Model F Mod Log

[msiegel]

Hi :) This is an ongoing log of my extensive Model F modification.

Status: Researching capacitive sensors

I'm planning the following:

[*]*no* case
[*]modified top plate & backplate. to be painted.
[*]original barrels, hammers, springs
[*]new acoustic/mounting foam
[*]original keys, new layout. replacement model m keys for backslash, enter, alt, space
[*]new matrix pcb (2-sided)
[*]modified insulator sheet
[*]new controller: Teensy dev board with GEEKey open source USB firmware; custom capacitance measurement circult; flywired & soldered to matrix
[*]new feet (cork, rubber, or silicone)
[*]original interlocking tabs; a few added nuts & bolts
[/LIST]

This will take awhile to implement :D


Bill of Materials - all prices shipped
IBM Model F AT Keyboard $60
Teensy microcontroller; RGB LED $30 from Adafruit
Replacement Model M keys and caps $20 from Clickykeyboards
PCB Etchant $11 picked up at The Shack (actually in stock!)
12x12" sheet of 2-sided "scissor-cut" copper clad ...

Tools
Screwdriver
Tape Measure
Nibbler
C Clamps
Saw
Hammer & Bag of Concrete (see recent posts)

[msiegel]

here's the initial layout i'm aiming for... pick one
it's a little like the hhkb, as i prefer smaller boards. hopefully the lack of dedicated arrow keys will be easy to adjust to.

edit: added arrow keys, because i use them a lot. see more recent posts.

- the top one re-uses the model f's backspace but has a model m backslash
- the bottom one uses a model m tab key to achieve an hhkb-like backspace
- model m enter key and bottom row keys are used in both
- i have no caps lock key, but i do use an option key on the mac


thanks to quadibloc for the lovely layout illustrations i've stolen from

[msiegel]

would a model f pcb fit in this case:

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=140333923993&ssPageName=ADME:B:WNASIF:US:1123

[ch_123]

I dont think that terminal board is a Model F.

[JBert]

"Space bar transplant" - what do you mean here? Replacing it with a lighter model M one?

I think it is possible to reuse the old one though, all I did was to bend the stabilizer spring so it had less up-force.

[msiegel]

"Space bar transplant" - what do you mean here? Replacing it with a lighter model M one?

replacing it with a *smaller* model m one

the model f pcb has capacitive pads for 4 meta keys on the bottom row... but the lonnnnng model f space bar is in the way of 2 of them. by substituting a model m space bar, a model f's bottom row can have the same layout as a model m

[msiegel]

I dont think that terminal board is a Model F.

it's not... but, if the curvature is right, and if the key-block openings are in the right places to match model f's switch pads, then it might make a good replacement for the giant old model f case

[msiegel]

I like the second layout better.  That's easy to do on a Model M (although I do like the function keys even better next to the space bar and having two of them).  Whats a OPT key?

your retr0right simulation is awesome XD

Option key is used on the Mac; like having that "Windows" key for Windows

[timw4mail]

Yeah, the white balance was off.

[rdh]

Since the Windows / ⌘ / ◇ keys are all really the same thing (they send the same scan codes), I like to refer to 'em generically as the 'logo' key.

Anybody got a good generic name for the Alt/Opt key?  (Do we need one?)

And don't forget the Retr0brite part of the mod.

Show Image

That pumpkin-orange color is appropriately Octoberish, I think.

[msiegel]

This is a great article on the Apple key history.

cool

hey, do we have a history section somewhere?

[ch_123]

No - the IBM entry in the wiki is kinda stalled

Im working on it. At least some of it should be up by this time next week.

[rdh]

Heh.  I just realized whose web site had the scan I trimmed to get this:



:-)

[msiegel]

Funny thing is although I started out on Apple IIs I can't remember anything about the keyboard.

iirc, it was like brown cherries but without a tactile bump

[Xuan]

Anybody got a good generic name for the Alt/Opt key?  (Do we need one?)

Another name for it is Meta from List machines, emacs still refers to it like that.

[lowpoly]

Here are the 73X3832 pics:

http://geekhack.org/showthread.php?t=7377

[quadibloc]

An important question for the Model F experts in this thread. When the key is depressed on a model F keyboard, does the plate go towards the printed circuit board... or away from it?

And you know why I'm asking - could 3278 keys be used on a Model F board, or not?

[msiegel]

An important question for the Model F experts in this thread. When the key is depressed on a model F keyboard, does the plate go towards the printed circuit board... or away from it?

