Introducing the GH36 Programmable Matrix Keypad

[domoaligato]

$20.00 sent your way bro. Thanks again!

[samwisekoi]

$20.00 sent your way bro. Thanks again!

Thanks very much to you and other donors!

 - Ron | samwisekoi

[domoaligato]

I will send more later when I can btw. I know that it is a little short of the total. but that is what I had in my paypal atm.

[samwisekoi]

I will send more later when I can btw. I know that it is a little short of the total. but that is what I had in my paypal atm.

No, don't.  You are good.  Also, thanks to the additional sponsors who donated since my last post.  It is truly appreciated.

Not required, mind you, but very much appreciated nevertheless.

Best to all,

 -Ron | samwisekoi

[domoaligato]

I am fully dedicated to this. I would love to see something like the duck ergo only open source and this seems to be the closest thing I have seen.

[dorkvader]

Prototype PCBs have been put in the mail.  If you sent me your address, I sent you a pair of PCBs.

Two PCBs arrived in the mail today, and I had some questions about them.

All PCBs have had the errant trace cut and then tested.

Some of the pins (the ones at the Teensy for Col7-11) were mis-labeled, so I blacked those out.

Well that answers the questions. Looking at the PCBs, I was almost sure that's what occurred.

I can confirm that metalliqaz' latest version (version number unknown to me at this time) of the Easy Keymap firmware-generator works perfectly.  The resulting firmware has been loaded onto the Teensy, and the keypad then tested on both Windows and Linux.

Go forth and prototype!

 - Ron | samwisekoi

p.s.  I advise against mounting the Teensy upside-down on top of the keypad.  However, Retro DSA works wonderfully and has all the 1x keycaps you could want.

I planned to do just that. We'll see if it works out or not. I also need to decide what exactly I'll be making.

I was thinking 1x tenkey and 1x gamepad to compliment my ergodox half. I'll be testing metaliqaz' firmware and soarers' firmware. Do you have a recommendation as to the transistor I use for the LEDs? all the ones I have in that size are very expensive hand-matched jfets.

I'll be adding LEDs to everything for fun as well.

[dorkvader]

Found an issue: R35 resistor powers location (row,column) (5,6) or (5.5,6). If you have the key "unsplit", the cherry PCB-mount stabilizer blocks it physically (on either side). Proposed fix: move it to the space between row4 and row5.




edit: I guess it's not too big a deal, as people can just use the ground side of R36 if they want to have one switch there, but people with a layout like yours (stab on (5,2) and (6,2) and split (5,1) and (6,1)) will still have an issue.

Also: another PCB error, R4, R5 and R6 are grounded to the ground plane on both sides. I will fix mine with a knife, but anyone wanting to backlight (6,1) (6,7) (5,1) (5,7) or (4,1) (4,7) will need to mod the PCB

Edit: opay, I'm going to have some fun with clipped diode leads off the floor wiring up the "power" side of those LEDs to that transistor.

[samwisekoi]

Wow, ok, lots of input.

I used a 31 cent transistor, Mouser P/N 512-MPSA29, which should handle up to 40 20mA LEDs.

I'm going to re-do those resistors, so thanks for prototyping that!

Also, sorry about the errors.  Let's get this one right before GH-122 hits fiberglass and copper!

 - Ron | samwisekoi

[metalliqaz]

If errors are found in the software, just let me know.  I posted a new beta tonight because I couldn't remember if I had posted the newest GH36 code in the last one.  Also backlighting still needs to be worked out.

[dorkvader]

Also, sorry about the errors.  Let's get this one right before GH-122 hits fiberglass and copper!

 - Ron | samwisekoi

Glad I could be helpful.

While I wait the transistors to arrive, I think I'll wire up the power side of the LED matrix and find some standoffs (I did decide to go for the underside teensy for now, mostly because all my teensies have pins on them already from PJRC and I don't feel like desoldering them).

Also backlighting still needs to be worked out.

As of now, there are only two pins connected to LEDs
One is the "num" pin (labelled B4 on the teensy), connected through a transistor to a single LED, and the "WASD" pin (D7 on the teensy) connected through a transistor to the WASD keys as well as the one to the left of "a and to the right of D.

