My custom unique looking "Rapture" keyboard build log

[dj027x]

So, I'm a fan of mechanical boards. (even though I currently use a microsoft sidewinder despite owning an apple extended2, dell at101w(my favorite) and an nmb) Im also a fan of modularity, and embarking on home projects. I just finished a replica of Milton's red stapler from 'Office Space', so it's time for a new project; A very unusual, unique, customizeable mechanical keyboard. Heres some the concept design i threw together with gimp:

I rejected tradition.
Instead, I chose something different.
I chose the impossible...
I chose Rapture;

More

The keyboard where the artist would not be bound by single colour leds,

More

Where the scientist would not be bound by petty non-programmable keys.

More

And with the sweat of your brow, it can become your keyboard too.

Yes, it looks unusual, with the grid layout, 1x1 windowed keys, and rather tiny (smaller than the poker 2) but let me explain why I chose these features (and feel free to leave constructive or destructive criticism, your choice  (seriously, I'd like to know how ugly you think it looks)):
-Teensy 3.1 microcontroller, seemed like the obvious choice for documentation, cost, ease of use, and functionality. I may switch to a de0-nano fpga if I need more jiggawatts later on.
-Programmable keys, made possible by the uC, so you can switch modes to get arrow keys, symbols, alternate layouts, macros etc.
-NKRO... why not? Its easier to make and simpler to understand than non-nkro as far as I'm concerned
-Its a keyboard, not a typewriter; why should I use staggered keys when ortholinear is more ergonomic?(this one's subjective)
-cherry mx clears cuz I like em. (I was considering whites just cuz they're more rare lol, and decided against matias cuz caps don't exist, though I do love the feel of alps)
-Blank keycaps, because A: screw qwerty and B: lends itself to fully programmable keys, and looks pretty fly.
-Small, for portability obviously
-Windowed keys, because full transparrents just look kinda cheap to me, and it definitely gives it a very unique look.
-Bioshock references, because I'm a geek, the quote seems relevant to choosing how you want your keyboard and Andrew Ryan is my hero, behind Nikola Tesla and Elon Musk. (Ryan is the fictional Elon! Or is Elon Comstock?  anyway...)
-INDIVIDUAL PER-KEY PROGRAMMABLE RGB BACKLIGHTING!!! Yep, the keys will be able to be programmed to respond to your touch, or give you beautiful colourful feedback. (heck, I intend to use them to display the memory contents of the microcontroller so you can literally use this board to program itself!!(yea, in machine code, but what do you expect?; I'm a comp eng  ))
There are definitely some hurdles to overcome, but that's what makes it fun =)
Glass ceiling: 200$

So what have I done so far? Nothing  Because I'm waiting on my parts to arrive 
The board has 48 keys (subject to change) but I ordered 50 clears, and 52 keycaps incase of defects. So I'm currently waiting on those to arrive so I can determine whether I will be able to pull off RGB backlights. MX switches only physically support 3mm leds, but afaik, 3mm rgb leds don't exist. So if I can modify them to accept 5mm rgbs, or square leds, the projects is a go. I will then order probably 100 leds, and begin the pcb design. I intend to have a multiplexed matrix design, with the teensy reading 6 keys per scan, and 8 scans per cycle; 6*8=48; nice and simple. I have the means to print, etch, drill, and solder the board at home, so that's what I will be doing. The led circuit will be a tad more complicated. Regarding the leds, if I go analogue, Ill need a resistor ladder, hefty current source, and 3 capacitors per led (probably not possible for my target dimensions) So it looks like I'm going the digital route: I would like to use the teensy's multiple built in pwm pins to drive the leds without sacrificing any clock cycles from key detection, but I'd still need 3 caps per led, and there are other logistics preventing that from working. (plus caps are a common point of failure in electronics). Afaik, my best option is getting some TTL ICs, and yes, I'll need 48*3=144 elements. (the FPGA route is looking more and more attractive). I should be able to get away with 144 extremely simple D-Latches in the form of 36 16-pin DIP ICs and still keep it tiny. (Unless I'm crafty with my design, this could end up requiring a multi-layer PCB solution). Phew, once that's all said and done... add joysticks? haha I toyed with that idea, but I'm content with my already complicated design. Maybe in the future. As for the lettering on the keycaps... Laser? (Don't yell at me, but they're also ABS :/) The final step (for assembly at least) would be buying or fabbing a slick case to hold it all together; Should be pretty simple. Once assembled just plug-and-play, err... program... the entire controller from scratch. (Hope I havn't wired anything wrong at this point!) Yup It's ambitious; I know. But I have to go through with it now that Iv'e put money into switches and caps. In for a penny, in for a pound! I'll make any practical or necessary adjustments along the way ("Eventually we adapted it from invisibility to just not being seen" lol) and I'll of course leave room for future improvement, and document every step for you! If there is enough interest I will design a (relatively) easy to use software interface; free, as in 'beer' AND 'freedom'!. BTW, keycaps are from SP, but it looks like they no-longer sell the packs that I bought.

[howthegodschill]

Damn, this is tight. I'm excited to see how this comes out in the end.

[CPTBadAss]

Please take pictures of the PCB etching process

[dj027x]

Please take pictures of the PCB etching process

Will do! I'm going to use the toner transfer method if you care to know. Possibly 2 separate boards with wire interconnects, depending on what size board I can get a hold of

[Findecanor]

Milton's red stapler huh ... 646 or 747?

There are people who have used RGB LEDs with Cherry MX, bent inwards under the keycap.
The ones in the Corsair keyboards have transparent casings with SMD:d LEDs underneath and they have been exclusive to Corsair during 2014 and I think they should become available soon... You could ask 7bit about them: he is often selling switches and has planned to sell the clear-cased ones. I don't think Cherry makes clear-cased MX Clear though...

As to driving the LEDs, there are a bunch of people everywhere who have made LED matrices run by an Arduino and a bunch of shift registers, but there are dedicated LED matrix controller chips out there that can drive matrices with PWM for 256 shades of grey scale.
The firmware for the Red Scarf II is made for an AVR µC (as in Arduino) and a shift register.

As a starting point for using the Teensy 3.0, you could use HaaTa's Kiibohd firmware. It supports it, the AVR in Teensy 2.0 and a few others microcontrollers.

[DrHubblePhD]

We need to talk, I am a HUGE bioshock fan!

[vvp]

I'm going to use the toner transfer method if you care to know.

We used it for Katy. It works without a problem with min track width 0.3mm and min clearance 0.25mm, via Φ1mm and via drill Φ0.5mm. Somewhat smaller features should be possible too. I think it is easier than trying to build circuits with jumper wires on a breadboard. No breadboards for me except for the very simple stuff.
Here is our first attempt at it: http://deskthority.net/workshop-f7/katy-keyboard-or-k80cs-key80-contoured-split-t8524.html#p191612
If you did not do it yet I have two good rules:

• If the toner print on the PCB does not look good do not even try to etch it. Clean it with acetone and transfer toner on the PCB again. It does not take long.
• Do not leave it in the etching bath for too long. Regularly check whether it is not done yet.

A positive photoresist is already bought if there would be need for finer stuff

[dorkvader]

Milton's red stapler huh ... 646 or 747?

It was their 747 in "rio red" although swingline did not make it at the time. The prop used in the film was the normal one that had been painted.

[dj027x]

Milton's red stapler huh ... 646 or 747?

646 of course!

It didn't turn out perfect; There's slight orange peel, and drip lines that the camera angle hides, and I had to cut out the damn logo by hand with an exacto knife, but it was literally the first time I ever did a paint-involved project, so I'm happy with the result considering my low skill level. It was very tedious and time consuming, but a welcomed experience.

There are people who have used RGB LEDs with Cherry MX, bent inwards under the keycap.
The ones in the Corsair keyboards have transparent casings with SMD:d LEDs underneath and they have been exclusive to Corsair during 2014 and I think they should become available soon... You could ask 7bit about them: he is often selling switches and has planned to sell the clear-cased ones. I don't think Cherry makes clear-cased MX Clear though...

My switches just arrived (im actually about to post an update here) and it looks like 5mm leds might just fit; I'll have to wait for the caps to get here to be sure though. I was definitely interested in clear body switches, but the only ones i found were matias, and I had no luck at all finding matias/alps compatible keycaps.

As to driving the LEDs, there are a bunch of people everywhere who have made LED matrices run by an Arduino and a bunch of shift registers, but there are dedicated LED matrix controller chips out there that can drive matrices with PWM for 256 shades of grey scale.
The firmware for the Red Scarf II is made for an AVR µC (as in Arduino) and a shift register.

As a starting point for using the Teensy 3.0, you could use HaaTa's Kiibohd firmware. It supports it, the AVR in Teensy 2.0 and a few others microcontrollers.