And you know why I'm asking - could 3278 keys be used on a Model F board, or not?

definitely toward. i just verified this with a disembodied keyswitch module

[JBert]

An important question for the Model F experts in this thread. When the key is depressed on a model F keyboard, does the plate go towards the printed circuit board... or away from it?

And you know why I'm asking - could 3278 keys be used on a Model F board, or not?

The hammers are pressed down when you press the key, just like a model M.

Whether it can detect keypresses from a 3278 depends on the controller. If it only detects changes in capacitance instead of a drop in capacitance, it might work.

[timw4mail]

You're expecting this to take a while, aren't you Ripster?

[msiegel]

can anyone recommend a laser cutting service, for stainless steel and plastic?

i'm looking at alternatives for fabrication of a new backplate and top plate / keymodule template.

[timw4mail]

Wouldn't aluminum be cheaper?

[msiegel]

Wouldn't aluminum be cheaper?

i don't know - do you think it would?

i was also considering plastic for the new top plate / template. if the plates were flat instead of curved (heresy, i know), then the foam pad might not be needed.

[timw4mail]

I'm pretty sure, as aluminum is a pretty common element, and stainless steel is an alloy of several different metals, and there are several different grades of stainless steel available...it just simplifies things to get aluminum.

[itlnstln]

I'm pretty sure, as aluminum is a pretty common element, and stainless steel is an alloy of several different metals, and there are several different grades of stainless steel available...it just simplifies things to get aluminum.

Aluminum is very difficult to manufacture as it has to be "freed" from other minerals, rocks, etc., usually bauxite.  Check it.

[JBert]

i don't know - do you think it would?

i was also considering plastic for the new top plate / template. if the plates were flat instead of curved (heresy, i know), then the foam pad might not be needed.

I think the foam is only there to stick to the spring tubes so they won't fall out when you separate the backplate from the front.

After all, now that the foam disintegrates even when slightly touched, it can't do anything structural or you would have felt the difference.

[msiegel]

hmm... it turns out the bottoms of the keyswitch modules are not perfectly flat.

they're very slightly curved.

i'm guessing this is by design, to match the curvature of the plates.

[msiegel]

Lasers for plastic?

for accuracy

plus, isn't abs kind of hard to drill

[lowpoly]

Check emachineshop.com  for something like this. They have a vast selection of materials. If this is going to hold the bent pcb then regular Aluminum will be too soft. Stainless will be nice but expensive. Most keyboards seem to have regular steel plates with some sort of surface treatment.

The emachineshop software is good for mostly two-dimensional objects. You should be able to bend it at a local tool shop later if the software doesn't allow for this.

Shapeways is another rapid prototyping service.

[msiegel]

Check emachineshop.com  for something like this.

fascinating... thanks lowpoly

[quadibloc]

Tell me when you have a Model F with a usable layout.

I'd have one already if I thought the ISO keyboard was usable - as I have two IBM 122-key keyboards, one a Model M terminal keyboard, and one the big Model F one for the 3270 PC.

I think the only way to mod a Model F to change the key locations would be to make your own printed circuit board to replace the one that came with the keyboard. Electronics hobbyists have done it, so it's possible.

[msiegel]

I think the only way to mod a Model F to change the key locations would be to make your own printed circuit board to replace the one that came with the keyboard. Electronics hobbyists have done it, so it's possible.

i'm interested in seeing their work. do you remember where you saw that?

I'd have one already if I thought the ISO keyboard was usable - as I have two IBM 122-key keyboards, one a Model M terminal keyboard, and one the big Model F one for the 3270 PC.

edit: how come the 3270 board isn't usable... interface compatibility?

[msiegel]

note: the plastic inserts in the barrels of a model f are a different shape than the model m's.

this means that using a model m's stabilized keys, such as the enter key, requires the model m's barrel inserts as well as its keys.

...now working on getting a model m, for ansi keycaps and barrel-inserts, and to test whether the membranes can be used as capacitive sensors...

[quadibloc]

i'm interested in seeing their work. do you remember where you saw that?

It used to be fairly common for electronics hobbyists to make their own printed circuit boards. Just single-sided single-layer, though. You could buy a kit where you got a board completely covered with copper on one side, and you stuck a mask to the copper, and then used an acid solution provided in the kit to remove the copper that was exposed.

edit: how come the 3270 board isn't usable... interface compatibility?