Or otherwise keys (3,3) (4,1) (4,2) (4,3) (4,4) and (4,5) are powered from the "wasd" pin.

If you flip it, that corresponds to keys (3,9) (4,11) (4,10) (4,9) (4,8) and (4,7) on the back.

You shoudl be able to drive that transistor with a normal PWM signal, as long as the frequency isn't over the transistors max rating (125 MHz for the ones I got).

In the current revision, the "LED1" and "LED2" pins (B5 and B6 on the teensy) are no longer connected to anything.

[samwisekoi]

Also, sorry about the errors.  Let's get this one right before GH-122 hits fiberglass and copper!

 - Ron | samwisekoi

Glad I could be helpful.

While I wait the transistors to arrive, I think I'll wire up the power side of the LED matrix and find some standoffs (I did decide to go for the underside teensy for now, mostly because all my teensies have pins on them already from PJRC and I don't feel like desoldering them).

I was serious when I said don't do the upside-down thing.  There is a long follow-on story, and it ends with JD sending me a replacement JD40 PCB.

On the other hand, a Teensy umbilical-cord thing would be very handy for such testing.  I might work one out.  I am thinking of a low-profile set of wires leading off to a small board with a socketed Teensy.  That way the only things at risk from soldering and desoldering would be a set of wires, not the Teensy or the PCB.

Anyhow, thanks!

Also, thanks to metalliqaz as well!

 - Ron | samwisekoi

[domoaligato]

I received mine today. I will probably be able to start trying to test it tomorrow.
Thanks!

[samwisekoi]

New mechanicals to address the issues DV has discovered.

 - Ron | samwisekoi

[HoffmanMyster]

My PCBs have arrived. 

"Unfortunately", I'm going out of town this weekend, so I won't be able to assemble it until next week.  It looks nice so far though!

[dorkvader]

New mechanicals to address the issues DV has discovered.

 - Ron | samwisekoi

oh cool, space for 1.5x caps at the top.

I think it should be noted that the components are supposed to go on the bottom. If you put them on the wrong side you'll end up without stabilizers or having to desolder. This is only true for a couple places.

So just put all the diodes and resistors on the bottom and call it good.

[samwisekoi]

All the passives are silkscreened on the bottom side only.

But where do they interfere with the stabs?  I tried to keep clear of all keep-away zones.

 - Ron | samwisekoi

p.s.  The space at the top could also be for a gap between rows.  Also, if you turbn the PCB, those 1.50 keys can be on the outside edges of a two-PCB split keyboard.

[dorkvader]

All the passives are silkscreened on the bottom side only.

But where do they interfere with the stabs?  I tried to keep clear of all keep-away zones.

 - Ron | samwisekoi

p.s.  The space at the top could also be for a gap between rows.  Also, if you turn the PCB, those 1.50 keys can be on the outside edges of a two-PCB split keyboard.

I meant that in many places you can put them on the top (they don't interfere with the MX switches). I prefer this style, so I can have a proper component side and solder side.

I'd just hate for a novice DIYer to start soldering components down and then realize he has to desolder them all.

Looking at the new image, I don't see any place where they interfere with the stabilizers, provide you solder them on the correct side.

[samwisekoi]

All the passives are silkscreened on the bottom side only.

But where do they interfere with the stabs?  I tried to keep clear of all keep-away zones.

 - Ron | samwisekoi

p.s.  The space at the top could also be for a gap between rows.  Also, if you turn the PCB, those 1.50 keys can be on the outside edges of a two-PCB split keyboard.

I meant that in many places you can put them on the top (they don't interfere with the MX switches). I prefer this style, so I can have a proper component side and solder side.

I'd just hate for a novice DIYer to start soldering components down and then realize he has to desolder them all.

Looking at the new image, I don't see any place where they interfere with the stabilizers, provide you solder them on the correct side.

Yes, I put all my passives and switches on the same side if I can.  I looked again and don't see anywhere that the passives interfere with the stabilizers.  Can you tell me the row and column positions of any that do?