The matrix controller chip seems like a good idea; I'll look into that tonight.
The one part I'm not particularly looking forward to is programming the usb protocol for the teensy(if it doesn't have a pre-made one. I haven't looked into it much yet), so I'd totally be into using some already made firmwares as a baseboard. Thanks for the info!
BTW, what lube would you recommend for the switches? I see in your signature you've got some lubed clears

[dj027x]

Milton's red stapler huh ... 646 or 747?

It was their 747 in "rio red" although swingline did not make it at the time. The prop used in the film was the normal one that had been painted.

It's unfortunate that the production stapler looks nothing like the movie prop lol, not sure what swingline was thinking there. Then again, if it did look like the prop, I would have missed out on the fun of making it myself.

[dj027x]

We need to talk, I am a HUGE bioshock fan!

Glad to hear it! My love for it comes from Bioshock more-so than Bioshock 2 or Infinite (Although they were both still good, and the end of infinite figuratively blew my mind haha)
Have you by any chance played System Shock 1 &/| 2? I believe it's where Bioshock got its inspiration from, so I got a few minutes in, but college work restricted me from sticking with it :/
(Though I may start back up again soon!) It's a pretty cool game if your into that old pixely stuff.

[DrHubblePhD]

We need to talk, I am a HUGE bioshock fan!

Glad to hear it! My love for it comes from Bioshock more-so than Bioshock 2 or Infinite (Although they were both still good, and the end of infinite figuratively blew my mind haha)
Have you by any chance played System Shock 1 &/| 2? I believe it's where Bioshock got its inspiration from, so I got a few minutes in, but college work restricted me from sticking with it :/
(Though I may start back up again soon!) It's a pretty cool game if your into that old pixely stuff.

I dont blame you for not liking bioshock 2, I usually tell people to skip it, but infinite was outstanding IMO. Especially if you play through burial at see which really closes up the series perfectly.

I actually haven't played system shock yet, I usually am not into that style, you know the huge inventory RPG like games. Although im sure its different from what I am imagining.

[dj027x]

We used it for Katy. It works without a problem with min track width 0.3mm and min clearance 0.25mm, via Φ1mm and via drill Φ0.5mm. Somewhat smaller features should be possible too. I think it is easier than trying to build circuits with jumper wires on a breadboard. No breadboards for me except for the very simple stuff.
Here is our first attempt at it: http://deskthority.net/workshop-f7/katy-keyboard-or-k80cs-key80-contoured-split-t8524.html#p191612
If you did not do it yet I have two good rules:

• If the toner print on the PCB does not look good do not even try to etch it. Clean it with acetone and transfer toner on the PCB again. It does not take long.
• Do not leave it in the etching bath for too long. Regularly check whether it is not done yet.

A positive photoresist is already bought if there would be need for finer stuff

Katy looks stellar, man! I hadn't considered a floating matrix like that. I am now though Yep I've used this method before, but I was using .762mm traces. I couldn't imagine going much smaller haha. I bought a micro drill set for my dremel tool for this type of thing, and its _just_ small enough for a .7mm id via. I use isopropyl alcohol to dissolve the paper after the hot water bath, then gently rub some etchant into it, and give it an acetone rubbing to get the toner off. I _might_ consider applying a solder mask this time, though I don't suspect I'll have any reason to do so. I bought a brother HL-L236ODW laser printer specifically for pcb making, and it has the option for extra heavy printing, so the print is usually pretty good, although the resolution could be better. .7mm traces is about the limit of quality for it. I believe I recall the photoresist method being slightly more expensive, which is why I went toner transfer, "if I'm correctly mistaken".

[dj027x]

I dont blame you for not liking bioshock 2, I usually tell people to skip it, but infinite was outstanding IMO. Especially if you play through burial at see which really closes up the series perfectly.

I actually haven't played system shock yet, I usually am not into that style, you know the huge inventory RPG like games. Although im sure its different from what I am imagining.

ARGH! I really need to play burial at sea. I got about 2 minutes into it before I got busy with other work. (I really should play all my steam games instead of only buying them. ive played roughly 20 out of 170 haha)
"Sale you say? Yea, I might play that someday... And I sure don't want to lose money on this... Better buy it now"

[dj027x]

Seem to be having some trouble posting my update. When I submit it, it takes me to a "create a new topic" page. hmmm. It has 32MB worth of pics. Would that do it?

[dj027x]

SORRY about this update being broken into several posts. It didn't let me upload more than 3 pics per post :/
Update 0: That's right, I zero index things.
The switches have arrived!

My pretties:

More

I'm a tad disappointed with the packaging; they came in this plastic bag, wrapped in brown paper. Some of the pins were bent upon arrival. I guess it keeps cost down, and for the most part, one gas to simply bend the leads back into place and all is well. Oh yea, my updates might have lots of pictures...

Muahahaha

More

So far, my 'obtained parts list' includes the teensy and switches. Still waiting on the caps to get here.
Here an hd pic of the teensy, cuz why not? (A big thanks to the geekhack admins for hosting this fantastic site btw. It is very much appreciated!)

More

[dj027x]

So the switches are actually a lot smaller than I imagined they would be. I guess the camera adds 5 grams?

More

More

Tiny buggers

[dj027x]

So it looks like a pretty tiny hole for the led there. There is a cylindrical slot to house the led body, and a rectangular cavity that leads downward to 2 tiny holes where the led pins com out the bottom.
I tried to get some decent shots of the front and back here. Something interesting to note: There are 4 holes on the back side, but the outer 2 are walled off from the inner 2, and I am stumped as to what the outer 2's purpose might be. I suspect cherry designed this body with the hopes that 3mm rgb leds might be developed soon, but why wall off the holes? Structure? Modding restrictions? In any case, that's only a guess as to their purpose.

More

More

[dj027x]

Here is the hole compared to a 5mm led

More

As-is, a 5mm led doesn't quite fit in there

More

[dj027x]

Popped it open to find some pretty mechanics, and I could actually see the lube between the stem and metal; nice quality.
WARNING: nudes

More

With the top off, it appears that the led can actually fit in there without affecting the operation of the switch. It will require some of the body be cut away though.

More

[dj027x]

I would have to cut off the south (when looking down at it) part of the top half of the body. This would allow the led to fit without opening up the wall that houses the mechanics, but it would remove a clip that is meant to hold the body to a pcb. I don't think this will cause an issue though.

More

Just for kicks, I threw a 3mm led I had lying around in there. Even if a 3mm has a flange, it doesn't sit quite right.

More

So, I'm just waiting for dem caps to arrive, and I can determine if I'll be able to get the 5mm leds in there without keycap collision. I love the way they feel too. The stems have a bit of wobble, but nothing I didn't expect. Another less picture-oriented update will be coming once my caps get here.
Also, I should actually be able to mill the case out of the acrylic found in flatscreen monitors. Ill probably use some acetone and dye to paint it. Probably guna have to buy the nuts and bolts to hold it all together though.
Sorry about this update being split into several posts. It wouldn't let me upload more than 3 pics in a post :/

[Findecanor]

646 of course!

Nice and glossy.   I made mine years ago from a 646 and decals I had bought from a guy on the Replica Prop Forum.

BTW, what lube would you recommend for the switches? I see in your signature you've got some lubed clears

I used RO-59KT. But after two years I am not so sure that the keys are so smooth as they once were.

With the top off, it appears that the led can actually fit in there without affecting the operation of the switch. It will require some of the body be cut away though.

You may also be able to cut away some from the LED without affecting how it works, as long as you don't cut down to the metal or silicon. Tedious work though.. 

BTW. Bi-colour LEDs exist in 3 mm. They are either three-pin (one common lead) or two-pin in different directions. Ducky has made a few keyboards with bi-colour LEDs, but I dunno which variant they use.

[vvp]

I've used this method before, but I was using .762mm traces. I couldn't imagine going much smaller haha. I bought a micro drill set for my dremel tool for this type of thing, and its _just_ small enough for a .7mm id via. I use isopropyl alcohol to dissolve the paper after the hot water bath, then gently rub some etchant into it, and give it an acetone rubbing to get the toner off. I _might_ consider applying a solder mask this time, though I don't suspect I'll have any reason to do so. I bought a brother HL-L236ODW laser printer specifically for pcb making, and it has the option for extra heavy printing, so the print is usually pretty good, although the resolution could be better. .7mm traces is about the limit of quality for it. I believe I recall the photoresist method being slightly more expensive, which is why I went toner transfer, "if I'm correctly mistaken".

You definitely can go down to min track width 0.3mm and min clearance 0.25mm, via Φ1mm and via drill Φ0.5mm. This is what we use with a 1200 dpi laser printer.  But I believe 600 dpi would be enough too. I believe it may be possible to go down to 0.2mm traces and 0.15 clearance but I did not try it.

As for as the drills. You can buy thin drills which have a thick stem. Here are some examples: http://www.proxxon.com/en/micromot/tools_drills.php?list
Although, it is good to have a drill press so that you do not break the drills.