My reference to usability was in terms of the layout. The main typing area of the 122-key keyboard is laid out the same way as an international 102-key keyboard, meaning that there is an extra key between Z and the left-hand shift key.

The controller apparently only has the issue of not sending key release codes, and there is supposedly a workaround for it according to posts on the subject here, but I'm afraid I haven't quite understood them well enough to understand what the workaround is: sending commands to the keyboard's controller, patching my computer's BIOS, using special driver software, or what.

[msiegel]

It used to be fairly common for electronics hobbyists to make their own printed circuit boards. Just single-sided single-layer, though. You could buy a kit where you got a board completely covered with copper on one side, and you stuck a mask to the copper, and then used an acid solution provided in the kit to remove the copper that was exposed.

this pcb material looks like it might be thin enough to be flexible:
http://www.goldmine-elec-products.com/prodinfo.asp?number=G17115

[msiegel]

ordered a "teensy" avr dev board from adafruit.

i just couldn't stand it any longer

this will be most of the hardware for the new controller.

[msiegel]

note to self: don't require many fasteners for assembly!

remember what a pain it was working with 42 screws/nuts for the model m mini mod

[msiegel]

you guys can chime in anytime... but this here is a log, man.

i have an older cellphone camera

[msiegel]

design idea for capacitance measurement:

AVR PWM --> (555 + Sensor Pads) --> Smoothing Capacitor --> AVR ADC

in this scheme the adc will have to measure ~100K samples per second
edit: hmm, 15K per second may be the limit

reference:
http://www.sokcapacitor.com/learning/2008/Capacitance-meter-with-two-555-timers-59.html

[msiegel]

the top and bottom plates are flat stainless steel, pcb is flat, and the keyboard has a rubber gasket running around the outside edge.

a small controller card is sandwiched between the steel plates, and a single multicolor led indicates caps lock, Fn key activation, and special modes.

main-cluster keys are as close as possible to their relative positions on an ansi layout board: esc is still at the upper left of the board, and backspace at the upper right. only tilde is moved.

[msiegel]

pic of the day: the lab, where ibm technology is analyzed

[msiegel]

Capacitive Sensor Design Resources

http://www.planetanalog.com/showArticle.jhtml?articleID=196902891

http://www.slideshare.net/ruthmoore08/capacitance-sensing-pccompatible-usb-capsense-matrix-keyboard

http://sensi.org/~svo/capsensor/index.en.html

http://www.cypress.com/?docID=2192 (PDF)

http://www.planetanalog.com/showArticle.jhtml?articleID=189602704

https://secure.cypress.com/?docID=2442  (PDF)

[msiegel]

Hey that pic looks fine!  You might want to think of making sure the controller picks up the "extra" pads - the two next to the shift keys and the one next to ENTER (can't tell from the pic if that one is there).  That way when you play with layouts later all you have to do is stuff the key wells differently with springs/hammers/keys.

thanks yes, i guess i should figure out how to use the existing pcb first, with a new controller... and wait until later to make a new pcb.

[msiegel]

it was working initially, with an old HP 286, but that computer is gone now

although i have a ps/2 to usb adapter which worked well with my model m, maybe there's not enough power available?

[msiegel]

there's a teensy development board coming in about a week

i still have to come up with a capacitance measurement circuit, figure out the ibm matrix, then switch mnemonix's usb code to use the avr's hardware usb, but still... it's coming.

edit: i just plugged the controller and matrix pcb into a usb converter... no output :-/
i'll try tim4

[lowpoly]

Why not bend the top and bottom plates?

How are they held together on the Model F, could you explain this?

[ch_123]

Someone needs to get this guy a Beam Spring keyboard. Nao.

[ak_nala]

Capacitive Sensor Design Resources

http://www.planetanalog.com/showArticle.jhtml?articleID=196902891

http://www.slideshare.net/ruthmoore08/capacitance-sensing-pccompatible-usb-capsense-matrix-keyboard

http://sensi.org/~svo/capsensor/index.en.html

Another article mentioned in the Site Editor's Note in the first link - -

http://www.planetanalog.com/showArticle.jhtml?articleID=189602704

EDIT: Also found a direct link to a PDF of the article in your second link that is a bit more convenient to view than the flash-based version on the slideshare site - -

https://secure.cypress.com/?docID=2442

[msiegel]

the plates are connected by sliding tabs. one of the tabs is longer, and bends to lock the plates together.





i'd love the plates to be curved, but i have no idea how to get the exact curvature

[msiegel]

Another article mentioned in the Site Editor's Note in the first link - -

http://www.planetanalog.com/showArticle.jhtml?articleID=189602704

EDIT: Also found a direct link to a PDF of the article in your second link that is a bit more convenient to view than the flash-based version on the slideshare site - -

https://secure.cypress.com/?docID=2442

thank you

[lowpoly]

the plates are connected by sliding tabs. one of the tabs is longer, and bends to lock the plates together.