I am moving them around to squeeze the Teensy in anyhow; that version was the "perfect world", but I thought it had no conflicts.

Please tell me where they are so I can move them.

Thanks!

 - Ron

[deci]

Oh wow, I've been looking for a way to make a custom matrix keyboard for awhile.
(I want something like a traditional keyboard just with the keys setup in a non staggered grid.)

Two of these side by side seems perfect. 
Definitely looking forward to the beta / prod version of this.

Now to find a case that might be able to fit these so I won'd have to fabricate my own.

[samwisekoi]

Was it me, the PCB or the firmware?

I've got my first fully-assembled GH36 built.  No matter which version of Easy Keymapper I use, R5C3 doesn't work.  I programmed it as an X, just under the S (R4C3) of my WASD cluster.

1. When I press X I get nothing. (Press and release, press and hold, anything.)
2. When I press and hold S, I get nothing.
3. When I release the S key, there is a pause, then a string of repeating XXXXXXXXXXXXXX until I hit Escape (on the GH36 or my keyboard.)

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

I've tested the switch, jumped the positions, checked continuity to the pins, jumped the pins, everything I can think of.

So is it me?  Did I mess something up or damage something?  Is it the PCB?  The one I used or the design?  Is it the firmware?  Is there some bug?

Everything else works great.  Has anyone else gotten this far?

 - Ron | samwisekoi

p.s. I am using 20140827 and 20140902.  GH36, single pad option.

[berserkfan]

I am following this thread with mouth open wide.

Awesome work guys! I am inspired to do more of my own modding thanks to you all.

[metalliqaz]

Was it me, the PCB or the firmware?

I've got my first fully-assembled GH36 built.  No matter which version of Easy Keymapper I use, R5C3 doesn't work.  I programmed it as an X, just under the S (R4C3) of my WASD cluster.

1. When I press X I get nothing. (Press and release, press and hold, anything.)
2. When I press and hold S, I get nothing.
3. When I release the S key, there is a pause, then a string of repeating XXXXXXXXXXXXXX until I hit Escape (on the GH36 or my keyboard.)

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

I've tested the switch, jumped the positions, checked continuity to the pins, jumped the pins, everything I can think of.

So is it me?  Did I mess something up or damage something?  Is it the PCB?  The one I used or the design?  Is it the firmware?  Is there some bug?

Everything else works great.  Has anyone else gotten this far?

 - Ron | samwisekoi

p.s. I am using 20140827 and 20140902.  GH36, single pad option.

Let me just make sure I'm understanding the issue:

- R5C3 is completely dead.
- R4C3 is broken, but manages to activate R5C3 after releasing the key
- the other 34 keys work correctly

If that's accurate, that is a fairly strong indication that the firmware isn't at fault.  If you have time this weekend Samwisekoi, I'd like to quickly review the PCB design.  I'm worried about missing pull-down.

[dorkvader]

I got it figured out.

on teh "teensy this side" side of the PCB, you can see the trace for r5c3 (upper left) contacts the lower left trace for r4c3, which also connects to the rows via the diode. I will snip this trace and wire to the "r5" expansion header or teensy pin directly.

Have to go now but can provide picture if needbe

edit: gott run but the error is in the purple circle on teh "s" key, green trace should not be connected there

here's the source image I modded (thanks ron) http://geekhack.org/index.php?action=dlattach;topic=61306.0;attach=74047;image

I only had a quick look i might be wrong, but that didn't look right wne I checked the PCB.

[samwisekoi]

Thanks DV!

Fix implemented and confirmed to solve the problem.  Two cuts and one easy jump are required.  I'll document them further in a later post.

(At work now.)

 - Ron | samwisekoi

[metalliqaz]

Eagle eyes 

[dorkvader]

Thanks DV!

Fix implemented and confirmed to solve the problem.  Two cuts and one easy jump are required.  I'll document them further in a later post.

(At work now.)

 - Ron | samwisekoi

Now that GH is back up, can you elaborate more?