The whole magic is: If the toner transferred nicely to the PCB. I.e. it sticks and did not deform much then it will etch nicely too (provided you will not leave it in the etching bath for too long).

We tried to print on an expensive glossy photo paper and transfer from that to PCB. That did not work well. A thin glossy paper from an old electronic parts catalog works very well. After we transferred toner from paper to PCB we put it to hot water bath (about 40°C), leave it there for a few minutes and then used an old tooth brush to rub the paper from the PCB. It is also important not to contaminate the paper or the PCB with oils before transferring (i.e. do not touch the surfaces).

I believe solder resist is needed only for really small feature sizes. No need for the stuff toner transfer method can produce.

Yes, photo resist is a more expensive and, in the case you are making the photo resist coating yourself, also more work. Although in that case it is not that much expensive on average.

Photo resist method can do smaller features. It is more consistent. Not my experience yet but that is what the tutorials claim :-)

[dj027x]

We tried to print on an expensive glossy photo paper and transfer from that to PCB. That did not work well. A thin glossy paper from an old electronic parts catalog works very well.

Did you have to find a blank page? I would expect whatever was printed on the page to be transferred as well. I was going to try some sticker backing paper next time, but perhaps I'll try some catalog paper.

[dj027x]

Keycaps just arrived! I'll do another update once I get home, and I'm going to spend my week in florida designing the pcb. I'll throw my design on here before finalizing anything so you guys can critique it.

[dorkvader]

Milton's red stapler huh ... 646 or 747?

It was their 747 in "rio red" although swingline did not make it at the time. The prop used in the film was the normal one that had been painted.

It's unfortunate that the production stapler looks nothing like the movie prop lol, not sure what swingline was thinking there. Then again, if it did look like the prop, I would have missed out on the fun of making it myself.

The 646 is their plastic version, a "lower end" of the 747. It makes sense to only do a "limited edition" (or special edition / colour whatever) on the higher end product.

I was planning on painting my vintage swingline model 31 (ca 1932) red and see if anyone gets it. I used to have a boston stapler but they tend to jam up too much.

Please let us know about the results of toner transfer. I read all I can about this before I try it myself to maximize my chances of success.

[dj027x]

I read all I can about this before I try it myself to maximize my chances of success.

That's always a good approach. I like to do a lot of hands on experimenting as well. Small scale testing is good (practice makes 'good enough').  The first time I did it was it an EE class where I decided not to send a board design to be fabbed like everyone else in the class, but to make my own. I had one shot to get it right or fail the class haha. My board ended up being like 8x smaller than everyone elses, had a ground plane, and pads on the sides to connect alligator clips. I did really well for my first time ( story of my life ) In fact, heres a pic:

[dj027x]

Update 1 (In several parts, I'm afraid:
TLDR: good news is I'm awesome and got everything to fit! Bad news is I have to mod every switch case and led to make it work.

So the caps arrived! I had to buy in packages of four. They all look very well made. No defects spotted yet. They're abs, but they've got a nice rough texture. Plus I have 8 extras if any start to smooth out.

More

Here's what the layout is going to be like:

More

I even with the top half of the case off, no way in hell 5mm leds are fitting under the caps. The cap sits at normal height, but you can't fully press the key down.

More

[FrostyToast]

May I ask why you are using 5mm LEDs?
Those things are obviously way too oversized.
The largest you can go for LEDs would be 3mm.

[dj027x]

I ground down the led body, which actually made the led a bit diffuse due to the now-rough outside, which in the case of rgbs is a good thing.
I then took a chunk out of the top of the switch body, and was able to leave the clip on the bottom half of the body intact.

More

I could get the led a bit smaller, but I know the RGB leds are going to be a tad bigger anyway, so I left it as you see in the next pic. Slid into the modified case, it looks like it just might fit under the cap.

More

[dj027x]

May I ask why you are using 5mm LEDs?
Those things are obviously way too oversized.
The largest you can go for LEDs would be 3mm.

Unfortunately, RGB leds don't come in 3mm. 5mm are the smallest.
But thats why I'm grinding them down to nubs!

[vvp]

Did you have to find a blank page? I would expect whatever was printed on the page to be transferred as well. I was going to try some sticker backing paper next time, but perhaps I'll try some catalog paper.

No I do not care what is printed on the page. If you can use a blank page then it is a bit better since it is easier to assess whether it printed well. I'm only making sure that there are no wrinkles on the page (i.e. checking that the surface is smooth).

The old print on the page will not transfer to the PCB. Realize that toner is about half plastics and the rest is black color (possibly fine sooth). The plastics is what you are using as mask for etching. The original colors do not contain it. Mkawa (or somebody else) can probably even tell you what is exactly in black toner and what is in standard print-press inks.

[FrostyToast]

May I ask why you are using 5mm LEDs?
Those things are obviously way too oversized.
The largest you can go for LEDs would be 3mm.

Unfortunately, RGB leds don't come in 3mm. 5mm are the smallest.

I swear that Kaliet used some RGB LEDs on some of his boards...
Go chat him up and see what he used.

[dj027x]

More

*click* (yea they're non-clicky, but I'm trying to be dramatic here)

More

[dj027x]

I swear that Kaliet used some RGB LEDs on some of his boards...

I took a look at some of his pics. It looks like he's using multiple color leds per board, but each switch is still single color :/
Though one of his boards did use RGB LEDs on the bottom of the board, as a backlight
I shot him a message to double check.
I'll wait on his response before putting money into hundreds of leds haha.

[dj027x]

So now that I know it's possible, time to order the RGBs!
Regarding another aspect of the project, I pulled the plexiglass(or whatever material it is) out of a junked monitor. Probably going to double it up, mill it and paint it to use it as a the case

More

Ah yes; almost forgot to check that the led works after all that grinding. (never drive a led you intend to keep alive without current limitation. I only lit this long enough to snap a quick pic)

More

See you next update. (Its probably going to regard the pcb design)

[dj027x]

As for as the drills. You can buy thin drills which have a thick stem. Here are some examples: http://www.proxxon.com/en/micromot/tools_drills.php?list
Although, it is good to have a drill press so that you do not break the drills.

That's exactly what I need haha. I've been using bits that are like .7 mm diameter from end to end, and they tend to wobble as a result of not sitting centered in the chuck.
I'll definitely pick up a set; not from there though, its a tad pricey with the shipping.

[FrostyToast]

Well Sprit does have this video clearly showcasing 2x3x4 RGB LEDs.
No idea where he got them from.

https://www.youtube.com/watch?v=o2oAjCjCaUE

[dj027x]

Ah my apologies; I should have clarified, 3mm rgb leds do exist, but I'm looking for separate channels for the r, g, and b components. You can see in the vid those leds only have 2 leads, and they actually transition through the colors automagically. Afaik, there's no way to set them to a specific colour unfortunately The reason 3mm controllable leds don't exist is because they just can't fit 4 pins (one for each channel and one common annode or cathode) within 3mm. Or at least they wouldn't make a profit doing it. I could use smd leds suspended by wires I suppose, but I think that would be more of a pain than grinding down 5mm leds

[dj027x]

update 1_1: I just had 2 ideas, I feel like an idiot for not thinking of the first earlier:
Firstly, Why the hell am I messing around with the typical semi-capsule shaped 5mm leds when I can just use square leds?
I was actually looking around for other led options when I saw a picture of them and I wondered to myself why I haven't considered using them yet, and now I probably will.
I was able to find some 2x5x5 mm ones, which will DEFINITELY fit in the modded cases (so long as 1 of the dimensions is 3mm or less and another is 5mm or less). I may be able to find even smaller ones with a bit of searching.
(Thanks to FrostyToast for suggesting other options; I probably wouldn't have found this otherwise)
2nd idea: The window in the keycaps is not diffused; it's quite transparent, so I'm worried that the light from the led either wont propagate evenly, or it will be blinding to look at the keys, because you'd just be looking straight into the leds. Well if that is the case, I should actually be able to rotate the switches and caps, so the led and window are on the north side of the key. The transparent bit on the top of the keycap (which is actually a little diffused due to the rough texture on the top of the keycap) would still light up, and be visible to the typist, and the window on the side of the keycap would allow the led to cast a glow onto the keys behind (north of) it.
Might cause a cool effect; I'll toy with the idea later on.

[dj027x]

Just bought 100 2x5x5 rgb leds. for 22$
Unfortunately they will take 5 to 27 days to arrive lol.
In the mean time, I'll be working on the pcb design
Honestly, I'm not sure why nobody has done this before (that I know of).
Perhaps I'll run into some insurmountable roadblock here, but till then, onward!

[neverused]

Honestly, I'm not sure why nobody has done this before (that I know of).

Do you mean rgb backlight a matrix keyboard?

[dj027x]

Do you mean rgb backlight a matrix keyboard?