Thanks. I guess just bolting them together would be easier then. The curvature makes it more difficult because of the tension from the pcb. Unless there is a way to perma-bend that too.

Bending sheet metal is done with three cylinders:





^^^ There are much smaller machines than these. A machine shop shouldn't be too hard to find near you. emachineshop may be able to do this as well. IIRC, you can add text to your CAD drawing requesting additional machining steps.

Here's a small one:



Page with more pics (site in german):

http://www.top-maschinen.de/biegen-kanten/biegemaschine-quantum-bm3/index.htm

[msiegel]

wow, those are awesome

the pcb is fairly thin, and is held flush against the backplate by two metal pins:




it attaches via these mounting holes:




since the pcb gets the same curvature as the plates, i believe there is a consistent pressure across its face (between the pcb and the barrels that lay on top of it).

[lowpoly]

Do the barrels apply additional pressure to the pcb?

[msiegel]

Do the barrels apply additional pressure to the pcb?

yes, this seems to be one of the reasons for the foam above the barrels.

also, the bottoms of the barrels are slightly curved, not flat -- so they match the pcb curvature.

the layers are:

top plate
foam
barrels
pcb
dielectric film
backplate

[msiegel]

Wondered about that.  I think the rubber mousepad type stuff might be a good substitute.  Cutting all the holes wouldn't be much fun though.   I have some really old mouse pads around here that look bad but aren't coming apart.

it's an interesting problem. the best material would compress well and be resilient, like an open-cell foam, but not fall apart like one

ibm engineers coated one side of the foam pad to seal it off -- the side that touches the barrels. closed-cell foam wouldn't compress well enough, so their choice seems like a good compromise, given the materials they could choose from at the time

i'm not looking forward to cutting holes

[msiegel]

Capacitive Pads

Let's see if we can figure out how these pads work...

*pulls out mr. blurryphone*


for each keyswitch, the front side of the pcb has 2 pads and pair of dots. they're all covered with solder mask, so they are not conductive. the hammer fits over these pads; it is turned away from them when a key is up, and turned against them when a key is pressed.

shadows of some heavy traces are visible through the pcb.




*flips pcb around*


the back side has a single pad bordered by traces, placed in exactly the same location as one of the front pads. it is also covered with solder mask.



...

[msiegel]

on the front side of the pcb, one of the two pads in a pair connects to a trace, while the other pad "floats" on the board, unconnected to any traces.

the back side pads are located directly below the floating front-side pads. the traces bordering back side pads are connected to ground. a ground plane borders the entire pcb.

traces on the front and back sides are thin and cross at right angles. this would minimize capacitive coupling between them.

edit:

the matrix has 16 columns and 8 "rows".

[msiegel]

on the front side of the pcb there are pairs of pads. one of the two pads connects to a trace, while the other pad "floats" on the board, unconnected to any traces.

on the back side of the pcb, there is only one pad per front-side pair. the back side pads are located directly below the floating front-side pads.

traces on the front and back sides are thin and cross at right angles. this would minimize capacitive coupling between them. the traces bordering back side pads are connected to ground. a ground plane borders the entire pcb.

from counting the traces entering each side of the pcb, the matrix can be seen to have 16 columns and 8 "rows". traces for rows are on the front side, and traces for columns are on the back side.

[msiegel]

Hey guys,

I need to fill in some blanks in my modified Model F layout...

so I'm looking for these Model M keys in original colors, their accompanying barrel-inserts, and any stabilizing bars:

• Backspace (plus its insert)
• Tab
• \ (backslash)
• Enter (U.S. version, plus its insert)
• Ctrl
• Alt
• Spacebar

I'd prefer one-piece keys if possible. I do need Alt & Ctrl to be original width rather than the narrower Unicomp 104 width.

What do you guys think is the least expensive way to obtain all these? :)

[msiegel]

$6 nonfunctional on Ebay.  Shipping seems high so maybe if you email him you can talk him down.

