhatsits on ebay. Just make your own for 10 euros.

GitHub will be updated shortly.

[tentator]

Hi ho capacitive community!
Great news on the horizon: this PCB works and is going to be supporting QMK!

Pandrew managed to do yet another miracle and after having ported QMK on the regular xwhatsit controller PCB, now managed to do the same also on this PCB!

So far we noticed just a few minor improements that could be done on this PCB for a next revision but it's fine to go also as is:
- resistor spacing is narrow for through hole components but I just used SMD ones and it worked as a charm
- a cap on the DAC output needs to stay unpopulated: just not solder the 10uF cap in position

For the rest I'm currently writing this post with my IBM 3278 Beamspring and the TH-xwhatsit controller and QMK firmware sporting autocalibration. It might still be we have to do minor finetunings and testing before releasing to QMK GIT (and the QMK guys actually being satisfied and accept the pull request). In the meantime we are welcoming other testers on other platforms and models.

photo5974081767235236865.jpg (379.04 KiB) Viewed 9462 times

Here you can see the smd resistors I've used to better fit the board, I like it:

photo5974171265763751590.jpg (158.16 KiB) Viewed 9462 times

This is the place where you shall not put the cap:

photo5974171265763751589.jpg (370.79 KiB) Viewed 9462 times

And here you see the resulting sandwitch from main board and beamsprings connector extension PCB:

photo5969709383153726043.jpg (270.09 KiB) Viewed 9462 times

photo5969709383153726042.jpg (284.44 KiB) Viewed 9462 times

The original 30pin beamsprings connector fits perfectly also on the extension:

photo5962967890816774914.jpg (169.92 KiB) Viewed 9462 times

I used standard header pins and header female receptors to interconnect the two parts.

So all in all: great stuff! -> Cheap and easy Model F and Beamsprings conversions ahead for everyone with the standard and convenience of QMK firmware. =)


Have fun with it!
tent:wq

[snacksthecat]

This is really great!

Did QMK support capacitive sensing already? Or is that what the pull request implements?

[tentator]

correct, it's the first time somebody manages to imlement capacitive sensing in QMK.. it's quite a breakthrough I'd say! Could be beneficial also for other capacitive boards! I was dreaming about this since a couple of years.. at least!

[Twst]

Really neat!
Good job everyone involved!

How is the pro micro holding up memory wise with all this capsensing and extra QMK stuff?

[tentator]

That's a good question and indeed one of the main problems for implementing qmk on those tiny chips.. It comes to the limit when calibrating the thresholds since we cannot keep in memory like one calibration point per key, but we have to make tradeoffs.. Especially for beamsprings this seems quite tight..
It's also one of the main reasons that layouts in qmk need to be compiled and flashed and are not reconfigurable on the fly /live..

Tent:wq

[kmnov2017]

I've updated the start of the thread with the Gerbers and BOM list. My BOM cost me just over 8 euros (with free ship to store) but not including the PCB (which costs about 1.5 euros per piece when ordered in multiples of 10 from JLCPCB).

[tentator]

You can remove from the BOM the item 8: C2 C_Disc_D3.0mm_W1.6mm_P2.50mm 1 10U

BTW, andrei did the whole firmware while not having the HW at hand while I was doing the testing "remotely", as-usual!

The most difficult part here was the new DAC and his protocol, which is different from the DAC onto the original xwhatsit BTW. And soldering mysterious blobs shorting pin 2 and 3..

Tent.

[Twst]

This is defently pushing the beamspring and model fs and other capacitive board longer into the 21st century for anyone interested.

Will build a couple of these for sure.

When you have it running on the pro micro, should it be possible to drop in a proton, teensy++ etc with better specs to give a little more breathing room for macros and stuff?
I heard the pro micro could brick if pushed over the edge - how close are you?

[tentator]

no I doubt you can brick it, but you could get to a not nice stack overflow maybe?
I don't know the proton and if that would be much better, but in the end we also wanted to use here something cheap and easy to source, which also is part of the "mission" statement, right?
But there are also secret plans to try to port qmk to another cheap SoC chip and see what happens there.. similar target as usual: around 10 bucks give or take, plus shipping.

[anthonymak]

neat, one suggestion, wish that the components are soldered to the other side, the through hole components can be seen.

[tentator]

Yes, that would also make sense so the sandwich could be done closer! Good suggestion!

[kmnov2017]

neat, one suggestion, wish that the components are soldered to the other side, the through hole components can be seen.

Its up to on what side you use the plugboard - either the top of the bottom of the main controller PCB.
You can even choose to ignore the plugboard entirely and use jump cables/ribbon cables instead.

[Twst]

no I doubt you can brick it, but you could get to a not nice stack overflow maybe?
I don't know the proton and if that would be much better, but in the end we also wanted to use here something cheap and easy to source, which also is part of the "mission" statement, right?
But there are also secret plans to try to port qmk to another cheap SoC chip and see what happens there.. similar target as usual: around 10 bucks give or take, plus shipping.