Even just per-key RGB backlighting. Isn't corsair's K70 and some razer thing the only ones ever to do it? And only just last year right?
This is definitely the first diy project I know of to do it

[dj027x]

Feel free to point out any error in my math.
I'm trying to determine the frequency of the signal the teensy will need to send to the leds to control them nicely.
Lets consider the board to just be a 12x4 screen.
Let's aim for 30Hz refresh rate, and 9-bit colour (3 bits per channel), for now.

Assuming the setup will consist of the teensy driving multiplexed d-latches, which in-turn drive one component of one led, driving the display would require 144 d-latches and 2 signals, both at:
(48 leds times 3 channels per led times 30Hz refresh rate times 2^3bit per channel=9 levels of intensity per channel = 48*3*30*9=) 38.88kHz

If I can miraculously find compact multiplexed 3-bit PWM ICs (I have my doubts that they even exist but one 18 pin IC could house 8 (might get hot)), I would be able to use 144 PWM drivers with 4 signals at:
(48 leds times 3 channels per led times 1 PWM driver per channel times 30Hz refresh rate = 40*3*30=) 3.7kHz

So there is a trade-off between complexity and speed (as there often is).

Now, the teensy 3.1 does have a 72MHz clock, but there's calculations it handles every cycle, so without actually testing it, there's no telling what freq I'll be able to drive leds at, although its safe to say it will be well below 72MHz. I'll investigate this further over the next few days, because the solution I determine here will affect my PCB design.

[dj027x]

Well I seem to have overlooked the fact that each led will also require 3 resistors. Because the leds are common anode, I will be able to use those fancy resistor ic's that have a common ground. No big deal, but the design will just require that much more realestate. This is definitely going to be a double sided pcb or multi-layered solution.

[dj027x]

School has started up again, so progress might become intermittent, however, the rgb leds have arrived!

More

The pictures may be of poor quality; I broke my phone's digitizer over break so I'm using an old phone to take the pictures.
Conveniently, the planck keyboard parts would work nicely for this project, however I don't like the way the keys protrude from the top of the planck.
I would like the bottom of my keycaps flush with the top of the case. I might be able to use the top plate for the planck though, to hold the switches and keep pressure off of the pcb.
Next step is modding each switch case to accept the rgb leds.
I'll update y'all when that's complete

[Evo_Spec]

School has started up again, so progress might become intermittent, however, the rgb leds have arrived!

More

(Attachment Link)
(Attachment Link)

The pictures may be of poor quality; I broke my phone's digitizer over break so I'm using an old phone to take the pictures.
Conveniently, the planck keyboard parts would work nicely for this project, however I don't like the way the keys protrude from the top of the planck.
I would like the bottom of my keycaps flush with the top of the case. I might be able to use the top plate for the planck though, to hold the switches and keep pressure off of the pcb.
Next step is modding each switch case to accept the rgb leds.
I'll update y'all when that's complete

Wow those are huge, i'm curious to see how you're going to fit it.

[dj027x]

Wow those are huge, i'm curious to see how you're going to fit it.

I've already tested that they fit. I'll throw up a picture when I get the first one assembled. Right now it's just a matter of drilling extra holes for the led leads.
I'm also looking into using kicad, so I should have a schematic relatively soon.

[dj027x]

leds seem to function nicely, and are decently bright even at low amps. They do however blow instantaniously when 9V is put across them, other like the typical 5mm red leds. I'm looking into various led matrix drivers at the moment.

[CPTBadAss]

Do those LEDs cause any interference with the switch actuation? And ugh, I need to learn more about this PCB etching process you and vvp keep talking about.

[dj027x]

Do those LEDs cause any interference with the switch actuation? And ugh, I need to learn more about this PCB etching process you and vvp keep talking about.

Iv'e been keeping an eye on this. As you may or may not be able to tell from the following pics, there is actually a relatively large amount of space between the led and keycap when bottomed out (1 to 2 mm).

More

 And fear not; I will document the pcb fabrication well =D
However, the good news is that they fit! And I didn't even have to modify the switch case any further than cutting that top bit as I explained before.
Bending the leads of the led to line up with the holes gives for good separation distance between the leads.
From here, its looking like I can make the keyboard matrix on the upper side of the pcb, and the led circuitry on the flipside.
Updated of a lighting test soon to come (gotta write some arduino code)

[DrHubblePhD]

Do those LEDs cause any interference with the switch actuation? And ugh, I need to learn more about this PCB etching process you and vvp keep talking about.

Iv'e been keeping an eye on this. As you may or may not be able to tell from the following pics, there is actually a relatively large amount of space between the led and keycap when bottomed out (1 to 2 mm).

More

(Attachment Link)
(Attachment Link)
(Attachment Link)

 And fear not; I will document the pcb fabrication well =D
However, the good news is that they fit! And I didn't even have to modify the switch case any further than cutting that top bit as I explained before.
Bending the leads of the led to line up with the holes gives for good separation distance between the leads.
From here, its looking like I can make the keyboard matrix on the upper side of the pcb, and the led circuitry on the flipside.
Updated of a lighting test soon to come (gotta write some arduino code)

That looks sick, interested if you can get everything fitting well!

[dj027x]

Hear Ye;
A video documenting my test-switch has been posted on youtube. My good phone is broken, so you get ****ty quality
(I realize I'm also not the most interesting person to listen to haha)

https://www.youtube.com/watch?v=UAS6XXlzCHo&feature=youtu.be

[dj027x]

One aspect that I dislike about pwm leds is that you tend to see the colors bleed when you move your eyes quickly. This is a little stressful for the eyes, but seeing as it would be impractical to use a dac for every component for every lead, let's hope I can get the pwm frequency high enough to fix this issue. One problem is that each of the three components of color in a single led must be activated in syncopation. That is; even during pwm, only one component can be lit at once. If two components are activated at once, current simply flows through whichever component has the slightly smaller drop-out voltage, and you only see one color.

[dj027x]

Diodes are in. I'm doing some research on suitable led matrix drivers. Ive found one that drives 16 leds. I should be able to use 9 of them to drive all channels of all of my leds

More

[dj027x]

So I have purchased the HT1632c led driver. It should be capable of handling all 144 signals on the board, controlled by a serial interface. so the teensy will scan for keypresses, send appropriate signals to pc, determine color of leds, and send a bitstream to the led driver to take care of it.

[dj027x]

Sorry for the delay. This driver is taking forever to ship from china once it's here, progress will pick up again.

[dj027x]

The little buggers are *finally* here!

More

Im going to just do a little testing to be sure the chip can light leds with sufficient brightness at a high frequency, then its on to pcb design!

[dj027x]

School and work have been retarding my progress, however it will pickup soon. I have bought some sets of gateron switches in the group buy. Id like to see how light shines through them, and if it looks pretty slick, I may rebuild this board with smd rgb leds instead. But I'm going to progress with the leds and switches I currently have for now

[Evo_Spec]

Not really sure what to say but i just wanted to leave a message here letting you know that i am keeping up with your updates!

look forward to seeing the progress.

[dj027x]

Not really sure what to say but i just wanted to leave a message here letting you know that i am keeping up with your updates!

look forward to seeing the progress.

Thanks! It's encouraging knowing others are interested in this

[stoic-lemon]

Yup, interesting project. Good luck with it and keep the updates coming.

[CPTBadAss]

Glad you're making some progress. Now let me see those etchings!

[dj027x]

Glad you're making some progress. Now let me see those etchings!

I'll post pics documenting the entire process as soon as I begin it. Although the amount of school work I have has been keeping me from making much progress on this.

[dj027x]

Just thought I'd let everybody know i haven't abandoned this project; I've been drowning in college work, but I should have some free time this weekend or next to make some solid progress

[berserkfan]

very happy to see more matrix layout fans.

sooner or later some one will decide to do a split layout with a matrix size I like... and then I'll pounce and probably make my last ever wallethack purchase!

[dj027x]

Now that school is out, I've been making slow but steady progress. I still hold a full-time job, so I have about 3 hours a day to work on this.
It turns out that the Holtek led drivers I got do not support RGB LEDs. RGB LEDs are fussy, in that you can only drive one color component at a time, otherwise if two or more colors are being driven simultaneously, all the current will flow through one path, you only see one color, and that used path is overpowered and more likely to burn out. To avoid this, one must syncopate the activation of each color component for each led. The teensy has 30 GPIO pins, which is _just_ enough to pull off using the teensy to drive the LEDs and monitor keystrokes with minimal external circuitry (i will need a simple 3 to 8 demux for monitoring keystrokes). Originally, I had hoped to offload the work of driving the LEDs to a dedicated chip, however I could not find one reasonably priced, or individually sold that suited this project, so I will have to either use a CPLD or FPGA to design my own, or let the teensy handle it. Due to the relatively beefy processor, I am pretty confident that this shouldn't be an issue, but only stress testing will tell. Right now, I am writing an extremely basic keyboard firmware for the teensy to use for testing to make sure the hardware and theory all works before I assemble the hardware and flesh out a killer firmware. While I had wanted to make this nkro, I have not been able to find good documentation on how to do this. It seems to be a limitation of USB. I have read that nkro can be done, but I have yet to see anybody explain how its done on an arduino or teensy. This week, I completed the extremely tedious task of setting up the LED matrix for further testing. I am going to hand write (or script) some test images and see how much the teensy can process in the background while still providing a steady image and smooth animation to the LED matrix. Here's a quick shot of the current setup. Not all the LEDs are hooked up to the teensy yet, but you get the idea.