I see where 2 cuts can be made, but I was planning on making the jump to the column pin on the expansion header or the teensy pin directly, or do you solder to the via? (Or scrape the soldermask off the trace and jumper it directly?)

I'm going to have to actually program the teensy I have on mine and test out all the switch locations. Then maybe I'll dig up some switches and keycaps from somewhere and solder them for actual use.

[samwisekoi]

I tried to post hi-res pics and everything!

I made TWO cuts, above and below the pad that trace crosses.  Then I jumped from the other row pad on that switch (remember the switches are mirrored) to the R5 expansion pad.

This morning I realized I could have just jumped it to the unused R5 pad on the next column.

But what I did worked great. The shorter jump would have been prettier.

In other news, I took the time to update the layout with fixes for ALL the issues found so far.  That is done, and I am now running LED traces to all the switches from Q1 (formerly NUM).  When that is done it will be PROTO_2.

 - Ron | samwisekoi

[dorkvader]

This morning I realized I could have just jumped it to the unused R5 pad on the next column.

But what I did worked great. The shorter jump would have been prettier.

Hmm, I don't think that would work. This is the one connection to the "row5" trace that goes horizontally from the teensy to the r5 expansion pad.

In other news, I took the time to update the layout with fixes for ALL the issues found so far.  That is done, and I am now running LED traces to all the switches from Q1 (formerly NUM).  When that is done it will be PROTO_2.

if you post a picture I'll look at it! This sounds like a great layout already.

[samwisekoi]

Revision of PROTO_1.  Officially v.140808c3 UNRELEASED.

300% of actual size of 5" x 5".

 - Ron | samwisekoi

[samwisekoi]

I cut the traces at points #1 and #2.

I soldered a wire between the pads at points #3 and #4.

Worked a treat.

 - Ron | samwisekoi

[dorkvader]

Revision of PROTO_1.  Officially v.140808c3 UNRELEASED.

300% of actual size of 5" x 5".

 - Ron | samwisekoi

I have a request: can we space the transistor pads further apart a bit? There's still room to solder it normally and bend it over. Just a thought.

I did the same with my prototype PCB https://farm4.staticflickr.com/3921/15137414926_2a7930224f_o.jpg

[samwisekoi]

Yes.  I hate those little legs.  2.54mm apart they shall be!

 - Ron | samwisekoi

[dorkvader]

Yes.  I hate those little legs.  2.54mm apart they shall be!

That's almost like 1/10".

[samwisekoi]

Yes.  I hate those little legs.  2.54mm apart they shall be!

That's almost like 1/10".

Or even 0.100".

[Hak Foo]

I recall hearing that in the Cold War era, Western electronics were built with 0.1 inch pin spacing, but Soviet ones were built with 2.5mm spacing.  I wonder if after the markets opened up, you had a bunch of people very delicately bending pins to try to force imported parts to fit.

[samwisekoi]

I recall hearing that in the Cold War era, Western electronics were built with 0.1 inch pin spacing, but Soviet ones were built with 2.5mm spacing.  I wonder if after the markets opened up, you had a bunch of people very delicately bending pins to try to force imported parts to fit.

Well, while we're off in the weeds, I think the specification for Cherry MX key spacing is physically 0.750" (Imperial) but it is listed as 19.05mm, and people who use the metric measurements have slight problems over long keyswitch runs.

<References available somewhere in the vastness of geekhack.org.>

 - Ron | samwisekoi

[dorkvader]

Well, while we're off in the weeds, I think the specification for Cherry MX key spacing is physically 0.750" (Imperial) but it is listed as 19.05mm, and people who use the metric measurements have slight problems over long keyswitch runs.

<References available somewhere in the vastness of geekhack.org.>

 - Ron | samwisekoi

42% dvorak thread guy mentioned it
http://geekhack.org/index.php?topic=61141.0

page 2 i believe.

[metalliqaz]

I've always used 0.750 inches

[berserkfan]

Smwisekoi

May I ask if it is possible to add new keys to this keyboard by adding hand wired modules?