QMK Proton C is supposed to be a drop in replacement for the pro micro, but with 40kB ram and 256kB flash vs the 2.5kB and 32kB of the micro. Costs about $16 so not going to break the bank, but going to double the price of the project
But hey, if the micro is good enough - it's good enough. Just thinking about future verisons and if it will work if you want lots of macros and other stuff going on

[adamcobabe]

Very cool. So are you going to adjust the design to not use the SMD resistors and fit the bent through hole ones better too? Why were the footprints so small in this design?

[kmnov2017]

Very cool. So are you going to adjust the design to not use the SMD resistors and fit the bent through hole ones better too? Why were the footprints so small in this design?

Yes. New design files will be uploaded in the next days. The existing footprint will still fit the resistors - but they'll need to be bent....

[gipetto]

I was thinking that capacitive sensing might be used to upgrade the ibm model m to nkey rollover. someone who owns a bolt modded model m could experiment with sticking kapton tape over a membrane contact (like how the varmilo switch is done), then using a capacitive controller to detect if the key was pressed when reassembled. I don't know is it easy to source the membrane ribbon sockets but if it was then you could have a simple drop in controller conversion project.

[anthonymak]

I was thinking that capacitive sensing might be used to upgrade the ibm model m to nkey rollover. someone who owns a bolt modded model m could experiment with sticking kapton tape over a membrane contact (like how the varmilo switch is done), then using a capacitive controller to detect if the key was pressed when reassembled. I don't know is it easy to source the membrane ribbon sockets but if it was then you could have a simple drop in controller conversion project.

the membrane is for switching method and is very different from capacitive sensing. they are not compatible to each other.

[kmnov2017]

This is defently pushing the beamspring and model fs and other capacitive board longer into the 21st century for anyone interested.

Will build a couple of these for sure.

When you have it running on the pro micro, should it be possible to drop in a proton, teensy++ etc with better specs to give a little more breathing room for macros and stuff?
I heard the pro micro could brick if pushed over the edge - how close are you?

The promicros already have double the storage than the original xwhatsit. Storage shouldn't be a problem.

[kmnov2017]

Did anyone else build these yet?

[tentator]

Andrei did!
And I did a couple more just now

[zzxx53]

Any news on when the firmware would be released? Can't seem to find it on the linked Github page or the zip files.

[kmnov2017]

Contact user pandrew for links to the firmware.

[pandrew]

Any news on when the firmware would be released? Can't seem to find it on the linked Github page or the zip files.

Hey guys, we haven't released this publicly yet, only sending the links to people who contacted us, because there's still some kinks we're still working on smoothing out.
Please contact me, or tentator if you want to try it, so we can help you out if there are any issues, and we'd like to know about any issues. Also we'd like to know when it works well too.

Cheers,
Andrei

[kmnov2017]

Works like a charm.

IMG_20200704_130636.jpg (4.11 MiB) Viewed 8561 times

[listofoptions]

updates pushed, sans resistor and trace changes (would have pushed last week, but a lot of weird stuff has been happening) I need to resize for better and larger resistors, or baring that; switch to large smd resistors, not sure which would be more amenable to use. Ive also included the pull request for BOMs, though i havent yet checked them for compatibility (museum work is getting in the way guys, im sorry)

[listofoptions]

kicad is still better than diptrace or FRAGGIN eagle. altium? well no mr moneybags its not, but at that price point literally nothing is good enough

I'm new to all the helpful software but I'm going to bet on Fusion 360's January addition of Electronics:

https://new.engineering.com/story/fusio ... cal-design

I've been pleasantly surprised with their parametric 3D modeling. One of their stated guiding philosophies is "to integrate as many tools as possible across a single workspace". I'd say I'm rooting for them because I want that. I want to 3D model my DEC keyboard's plastic assembly to have a rail that will slide in a row of Amoeba single switch. I'm hoping I can bring in mtl's schematics and layout. That article mentions a lot of great things about knowing the size of both and telling you when there's a problem. I'll let you know if any of it works out.

Not sure why it's FRAGGIN eagle to you but I'm hoping with the integration work they've fixed some of what you didn't like. And it's free to hobbyists.

i wish text was more expressive! eagle lacks many decent features kicad has (well and the opposite can be said too, though i brain fart on what eagle has aside from routing, that kicad doesnt) but my main complaint with eagle is small board support (last i used it i was limited to just over a eurocard and 2 layer, which for most of my projects is essentially useless). anywho, use the tool youre used to! its no fun otherwise!

[DMA]

BTW, I would recommend a DAC with some faster bus. If you can reprogram it between scans, you can have per-column thresholds - which is not as good as per-cell, but close enough for practical purposes.

[pandrew]

@DMA with my qmk firmware we are already reprogramming the DAC. We are dividing up the scan into threshold bins, so keys that have similar signal levels will use one threshold. The number of bins is configurable, and will have an effect on performance.