More

[dj027x]

The full matrix in all of its colorful glory

More

[dj027x]

A video of a simple animation is up at https://youtu.be/y0u0rfIS4LY
I was unable to find a dedicated led driver that suits this project for a reasonable price, so I'm testing wether the teensy can handle driving the leds by itself. So long as you aren't running physics simulations on this board, it should be able to animate the leds quite nicely; even run simple games like tetris or conway's game of life, albeit on a 12x4 screen. Based on current performance, the teensy should be able to render pretty animations and monitor and process keystrokes simultaneously with no issues. I also had the idea of having it record every keystroke you make, so it can display essentially a heat-map of keys most often pressed, as a tool to help you design a custom layout. My hopes were to allow one use the board to program itself through an assembly code of sorts. The led's colour or flashing would represent a particular instruction or value in memory. One would then be able to tap the key, or use function keys to alter the values, thus altering the behaviour of the keyboard. Whether this would be most easily done by creating a functional language within the teensy, or writing my own microcontroller to an FPGA is yet to be determined.

[shaymdev]

My hopes were to allow one use the board to program itself through an assembly code of sorts. The led's colour or flashing would represent a particular instruction or value in memory. One would then be able to tap the key, or use function keys to alter the values, thus altering the behaviour of the keyboard. Whether this would be most easily done by creating a functional language within the teensy, or writing my own microcontroller to an FPGA is yet to be determined.

  you are hard core!

I'm not usually into LED lighting the keyboards...but this project is looking pretty awesome...and might be making me wish for something like this in my build.

[Evo_Spec]

awwwww yeahhhh, rainbow!!!!

[dj027x]

I believe the slight blinking that can be seen when running the rainbow effect can be easily mitigated by using small approximate look-up tables for trig functions instead of calculating values on the fly. Now, I just need a multiplexer to monitor the keystrokes. I am going to _finally_ begin designing the pcb, and keep you updated along the way. I have decided to keep it 6kro for now to allow the keyboard to work with bios. I will likely add a switching feature to this in the future to allow the user to change from 6kro to nkro. Some amount of hackery is needed to actually get the teensy to nkro, but im up for it, when the time comes.

[dj027x]

2 anouncements:
A: F*** script kiddies
B: Lots of progress has been been made! Updates soon to come

[Mad_Maxx]

Awesome project!
It's great to see people forging into the great unknown, with nigh but a workshop and an idea.
Where did you get the idea for putting the quote on the caps like that? It's freakin sick.

Good luck!

[dj027x]

Where did you get the idea for putting the quote on the caps like that?

I'm all for simplicity, but the completely blank keycaps were just too boring. I figured some barely visible text would just give it a nice accent, without being too flashy.
As for my choice of quote, I just love the character of Andrew Ryan; His vision, his philosophies. The quote (at least my interpretation) is a reminder not to be a metaphorical slave; If you want something, you gotta go out there and get it. Its also a little allusion to how free this keyboard is going to allow you to be, as far as keyboards go of course. A contending quote was "No Gods or Kings. Only Man" I actually may end up going with that one in the end

[dj027x]

So, here's the skinny:
I've decided to go with gateron blues. I used a dremel tool to remove the plastic bit, which you can see in the video.
There was that ~10Hz flicker I mentioned in the last video; I fixed that entirely by setting the teensy to 96 MHz overclocked.
After playing around with some code, I've gotten it to run some mildly interesting animations, which you'll see in the video. The moving rainbow, a breathing effect, pong, conway's game of life, and 'ghosting', but not the bad kind, the key you press just lights up and fades away.
I actually reeeeaaally want to program tetris into it next.
Anyway, materials have been ordered: 12x12 double sided copper clad FR4 board, a 3 to 8 decoder http://www.digikey.com/product-search/en?vendor=0&keywords=568-1417-1-ND, and an A105054-ND http://www.digikey.com/product-detail/en/5-146130-6/A105054-ND/1125654 header to solder to the underside of the teensy so we can access those much needed GPIOs.
I've decided to just use Express PCB to create the schematic instead of going through the hassle of learning kicad, and have I got a treat for you;
current mockup:

More

single block:

More

And for those playing along at home, just let me know if you want the actual express pcb file, and i can email it to you.
Now for some game changing developments, I have actually looked into, and subsequently been getting really excited about getting a 3d milling machine. In particular, the xcarve https://www.inventables.com/technologies/x-carve, a desktop sized 3 axis milling machine with up to 75 micrometer accuracy. With this, I can mill my case out of plastic, cut out the circuit board and drill the holes in one go (2 for double sided circuits) instead of etching it with chemicals, and even shave just a hair-thin layer off of the keycaps to engrave the lettering! The head could also be swapped out to create a lathe, 3d printer, laser cutter, pick and place machine, and drill press! Given the parts I have laying around, I could build it for $600. That's pretty damn good in my opinion. Plus, I may decide to create my own switched in the future, which I could use this machine to create =D
So, I'm currently in the process of finishing up the design and then I'ts just a matter of creating the board, assembly, and programming the teensy.
Please let me know if theres anything else you want me to cover!

note: link to video coming soon

[dj027x]

link to video:

[No_joke]

Really nice work!
I'm liking the progress on this
What are decoders used for?
CNC routers are amazing. I've been wanting to build one for some time now.
If you can justify the expense I think you'd definitely have many uses for it.
Now, building your own switches... Interesting! Has it been done before?
Wouldn't injection moulding be a better technique for that?

[dj027x]

Really nice work!
I'm liking the progress on this
What are decoders used for?
CNC routers are amazing. I've been wanting to build one for some time now.
If you can justify the expense I think you'd definitely have many uses for it.
Now, building your own switches... Interesting! Has it been done before?
Wouldn't injection moulding be a better technique for that?

Thanks! the decoder is used because the teensy does not have enough GPIOs to control the LEDs and monitor keystrokes simultaneously, at least, not the way I've wired it up. The teensy will use 3 outputs to select 1 of 8 sets of 6 keys, with the aid of the decoder. It will then check its 6 inputs, and depending on which set of 6 keys it is powering, it can figure out which keys are being pressed. Not sure if switches have been built on this scale before. Yea, I would probably use the mill to create the mold out of aluminum which would be much cheaper than having a third party create it.

[dj027x]

I'v been continuing to develop the PCB layout, so here's a quick update:

More

I've worked out a place to fit the decoder on the main board. I want the board to have as thin of a bezel as the poker ii. This means that I cannot attach the teensy on any side of the board outside the switch  matrix, otherwise I'd have to make the bezel bigger to accommodate it.  This means that I will either have to fit the teensy onto the main board, or a second board below the main one, which will make the keyboard thicker. Ideally, I would like to keep this board thinner than the poker ii as well, but the poker ii has space beneath the space bar to house big components. I think I will try to mount the teensy on a second board near the top of the switch matrix, so that I can play off any extra thickness as a means to tilt the board. however, if I am able to fit the teensy on the main board, I could achieve a thickness of approx. 2.5 cm with keycaps on; 33% thinner than the poker ii!

[dj027x]

Just made a small fix to the wiring. I didn't like running the active high enable line between the led contacts, so I moved some stuff around and brought the line around the switch components.

More

I am finding that circuit design is pretty cool on this scale. It's like a big organic puzzle, where you determine the problems, and how to fix them. Engineering is great!
Luckily, this should all work in the lower MHz range, so I shouldn't have to worry about length matching, crosstalk, joint angles etc.

[Ramage]

link to video:

That's definitely borderlands 2 in the background.

[neverused]

Quick question, in your pcb layout are you going to use diodes and leds in the switch? I know that the switch allows for diodes to be placed in the housing, but that would likely conflict with any Ieds. I didn't see a separate solder pad or location for either component.

[dj027x]

Quick question, in your pcb layout are you going to use diodes and leds in the switch? I know that the switch allows for diodes to be placed in the housing, but that would likely conflict with any Ieds. I didn't see a separate solder pad or location for either component.

I'm actually putting the LED where the diode usually goes, and have moved the led to the left of its corresponding switch. It's hard to tell what goes where without the silkscreen layer, plus I didn't actually label the led pins, because they're inside the switch. But here's a quick sketch:

More

[neverused]

Quick question, in your pcb layout are you going to use diodes and leds in the switch? I know that the switch allows for diodes to be placed in the housing, but that would likely conflict with any Ieds. I didn't see a separate solder pad or location for either component.