The inserts are the problem.  You have to get a keyboard for those.

thanks ripster... & that's a good point: shipping for a whole KB is much higher than for a few keys

hmm...

[ak_nala]

Also, that keyboard being a 1391401 from 1989 almost certainly has stabilizer bars instead of inserts and, of course, two-piece keys.

You would need a later model to get both one-piece keys and inserts, though you could get inserts in a later Blue Label 1391401 if you don't mind two-piece on the non-insert keys.

[msiegel]

Also, that keyboard being a 1391401 from 1989 almost certainly has stabilizer bars instead of inserts and, of course, two-piece keys.

You would need a later model to get both one-piece keys and inserts, though you could get inserts in a later Blue Label 1391401 if you don't mind two-piece on the non-insert keys.

thanks ak_nala

it looks like keys with stabilizer bars probably will work -- *if* the bars are mounted facing toward the user rather than away ...i couldn't quickly find a photo of them around here

i'll still need the inserts, but other than that i'll take what i can get

[kishy]

Just a thing about the pictures...

If your phone has a "macro mode" or an option or switch which has a little flower icon, use it. It allows for better quality closeup shots.

I haven't seen a camera phone yet that doesn't have the feature...it's usually hidden pretty well though.

[msiegel]

If your phone has a "macro mode" or an option or switch which has a little flower icon, use it. It allows for better quality closeup shots.

thanks kishy

hmm... reviews indicate mine doesn't have autofocus *or* macro mode
it's an old Nokia E65.

[msiegel]

this requires patience

[msiegel]

om nom

[msiegel]

Sorry to get back onto the whole foam thing but it looks like the holes are keyed so the foam really doesn't even hold the wells in place.  More sound insulation than anything else.

the side of the top plate that faces the wells is not perfectly smooth -- there are some protrusions from plastic stabilizers which are riveted through additional holes in the plate.

the foam not only cushions wells against the top plate, with the pressure and friction providing a more secure mount, but also greatly improves the tolerances for assembly.

any sound dampening might just be an added bonus.

[msiegel]

updated layout, with modified top plate illustrated (dark gray)

blue keys are model m replacements (on order from clickykeyboards)
yellow keys are for cursor control, Up Left Down Right

it will be small, because there's no case

[msiegel]

Nice layout.  It looks like it will be tricky getting everything aligned - after my N52 mod I appreciated the small tolerances involved.

thanks

the existing cutouts should work for all but the bottom row. yeah... we'll see about aligning things with 1/10 mm accuracy

[msiegel]

parts arrived today from clickykeyboards!

the backslash, enter, alt, and space bar are replacement keys from model m, to fill in blanks in the new layout.

the entire layout is shown here loose. the keys are not attached to anything yet...

this layout features:
- caps lock = Function
- num lock = "logo" (windows/opt),
- the original number-pad arrows,
- esc & backspace are in ansi-like positions at the corners of the board
- alt and space bar are in same locations as on model m
- no missing characters for u.s. users

apparently this is model 8426

[Mnemonix]

this layout features:
...
- the original number-pad arrows,

Very nice!  The arrangement of the arrow keys reminds me of my old Commodore 128.
Checked a picture... the sequence was a bit different on it: Up, Down, Left, Right.

[Shawn Stanford]

This thread is teh ****e!

[msiegel]

Is the pad active next to the Rshift key?   I still like the traditional T style better.  It  would be offcenter by a 1/4" but still not bad.

My Model M International has that pad active.

thanks... reminds me to put a pad under every barrel, when i make the new pcb

[Shawn Stanford]

...snip... this layout features: ...snip...

Show Image

Where are Ins, Del, Home, End, PgUp & PgDn?

[msiegel]

Where are Ins, Del, Home, End, PgUp & PgDn?

good question!

the keymap will be user-definable...

i imagine:
Del = Function + Backspace
Ins = Function + ~
Home, End, PgUp & PgDn = Function + arrows

[Shawn Stanford]

So, the attention keypress will be (for instance) Ctl+Alt+Fn+Backspace?

[msiegel]

So, the attention keypress will be (for instance) Ctl+Alt+Fn+Backspace?

XD heck, i don't know...

that one might call for a user-defined macro


edit: but yes, that's correct in the scheme i'm currently imagining

[ch_123]

OK, here's a pic from Chloe showing those pins poking through the foam in a XT Model F.

On the AT Model F, there's no foam on the top plate. Only between the two plates, so most of it is unexposed and in good condition.