You see, 6x6 isn't exactly what I want. But it is close. I basically want to add thumb keys and extra function keys, so I am looking at this from the ebay seller wonderco:

http://www.ebay.com/itm/1-pc-White-Keyboard-2x5-keys-10-keys-Metal-Panel-/130303141925?pt=PCA_Mice_Trackballs&hash=item1e56ac2825

Am thinking of getting two GH36s, then adding 20 keys per side to get a total of 112 for my matrix layout ergodox. I'll have to hand wire the extra modules of course.

Do you think it is possible to add the 20 keys on each side to the teensy controller, or am I supposed to get even more teensies to handle the extra keys? i think running my setup on 4 teensies would be both expensive and awkward.

[jdcarpe]

Smwisekoi

May I ask if it is possible to add new keys to this keyboard by adding hand wired modules?

You see, 6x6 isn't exactly what I want. But it is close. I basically want to add thumb keys and extra function keys, so I am looking at this from the ebay seller wonderco:

http://www.ebay.com/itm/1-pc-White-Keyboard-2x5-keys-10-keys-Metal-Panel-/130303141925?pt=PCA_Mice_Trackballs&hash=item1e56ac2825

Am thinking of getting two GH36s, then adding 20 keys per side to get a total of 112 for my matrix layout ergodox. I'll have to hand wire the extra modules of course.

Do you think it is possible to add the 20 keys on each side to the teensy controller, or am I supposed to get even more teensies to handle the extra keys? i think running my setup on 4 teensies would be both expensive and awkward.

You would just need to wire it up in a matrix and connect the columns and rows to the appropriate expansion pins on the side of the GH36 PCB. You would be limited by the number of rows and columns supported by the controller, which in this case is 6 rows and 12 columns.

[berserkfan]

Smwisekoi

May I ask if it is possible to add new keys to this keyboard by adding hand wired modules?

You see, 6x6 isn't exactly what I want. But it is close. I basically want to add thumb keys and extra function keys, so I am looking at this from the ebay seller wonderco:

http://www.ebay.com/itm/1-pc-White-Keyboard-2x5-keys-10-keys-Metal-Panel-/130303141925?pt=PCA_Mice_Trackballs&hash=item1e56ac2825

Am thinking of getting two GH36s, then adding 20 keys per side to get a total of 112 for my matrix layout ergodox. I'll have to hand wire the extra modules of course.

Do you think it is possible to add the 20 keys on each side to the teensy controller, or am I supposed to get even more teensies to handle the extra keys? i think running my setup on 4 teensies would be both expensive and awkward.

You would just need to wire it up in a matrix and connect the columns and rows to the appropriate expansion pins on the side of the GH36 PCB. You would be limited by the number of rows and columns supported by the controller, which in this case is 6 rows and 12 columns.

This sounds good. So I will need 1 teensy 2.0 per side and a total of 2 free usb ports on my computer right? I might just buy from the ebay seller first to prepare.

[samwisekoi]

Smwisekoi

May I ask if it is possible to add new keys to this keyboard by adding hand wired modules?

You see, 6x6 isn't exactly what I want. But it is close. I basically want to add thumb keys and extra function keys, so I am looking at this from the ebay seller wonderco:

http://www.ebay.com/itm/1-pc-White-Keyboard-2x5-keys-10-keys-Metal-Panel-/130303141925?pt=PCA_Mice_Trackballs&hash=item1e56ac2825

Am thinking of getting two GH36s, then adding 20 keys per side to get a total of 112 for my matrix layout ergodox. I'll have to hand wire the extra modules of course.

Do you think it is possible to add the 20 keys on each side to the teensy controller, or am I supposed to get even more teensies to handle the extra keys? i think running my setup on 4 teensies would be both expensive and awkward.

You would just need to wire it up in a matrix and connect the columns and rows to the appropriate expansion pins on the side of the GH36 PCB. You would be limited by the number of rows and columns supported by the controller, which in this case is 6 rows and 12 columns.

This sounds good. So I will need 1 teensy 2.0 per side and a total of 2 free usb ports on my computer right? I might just buy from the ebay seller first to prepare.

Can you post a mock-up of your desired layout?