I'm actually putting the LED where the diode usually goes, and have moved the led to the left of its corresponding switch. It's hard to tell what goes where without the silkscreen layer, plus I didn't actually label the led pins, because they're inside the switch. But here's a quick sketch:

More

(Attachment Link)

Oh very nice

[dj027x]

Damn this thing is getting compact! It's still very messy, however I have fit everything within the bounds of the physical switches! There are probably still some inefficiencies somewhere that I have not yet caught, but it's coming along. I also noticed there were some broken connections in my last update; oops. Those have been fixed now:

More

[dj027x]

well, my current design is really compact, however because of the key layout that I want, there is no room to fit the teensy on a single board. I am going to redesign the circuit on 2 boards, which will then be sandwiched closely together and connected with a ribbon cable. This will allow for a thin bezel, and the thickness of the board should still end up being roughly as thick as the poker ii. I've bought a 40-pin zif clamshell connector, some short 40 pin flex-cable, solder paste, solder mask, and micro usb female type b ports. The solder mask and paste should make assembling this thing a cinch, and I may put several usb ports around the edges, so one can connect the board wherever they please, however its still up in the air. At 70 cents for 10 ports, I figured buy it now, make decisions later. Updated circuit design pics likely to come tomorrow.

[joey]

Why not connect directly to the teensy with a ribbon cable, rather than a second pcb that has a teensy on it?

[dj027x]

Why not connect directly to the teensy with a ribbon cable, rather than a second pcb that has a teensy on it?

I would need to build a breakout board to attach the cable to the teensy anyway. I guess I could solder the ribbon cable wires directly to the teensy leads, but I think that would get pretty messy. You can see how crowded my design currently is in the pictures above, however if I just move that 3 to 8 decoder onto another board with the teensy, the design becomes much less cluttered, and less prone to errors both in designing and etching.

More

[Dodgy]

Man, this is looking freaking awesome 

[dj027x]

Thanks! I know things are moving slowly, but I only get about 3-4 hours a day to myself. I can however guarantee everybody that this board will be done in under 6 months. I'm rapidly approaching fabrication time!

Dude, I love the work you're doing with the keycaps! At the start of this project, I was looking for keycaps like yours: https://geekhack.org/index.php?action=dlattach;topic=54513.0;attach=95645;image but black instead of blue. Do you think you could make keycaps with black opaque sides and a transparent diffuse top? I would love to put those on this board! Taking it a step further, would putting a legend in the transparent area be feasible?

[dj027x]

8mm ribbon cables came in. Just waiting on cable sockets and other building materials to get started. I'm ordering the xcarve mill, or some variant shortly, so I may forgo etching the board in favor of milling it instead.

More

[neverused]

Any progress on this?

[dj027x]

Any progress on this?

Still working out the 2-board design. I have ordered all the parts for the milling machine, so once those arrive, I just need to mill the boards! I'm also waiting on the arrival of the planck keyboard grid top plate to get here. Just a waiting game at this point. I'll be sure to post pics of the new 2-board design when I get something good worked out. Probably sometime this week

[njbair]

I stumbled upon this thread a while back but forgot about it because it wasn't coming up under the Watched tab. Definitely want to keep an eye on this, and looking forward to seeing your new design.

[dj027x]

I stumbled upon this thread a while back but forgot about it because it wasn't coming up under the Watched tab. Definitely want to keep an eye on this, and looking forward to seeing your new design.

Thanks! I'm glad to hear people are interested in the results!

[dj027x]

Just a little update; the 3d milling machine should be here, and assembled in 3 weeks at most. Soon as that's done, I will mill out the main boards! There is actually one circuit board I need to make for the milling machine, which I will use the toner transfer method to etch.

[vvp]

I will mill out the main boards!

That sounds interesting. Post pictures when you are done.

[spfft]

I'd love to hear your thoughts on the mill + getting it up & running. Very exciting build log

[dj027x]

I'd love to hear your thoughts on the mill + getting it up & running. Very exciting build log

I'd be happy to! I will probably put that info up somewhere else and link it in here in an attempt to keep this thread keyboard related.

[dj027x]

Well, it was requested I show the etching process, however, I decided to mill the pcb instead of etch it, as that will result in a much cleaner result. I did however need to etch a tiny board to controll a power supply for the milling machine, so I just posted a couple pictures here: http://imgur.com/gallery/sHXoJ

[AGmurdercore]

I just came in to say that i am really inspired by your work! Please finish this up and make as many guides as possible because i want to make on RGB board as well now lol

[dj027x]

I just came in to say that i am really inspired by your work! Please finish this up and make as many guides as possible because i want to make on RGB board as well now lol

Will do! I intend to document the build process very well. Lots of pics and explanations. At the moment, I'm waiting on a front plate to arrive that will hold all of the switches. Then I just have to mill the pcb and case, then assemble! PCB design is also pretty close to being finished

[AGmurdercore]

I just came in to say that i am really inspired by your work! Please finish this up and make as many guides as possible because i want to make on RGB board as well now lol

Will do! I intend to document the build process very well. Lots of pics and explanations. At the moment, I'm waiting on a front plate to arrive that will hold all of the switches. Then I just have to mill the pcb and case, then assemble! PCB design is also pretty close to being finished

Glad to hear that you are progressing so well! Good luick i will keep and eye on this

[dj027x]

The new 2-PCB solution design is coming along nicely. Having 2 PCBs instead of one is inevitable on this board, since there is not space on the board to hold the teensy and decoder. Even though I got some ribbon cables and sockets, I have opted to simply use some strategically placed posts to act as interconnects between the boards, as there isn't even room on the board for the ZIF socket, plus trace management is a lot easier using interconnecting posts. If I can manage to do so without sacrificing overall thickness, I will use sockets to allow the 2 boards to come apart. This would be ideal because it would allow the board to be serviced if anything broke. With 2 PCBs, the total thickness of the board is projected to be 32mm to 36mm; still thinner than the Poker 2! Now that I'm doing a two board design, I really need to think about the physical construction process of it all. Components will need to be soldered in a specific order, and pads need to be on specific sides of the board, since the board is not through-hole plated. Because of this, I had to significantly redesign the switch matrix, however I am very close to the final solution. Here is the current standard cell:

More

As Iv'e said before, as soon as the top plate and remaining X-Carve components arrive, this project should wrap up quickly. The board designs should be done by the time they arrive, so it will just be a matter of milling the board, applying soldermask, soldering/assembling, and loading a simple firmware to the teensy. I'll also mill the case once the rest of it is assembled.

[dj027x]

Another little update to the standard cell. I rearranged things so that there will be unoccupied space between the corners of every key.

More

<---Free space!

This will make inter-board routing super easy, and allow 5 support posts to be installed. X-Carve components are arriving tomorrow =D

[dj027x]

Welp, the x-carve is assembled and functional! Also, my grid top plate finally came in! Here is a comparison between my sidewinder, and rapture:

More

Additionally, the top PCB design is complete!:

More

Furthermore, I ordered one more 4 to 16 decoder; I have decided to add a relatively unusual feature to this keyboard; if I use this decoder, I will end up with 9 unused pins on the teensy. I have decided to route these to headers on the side of the board, meaning this keyboard will have 9 GPIO pins, plus gnd and power! I have considered exposing the lines for all other pins as well, although I'd have to urge caution in daing anything with them beside reading output, as they are vital to the proper operation of the keyboard.
'Why would I want my keyboard to have GPIO pins?' you might ask. Well, if you don't use them, they will be nicely concealed by the case, and you won't even notice them. If you are adventurous however, you can wire up your own peripherals to communicate with the keyboard! Devices that interface via USB, I2C, JTAG, what have you, all can be hooked up, provided proper protocols are installed on the teensy. I will make as many of these pins as I can the teensy's analogue pins as well, so you can interface with analogue and digital devices.
So yea, there is going to be another delay while I wait for that part to arrive, however once that gets here, no obstacle other than the firmware stands in the way of having a working unit. The last step will be that case to enclose everything. And, we're still geared to end up smaller than the poker 2!

[dj027x]

Well, I realized that ExpressPCB is a ****ty, locked down, proprietary piece of **** that does not allow you to export a gerber file of your design. So, spent the past 2 days converting my top board into a KiCAD project! I found some good models and footprints for the switches, and with pcbnew I am able to render for all to see, an exciting preview of the final thing:

More

I am currently just tweaking my toolchain in terms of design > gerber files > g-code, and designing the 2nd pcb. Going to do a test print with some scrap PCB soon to test how the components fit (Over 500 pins so far!).