[msiegel]

On the AT Model F, there's no foam on the top plate. Only between the two plates, so most of it is unexposed and in good condition.

right, that shot shows what's under the top plate

[msiegel]

Everything looks pretty similar from the F AT vs F in the pics except the metal back plate and the notches in the plate are in a different part relative to the circle.  Controller obviously.  Wonder if springs/hammers are the same?

yes, they're quite similar.

the top plates have a different surface treatment: painted on the XT, and what looks like black oxide on the AT.

the hammers are slightly different in shape, as are the barrels (one difference being those notches you mentioned).

not sure about the springs. one day i'll have to get a set of XT springs/barrels/hammers to compare and build with


on an unrelated note, the switch modules from (3) 122-key model F terminal boards could be used to build (5) of these mini keyboards

[msiegel]

Is the pad active next to the Rshift key?   I still like the traditional T style better.  It  would be offcenter by a 1/4" but still not bad.

hmm... if you did that, then the ~ could be moved to the lower right corner of the board, and a standard double-wide backspace would fit in the top row

[JBert]

Everything looks pretty similar from the F AT vs F in the pics except the metal back plate and the notches in the plate are in a different part relative to the circle.  Controller obviously.  Wonder if springs/hammers are the same?

One difference I've found is that the backplate of the key assembly is a lot more "loose", i.e. you can slide it open by hand and the barrels can still move a bit when the thing is assembled. Compare this with an XT where I needed a hammer to shift the plates and where the barrels are clamped down.
Maybe as a result, you get the different key feel like a de-revetting model M compared to a complete one.

[msiegel]

making progress, slowly and steadily

[lowpoly]

Maybe use a Dremel?

[chimera15]

Really interesting.  What are you using for the electronics of it?  The same controller?

[msiegel]

Really interesting.  What are you using for the electronics of it?  The same controller?

the controller is a Teensy mini AVR development board, which will be running open-source GEEKey software.

this mod will also use a homemade capacitive sensor PCB, thereby replacing all of the Model F's original electronics

[chimera15]

the controller is a Teensy mini AVR development board, which will be running open-source GEEKey software.

this mod will also use a homemade capacitive sensor PCB, thereby replacing all of the Model F's original electronics

So cool,  I have to get one of those teensy's.

I'll have to look up GEEKey, thanks a bunch.

What's the capacitive sensor going to do?

[msiegel]

What's the capacitive sensor going to do?

the capacitive matrix pcb senses whether each key's hammer is in the down or up position (close to, or far away from, the pcb)... in order to register keypresses.

this is similar to the model m's membrane sheets, but instead of relying on the conduction (or not) of electricity, it uses "capacitive coupling" of a test signal:

if a hammer is close to the pcb, a test signal will pass through sensor pads and be detected; but if a hammer is far away, the signal is blocked.

[msiegel]

This is where it gets interesting.  How are you going to reverse engineer a Model F capacitive circuit??

operation of the sensor matrix seems very straightforward, which leads me to believe the circuit is so simple it does not even use an oscillator.  

so far the theory is:

row traces are on the pcb's front, and the controller chip applies test pulses to each row. column traces are on the pcb back, connected to a few (inverter?) chips that clean up any test pulses that make it all the way through to the rear pads. then the pulses (if they made it through) are read by the controller chip.

no diodes

it looks like all the hard work is being done by capacitive coupling between the two front pads, mediated by the hammer.  the magic of capacitive coupling is also used to pass the signal *through the pcb* from the second "floating" (electrically unconnected) front pad, to the back-side pad.

[chimera15]

the capacitive matrix pcb senses whether each key's hammer is in the down or up position (close to, or far away from, the pcb)... in order to register keypresses.

this is similar to the model m's membrane sheets, but instead of relying on the conduction (or not) of electricity, it uses "capacitive coupling" of a test signal:

if a hammer is close to the pcb, a test signal will pass through sensor pads and be detected; but if a hammer is far away, the signal is blocked.

Do you have to do that because the teensy won't register the electrical connection or something?  Why not just use the normal matrix?  Oh because you ruined the matrix when you hacked the board apart?  Why not just rewire it? Or it's easier to do it this way?

[msiegel]

Do you have to do that because the teensy won't register the electrical connection or something?  Why not just use the normal matrix?  Oh because you ruined the matrix when you hacked the board apart?  Why not just rewire it? Or it's easier to do it this way?

ideally i want to gather enough information, materials specs, cad files, source code, assembly techniques etc... so that keyboards based on model f technology can be manufactured again, in a modern and affordable form.