Also, what LEDs will you need/want?

 - Ron | samwisekoi

[samwisekoi]

Ok, attached is a high-res image of the second edition of the PCB layout.  This has all-new trace routing and passive locations.  I added many new features as well, including:

 - Full backilghting in addition to WASD backlighting.
 - Selectable NUM and CAPS LEDs (jumper required to choose NUM/CAPS or LED1 circuit
 - Optional 1.50x, gap or original flavor top row.  (Could be side row in rotated version.)
 - Consistent diode orientation plus marking on both sides and square pads.
 - Relocated side and center mounting holes for use in cut configurations.
 - I believe all passives are outside of stabilizer and switch footprints.

Expansion ports not yet wired, and some unused horizontal traces remain for that purpose.  Same with the traces running into the borders.  None of these should be shorted, but I may use them for the expansion pads.  Please report shorts.

But other than that, this is what I think the Beta board will look like.

Please let me know anything you see that looks wrong!

Thanks!

 - Ron | samwisekoi

[moemoe666]

awesome work
worth waiting and also want a keycap set signature for this )

[samwisekoi]

Hmmm.  Looking at that circuit after a few days away, it seems like the CAPS LED should possibly be on the other edge of the PCB, so that it is in the correct spot when the PCB is reversed.

Anyone have thoughts on that?  Prototypers?

 - Ron | samwisekoi

[dorkvader]

Hmmm.  Looking at that circuit after a few days away, it seems like the CAPS LED should possibly be on the other edge of the PCB, so that it is in the correct spot when the PCB is reversed.

Anyone have thoughts on that?  Prototypers?

 - Ron | samwisekoi

I think it's most important for it to be in the correct spot for the split KB case,

But I don't use lock lights, so I plan to just wire it to the LED matrix, were I to get one.

[domoaligato]

Hmmm.  Looking at that circuit after a few days away, it seems like the CAPS LED should possibly be on the other edge of the PCB, so that it is in the correct spot when the PCB is reversed.

Anyone have thoughts on that?  Prototypers?

 - Ron | samwisekoi

I think it's most important for it to be in the correct spot for the split KB case,

But I don't use lock lights, so I plan to just wire it to the LED matrix, were I to get one.

+1

if all of the leds are part of a matrix we canbdeside individual functionality however we want right?

[dorkvader]

if all of the leds are part of a matrix we canbdeside individual functionality however we want right?

Not as configured.

The "lock" LEDs (capslock, numlock, layer-lock, whatever) need to be individually controllable by the microcontroller. We don't have enough pins to address each LED individually, so only specific LEDs (or LEDs in specific locations rather) are individually addressable: the rest will turn on/off all as one unit. If you want your numlk LED to light up when you press numlk, then it needs to be separate.

Now you can wire up a whole LED matrix to be individually addressable, and turn off any/all LEDs at will, and make any LED the numlk or capslock LED, but this will require some specific LED driver chips and/or a lot of different hardware and work.

Right now each potential "lock-LED" location can be configured to be a lock-led or just a normal LED and turn on with the rest. To do this with all LEDs (and make any one a lock LED) with the PCB design would require a huge amount of traces and a jumper for each LED. (and likely some other jumpers as the PCB design would likely be too complex for a 2-layer board, etc etc)

You can make any LED location be your lock LED, but it would require you to cut traces and direct-wire the LED to the correct pins. No fun but workable.

[GenKaan]

Been wanting to make a hitbox for fighting games thats almost small enough to have in your pocket, would it be possible to make a custom layout like a "hitbox"?

Something like this:


Im playing on PC so it dose not matter really what PCB thats being used. Only need 12 buttons + what ever for lighting

[samwisekoi]

Been wanting to make a hitbox for fighting games thats almost small enough to have in your pocket, would it be possible to make a custom layout like a "hitbox"?

Something like this:

Show Image

Im playing on PC so it dose not matter really what PCB thats being used. Only need 12 buttons + what ever for lighting :)

The grid is square, so other than the stagger, yes.  Also it can only be six columns wide.

Is that for two hands?

 - Ron | samwisekoi