[dj027x]

Attempting to do a test print revealed that the program pcb2gcode outputs gcode that grbl does not understand this breaks my toolchain, as I need to convert the drill files from kicad or eagle into gcode files to mill the holes. I have searched endlessly, and found no linux software to do this that isn't miserably out-of-date. LinuxCNC is an option, however that is basically its own distro, and also pretty dated. Luckily, every other part of the toolchain works, and this drill-to-gcode conversion is a relatively simple operation; just a matter of parsing and creating text files. So, I am writing a bash script to take care of this issue (sure would be nice to have my keyboard done to write it with). The script should be done by tomorrow, and I should have a test print sometime this week. Anyway, heres a picture of what the results will be once I get this conversion issue sorted out

More

[lucaslink]

stoked on this build!

[dj027x]

Welp, I finished the script! Now I can take an Excellon drill file,  accounting for the offset caused by the diameter of the endmill, and convert it into the appropriate g-code which will mill all of the holes using a ramped helix, instead of just plunging the endmill into the material. If anyone cares to use the script or contribute to it, it's on github: https://github.com/DJ027X/drl2ngc
Now then, enough Linux/CNC stuff, this is a keyboard forum; I should have a test run of the PCB done in a couple days, and I'll be sure to post pics when it's done.
If the results look good, I will wrap up the bottom PCB design and begin milling, soldermasking, and tinning!

[dj027x]

Just a quick update; I'm almost done with the 2nd pcb design, and I've decided to do the test print right before the full print so as to keep variables like machine settings, temp fluctuations, etc, as constant as possible between test printing and production printing. I started this blog on jan 13 this year. By the looks of it, this board should be done before jan 13 of next year! Just one year of what probably amounted to a couple hundred hours of designing/engineering to produce a one of a kind keyboard from scratch! I'll bet that's better than some actual keyboard companies can do, and they have full teams of designers!

[Hexterdude]

On the aesthetics part of the keyboard have you decided if you're going to design a rapture themed keyset or case for it?

[dj027x]

I am just going with a plain black case, and plains black windowed keys. I'm still considering etching a bioshock quote on the keycaps, but haven't decided yet.

[spfft]

So pumped to see this!

[dj027x]

Well, you shouldn't have to wait long. Just a little more fiddling with the pcb design and I'll mill it out and assemble! After that, I still need to make the case and load the firmware onto it.

[dj027x]

Well, I'm going to renege my promise to put GPIO ports on this board. After some thought, I decided that there would really be no benefit to using GPIOs from this board considering the amount of processor power the teensy is already going to be devoting to operating the keyboard. Plus, the layout of the 2nd board is already a nightmare. Routing the unused pins to the edge of the board would take a lot more time than I really think it would be worth. Beside, you can always use the USB cable to talk with other devices. In any case, the designs for both PCBs are complete for the most part; I'm just going to run through verification to ensure that everything is connected properly and that it will be possible to physically construct everything. Here's a quick screenshot of both boards.

More

After I verify that the designs are valid, it'll be time to mill them out, apply soldermask, tin plate, and assemble! Then I'll load a quick proof of concept firmware onto the teensy before fleshing my own out. Simultaneously, I'll be designing the case that it'll all fit into. Like the very first concept image I posted on this thread, I'm going to try to keep it very sleek and simple. I'll update again once I begin the PCB fabrication process.

Edit: By the way, if anybody knows of some basic circuit routing techniques, I'd love to hear them! This thing was a pain in the ass to route. I figured I'd make all vertical traces on one layer, and all horizontal layers on another layer, but limited space, and number of connections made that unfeasible.

[dj027x]

Verification is all done! I found a couple major errors, and fixed them up, and verified that all the connections exist that must exist, and that they all go to the right place, both PCBs align properly, and the design will by feasible to physically construct (no solder joints between the sandwiched PCBs) Here's a picture of the final design. If you'd like the files, pm me and I'll send them to you, otherwise, I will throw them up on github once this project is all done.

More

Now, I need to convert the gerber files to gcode (should be done within the day) and I can prepare to mill out the boards!

[spfft]

This is so great! Making me really want to attempt something similar on my own. With Topre.

[dj027x]

This is so great! Making me really want to attempt something similar on my own. With Topre.

Ooh, that would be challenging, as topre keys are not self-contained units. You'd need a mesh and custom moulded frame to make it work. Not impossible if you've got a CNC mill though!

[dj027x]

Well, I have begun running some test prints just to make sure everything is dialled in properly, and that I will be able to manufacture the difficult bits just fine. For anybody who cares to know, I'm using PCB engraving bits to mill the traces, and PCB drill bits for the vias and small stabilizing legs on the keyswitches (links for these items are at the end of this post). I found out the hard way (after breaking 3 bits) that the PCB drills are strictly drills, not endmills (they were advertised to me as endmills). They will not make lateral cuts, which became clear to me only after I inspected the bits to find that their teeth were not properly shaped to make lateral cuts, only plunge. I have not yet decided what to use to mill the hole for the large center peg on the keyswitches, however that update will be coming next. For now, I just waned to show you some PCB porn; I was testing to see if the engraving bit would be able to cleanly cut the tiniest parts of the board, which it most certainly was.

First, here is the difference spindle speed makes. The left part was cut with the spindle running at roughly 1/4 speed (not sure what the actual RPM were, but it was at 25% duty cycle at 48V). The right part (where the cut is much cleaner) was done at full speed

More

So we can make clean cuts, time to test fidelity with my fingerprint for scale.

More

Hot damn that's a clean cut! I used a wire brush to clean most of the dust out of the trench. I'd probably be better off even going a bit more shallow than this. By the way, if you're doing this DON'T BREATHE THE DUST. Breathing in fiberglass dust is a sure way to get pneumonia. Now, the fidelity is good, but are the cuts accurately aligned?

More

They sure are! Looks like the bottom PCB is fab-ready! I'll mill it out shortly after I run some tests pertaining to the top PCB. I also came up with a strategy to align top and bottom sides when I mill them out, but I neglect to show pictures for that. Essentially, I just put a bounding box around all of the designs with a hole in each corner. This way, I can make sure all the holes line up before each print. That's all for now. I'll update with test results for the top board soon, then it's production time! I'll just leave you with a couple more PCB pics.

More

More

engraving bits: http://www.amazon.com/Autek-Titanium-Carbide-Engraving-J3-3001Tix10/dp/B00HC98K2C/ref=sr_1_2?ie=UTF8&qid=1448229038&sr=8-2&keywords=pcb+engraving
drill bits: http://www.amazon.com/10Pcs-Print-Circuit-Board-Carbide/dp/B00IMXHXGG/ref=sr_1_3?ie=UTF8&qid=1448229138&sr=8-3&keywords=PCB+milling+bits

[dj027x]

Boy, things are moving right along; I have finished testing all that I care to test as far as the top PCB is concerned; I made a little test bed consisting of 2 switches and one diode. I loaded up a pattern that milled one hole in each corner, for alignment between runs. First, I milled all of the vias, then the traces, then the holes for the legs on the switch using a fishtailed burr which was really loud, (didn't bother milling the backside traces), sanded away any burrs, and tried to fit a switch in, however my super tight tolerances, along with a tad of misalignment didn't allow the switch to fit. I widened the hole a bit by running the burr around it again, and everything fit together nicely. Here are the pics. I intend to mill the bottom PCB within the week, once perfect my soldermask application method.

More

More

More

More

More

[dj027x]

Happy thanksgiving everybody! I'll be milling out the bottom pcb today. Instead of milling out all of the holes for the legs on each switch, I have decide to just buy the proper size drill bits, and drill them out. Drilling will be MUCH faster and quieter than milling the holes. So, I'm off to my hardware store (hope they're open today) to pick up some drill bits, then I'll soldermask an tin the board. See ya soon.

[neverused]

Happy thanksgiving everybody! I'll be milling out the bottom pcb today. Instead of milling out all of the holes for the legs on each switch, I have decide to just buy the proper size drill bits, and drill them out. Drilling will be MUCH faster and quieter than milling the holes. So, I'm off to my hardware store (hope they're open today) to pick up some drill bits, then I'll soldermask an tin the board. See ya soon.

Can't wait to see your results!

[dj027x]

hardware store was closed yesterday, and I discovered today that they didn't have what I was looking for. Looks like I'm ordering the proper sizes from amazon. :/

[dj027x]

Some unfortunate news: I'm pretty much stuck at the moment. I need either a set of drill bits with a 1/4 or 1/8" shank, or an adjustable chuck that fits where an ER11-A collet usually fits. As far as I know, neither of those things exist. I *could* mill out the holes, but there are 184 holes, and milling them is extremely loud. So, I can't mill out these holes while I'm still in my apartment, however I will be moving back into my house early january. Unfortunately, it looks like this project may be put on hold until then, unless I can find a proper set of drill bits, or an adjustable collet. Either way, rest assured that this project will definitely be completed in either January or February at the latest.