[chimera15]

ideally i want to gather enough information, materials specs, cad files, source code, assembly techniques etc... so that keyboards based on model f technology can be manufactured again, in a modern and affordable form.

So you're using that method so that you can manufacture it down the road, not for any practical reason?

[msiegel]

So you're using that method so that you can manufacture it down the road, not for any practical reason?

practically, the bottom row of the original pcb doesn't match the new compact layout, and there are too many rows+columns for the Teensy controller

[chimera15]

Ah that's interesting.. how many keys will the teensy support?

[msiegel]

Ah that's interesting.. how many keys will the teensy support?

i'm planning a 15x5 matrix (using 20 pins), and saving a few more pins for an rgb led...

a maximum of 75 keys are possible in this configuration.

the compact layout i have in mind only includes 66 sets of sensor pads though

[Mnemonix]

i'm planning a 15x5 matrix (using 20 pins), and saving a few more pins for an rgb led...

Great to see progress here!

A 15x5 matrix will be quite dense and introduce a lot of possibilities for ghost keys to occur; at least on a conductive matrix w/o diodes this would be the case. If there is no connection between the pads, there is no way for ghosting on a capacitive matrix, right? That would be quite cool, actually.

Otherwise I would be a bit worried about the Function key you wrote about somewhere because it needs to be detectable in any key combination, thus requiring a row of its own.

EDIT: A 12x8 matrix seems attractive, too. It would allow for 96 keys and looks nicer from a software point of view.

[msiegel]

A 15x5 matrix will be quite dense and introduce a lot of possibilities for ghost keys to occur; at least on a conductive matrix w/o diodes this would be the case. If there is no connection between the pads, there is no way for ghosting on a capacitive matrix, right?

yes, this is still an open question... although there's no conduction path between "switches", i don't know whether there's still a way they can interfere with each other.

i plan to test this. hopefully a 2x2 matrix made of cardboard and aluminum foil will be good enough to know for certain

[JBert]

Well, your research can also be handy for buckling beam keyboards in case those need a new controller board.

[Mnemonix]

i plan to test this. hopefully a 2x2 matrix made of cardboard and aluminum foil will be good enough to know for certain

I'd guess so.

Great, now I want a Model F, too. Congrats.

[microsoft windows]

I never knew Model F's had green Alt keys.

[ch_123]

They don't... Where did you get that idea?

[msiegel]

They don't... Where did you get that idea?

that's a model m alt key in the photo

[msiegel]

Regarding Ghosting it seems to me that it doesn't matter what kind of switch mechanism is used.  It's inherent to the way the matrix is laid out.  

The other thing is keybounce and other issues and how a capacative circuit handles that.   Dunno - seems complicated. May be wrong - I'm not an EE.

i'm not sure either, yet... but i suspect the capacitive technology is nowhere near as complex as i had assumed

[JBert]

Regarding Ghosting it seems to me that it doesn't matter what kind of switch mechanism is used.  It's inherent to the way the matrix is laid out.  

The other thing is keybounce and other issues and how a capacative circuit handles that.   Dunno - seems complicated. May be wrong - I'm not an EE.

Actually, it's easy once you think of it: the capacitance varies with the distance between the hammer and board. When the hammer bounces, it only bounces a fraction of the distance it should usually overcome to be considered closed so the capacitance won't vary much. Hence the only thing you need is a "Schmitt trigger" which has two states: [capacitance is less than "switch closed" threshold] or [capacitance is greater than the "open switch" threshold]. Anything in-between is considered hysteresis so you simply stay in the current state.

Compare this with a regular switch: either you have a running current or you don't. Bouncing is a form of the latter. Result: you need timing to see if the switch stopped bouncing when you lost contact.

Capacitance sensing clearly wins here in simplicity...

[msiegel]

Hence the only thing you need is a "Schmitt trigger" which has two states: [capacitance is less than "switch closed" threshold] or [capacitance is greater than the "open switch" threshold]. Anything in-between is considered hysteresis so you simply stay in the current state.

the trigger sounds easy to simulate through software, and i'll bet a single pulse to the sensor pads can be used to sample a switch's state.

maybe the threshold function can be performed by just reading a digital input on the uC... if voltage of received pulse > logic high threshold (2.5v), then switch is closed and input reads "1".

that'd be awesomely simple

[msiegel]

excellent find... thanks ripster, and talis

i'm thinking of an alternate technique, which i believe model f uses:

when a hammer is too far away from its pads, there's too little capacitive coupling between a pin generating a test pulse and an input pin (both on the microcontroller)... so the test pulse is too weak to be detected.

but when a hammer is close enough, coupling is strong and a pulse comes through.

where's talis, to debunk this theory?