[jaffers]

'ed

But if you are having trouble clamping? smaller drill bits than you want to get a collet. stick it in the chuck and stick your drill bit in there. You should be right as rain. Great work so far on the build man, interesting design

[dj027x]

And now for some good news: I ordered a set of collets of various sizes both metric and imperial. The imperial set should get here monday, so I can resume milling the PCBs! I was scared I was going to have to put this project off for a month haha. The collets were roughly $20 a set, however they should last a lifetime, plus they allow me to use nearly any diameter drill bit or endmill I want, so I'm no longer restricted to bits with a 1/8" or 1/4" shank. Man, I've probably put almost $2000 into equipment for build these boards haha. In any case, the BOM cost for one keyboard should be a tad under $200, if my dead reckoning is accurate. I'll have more updates monday!

[lucaslink]

And now for some good news: I ordered a set of collets of various sizes both metric and imperial. The imperial set should get here monday, so I can resume milling the PCBs! I was scared I was going to have to put this project off for a month haha. The collets were roughly $20 a set, however they should last a lifetime, plus they allow me to use nearly any diameter drill bit or endmill I want, so I'm no longer restricted to bits with a 1/8" or 1/4" shank. Man, I've probably put almost $2000 into equipment for build these boards haha. In any case, the BOM cost for one keyboard should be a tad under $200, if my dead reckoning is accurate. I'll have more updates monday!

yay! I was going to offer to mail you my collets for use to get the project done. i have a bunch of various sizes.

[dj027x]

Man, Sorry for the lack of updates; I had to move from Minnesota to New York last month. But, I'm all set up and ready to continue the build! All of my collets and drill bits are here, so It's time to begin milling the main boards!

[ZefyrPyon]

Boy, things are moving right along; I have finished testing all that I care to test as far as the top PCB is concerned; I made a little test bed consisting of 2 switches and one diode. I loaded up a pattern that milled one hole in each corner, for alignment between runs. First, I milled all of the vias, then the traces, then the holes for the legs on the switch using a fishtailed burr which was really loud, (didn't bother milling the backside traces), sanded away any burrs, and tried to fit a switch in, however my super tight tolerances, along with a tad of misalignment didn't allow the switch to fit. I widened the hole a bit by running the burr around it again, and everything fit together nicely. Here are the pics. I intend to mill the bottom PCB within the week, once perfect my soldermask application method.

More

(Attachment Link)

More

(Attachment Link)

More

(Attachment Link)

More

(Attachment Link)

More

(Attachment Link)

Are you able to edit the PCB to accommodate a TKL and 100% design? I would KILL to find a DIY RGB keyboard alternative, but all the software out there is crapola.

[dj027x]

Unfortunately, there would be hardware limitations that would make that change difficult. I chose all of my hardware to be only as powerful as it needs to be for this particular design. It should be possible though. I mean, the there are already 100% RGB boards out there; I was just looking for something far more personalized, and compact

[ZefyrPyon]

To be honest, anyone can really pick up an iron and learn to solder, but I just don't have enough electrical knowledge to make my own PCB. Do you have any recommendations for me?

[dj027x]

To be honest, anyone can really pick up an iron and learn to solder, but I just don't have enough electrical knowledge to make my own PCB. Do you have any recommendations for me?

Well, no doubt, there is a lot of knowledge that goes into PCB design, but it is all pretty easy to understand. My knowledge of electronics has come from a mix of personal and academic experiences. I learn the theory from college courses, and I implement that knowledge in a practical way in personal projects. Before college, I tried very hard to teach myself electronics, but didn't really get anywhere. A professor is simply a far more powerful resource than a book or youtube video because professors often have industry experience; they know both the theory and practical applications of electronics. That being said, the next best thing to a college course is a college textbook (which I never attempted to use while teaching myself, so it may work very well). They are usually $50 to $150, however you should be able to find torrents for them if they're over 1yr old. The textbook I used in college (which I find to be rather well constructed) is introductory circuit analysis http://www.amazon.com/Introductory-Circuit-Analysis-12th-Edition/dp/0137146663
Now, if you're talking specifically PCB manufacturing, as opposed to circuit design, all of my knowledge of that actually is self-taught. I simply researched the various different ways of making PCBs at home, and picked my favorite one. As far as actually designing the PCB, you just need to learn how to use the proper software (I recommend either eaglePCB or KiCAD)

[dj027x]

The top board should be milled within a week or so.

[FLFisherman]

The top board should be milled within a week or so.

Milled?

[mastermachetier]

To be honest, anyone can really pick up an iron and learn to solder, but I just don't have enough electrical knowledge to make my own PCB. Do you have any recommendations for me?

Well, no doubt, there is a lot of knowledge that goes into PCB design, but it is all pretty easy to understand. My knowledge of electronics has come from a mix of personal and academic experiences. I learn the theory from college courses, and I implement that knowledge in a practical way in personal projects. Before college, I tried very hard to teach myself electronics, but didn't really get anywhere. A professor is simply a far more powerful resource than a book or youtube video because professors often have industry experience; they know both the theory and practical applications of electronics. That being said, the next best thing to a college course is a college textbook (which I never attempted to use while teaching myself, so it may work very well). They are usually $50 to $150, however you should be able to find torrents for them if they're over 1yr old. The textbook I used in college (which I find to be rather well constructed) is introductory circuit analysis http://www.amazon.com/Introductory-Circuit-Analysis-12th-Edition/dp/0137146663
Now, if you're talking specifically PCB manufacturing, as opposed to circuit design, all of my knowledge of that actually is self-taught. I simply researched the various different ways of making PCBs at home, and picked my favorite one. As far as actually designing the PCB, you just need to learn how to use the proper software (I recommend either eaglePCB or KiCAD)

Hey man amazing project and amazing work . I am very interested to see the results. Just a few quick questions on your process into making the pcb. What is the name of the process and are there any resources that are handy for replicating. Also when you are finishing the pcb here is there a reason you are milling instead of the process you used earlier or showed in that first pcb your created. Also Is your goal when you publish your pcb design to make it be able to be sent any where that produces custom pcbs and also be able to make it yourself at home?

Thanks again for the time you are putting into the process.

[dj027x]

Milled?

Hey man amazing project and amazing work . I am very interested to see the results. Just a few quick questions on your process into making the pcb. What is the name of the process and are there any resources that are handy for replicating. Also when you are finishing the pcb here is there a reason you are milling instead of the process you used earlier or showed in that first pcb your created. Also Is your goal when you publish your pcb design to make it be able to be sent any where that produces custom pcbs and also be able to make it yourself at home?

Hey, glad to hear you are interested in the project! Are you referring to the process to design the PCB, or to fabricate it? To design, I simply determine the components I will need, and how they must be arranged, if the user is going to see them. From this, I determine where they sit on the PCB, and from there it's pretty much trial and error; attempting to find a circuit design that connects everything properly without ridiculously long traces or any erroneous vias. If you look VERY closely at past versions of the key matrix layout in the photos I've posted, you will notice various errors in the design, hence why I made newer versions (and HOPEFULLY eliminated all of the errors). If you are referring to the PCB fabrication process, I believe it's just referred to as 'milling a PCB'. I actually buy a copper-clad FR4 (FR4 simply means 'fiberglass') board. I then use the milling machine to cut troughs just deep enough to cut through the copper layer, but not the rest of the board. Resources to replicate it are pretty scattered around the web. It's taken a lot of programming/scripting to fine-tune my milling machine. However if you are interested, my milling machine is a modified X-Carve, and can be purchased for about $700 at inventables.com. The software I use to create my circuit boards and convert them into a path for the milling machine to follow is EaglePCB or KiCAD (for designing the board) Visolate (converts copper trace paths into toolpaths) drl2ngc (a script I wrote myself to create a toolpath based on the vias that need to be drilled for the design) Camotics (to get a rendered preview of what a toolpath will end up cutting out) and Universal G-Code Sender (communicates with the milling machine). Keep in mind that all of these tools are for GNU/Linux. I don't know if they have windows/mac alternatives. As for why I chose milling over etching; I find that milling circuit boards is faster, far more accurate, repeatable, efficient, controlled, cost effective, and less messy (it creates fiberglass dust, but there are no hazardous chemicals involved like there are in the etching process, and I have a gas/particle mask and air filter to filter out the dust). When the project is done, I'm going to release the project files in a format that most PCB manufacturers will accept, under some Free and Open Source license so people can improve and redistribute the design if they desire. Because the design is only 2 layers, and has relatively thick traces, it should be relatively easy for anybody to make at home, no matter what fabrication process they'd like to use.

[dj027x]

Just so y'all know, I've had to put this on the back burner while I focus on college work. Shouldn't be long till I wrap this project up

[TheJonas]

Just so y'all know, I've had to put this on the back burner while I focus on college work. Shouldn't be long till I wrap this project up

Good luck! Looking forward to seeing the final result