[msiegel]

both plates cut to size




top plate - cutouts made for new bottom row layout




the tabs for shift key stabilizer bars have been destroyed in the process of cutting. this mod will use shift keys that do not have stabilizer bars

[ch_123]

Cool stuff.

I assume extra barrels are going to go into the space you cut out beside the space bar? Are you making your own keyed strips?

[chimera15]

awesomeness.  Might end up doing this mod myself if you can help me figure out how to program the teensy.

I'm still really curious about the details of how you're going to interface the matrix with a new pcb and everything too.

[microsoft windows]

On most buckling spring keyboards, the shift keys don't need stablizer bars anyway.

[ch_123]

That's true of Model Ms made after 1989. Not the Model F or older Model Ms. Although perhaps the newer stabilizer slots would fix that - or at least timw4mail said as much.

[ak_nala]

Perhaps a step drill bit in a drill press? Then nibble, saw (jeweler's saw) and/or file the notch?

Seems a simpler and less expensive solution than a specialized punch, especially with a thick plate and the risk of deforming the curve in the process of punching (but then my experience with punches is pretty limited).

[msiegel]

these are the tools used so far.

since this is a prototype, i'm thinking of using wood for the left side of the bottom row. if all goes well i'll re-use a strip of four cutouts for the right side

next i plan to take a break from metalwork to address the electronics.

[msiegel]

Is the concrete to make the heaviest keyboard ever?  

it's better than lead

i used the hammer and bag of concrete to get the top plate back in shape after accidentally bending it during cutting. between the nibbler and the clamps it got a few big dents

[msiegel]

i thought the Monolith was touch-activated...

[chimera15]

Is that somewhere in Japan?

[PRISONER 24601]

This thread/project is amazing. kudos.
Have you tried contacting IBM for the controller information? I mean, this patent is at least 100 years old it can't be a big company secret or anything. Probably not. At the worst, they'd just hang up on you.

[microsoft windows]

Here's a mouse to go with it.

Show Image

I wish that was my backyard.

[msiegel]

apparently those are in Xiamen, China

[chimera15]

apparently those are in Xiamen, China

Cool, wonder if I can see them in google earth. lol Maybe China isn't that detailed though....

 Edit: Well couldn't find them, but that's a real pretty city.  Lots of sculpture.

[msiegel]

This thread/project is amazing. kudos.
Have you tried contacting IBM for the controller information? I mean, this patent is at least 100 years old it can't be a big company secret or anything. Probably not. At the worst, they'd just hang up on you.

thank you

i'll start by studying this patent. page 9 describes how a capacitive sensor matrix works.

the patent is for ibm's beam-spring keyboards, but it looks like the same type of sensor matrix that's used in the model f.

edit: details on ibm's capacitive sensor matrix technology. maybe this explains why there are a bunch of NOR chips on the model f controller card

[msiegel]

keys mounted, to keep them sorted out.

the keys will be about 3mm taller when springs and hammers are installed -- the barrels will be quite visible




this prototype has a thin bezel on the sides.

for the "official" version, i'm planning a 1/2 inch bezel on the sides and front, about 1 inch at the top... as well as a 1/4 inch-tall rubber gasket sandwiched between the plates, that sticks out 1/16 inch all the way around the board as a bumper. the plates will be cnc-machined 20 gauge stainless steel

[chimera15]

Looks awesome!  This is the first thing I've typed using dvorak without looking at a cheat card!

[msiegel]

Looks awesome!  This is the first thing I've typed using dvorak without looking at a cheat card!

congratulations!

i hope i can learn even half that fast when the time comes for me to study touch typing

[msiegel]

low-Z amplifier design for capacitive sensors
http://www.capsense.com/capsense-wp.pdf

[msiegel]

sensor matrix note:

according to a formula in the ibm capacitive matrix patent, it's likely the sensor pad output current is in the range of 1 to 10 microamps, when fed with a 5 volt pulse to test the switch state.

can an AVR sense that little current directly? if not, op amps will probably be needed on the 5 sense lines.