DK1 Build Thread: Microsoft Natural Ergonomic 4000 with Cherry MX Switches

[VoteForDavid]

I finished swapping out the first bag of springs.  I need to break into the second bag to have enough for the full keyboard, then go through and test each switch for sticking.  Already I know several stick a little or "click" up instead of rising smoothly, and I'll swap their springs into switches that don't stick.  The reject pile already includes several bases that are anywhere from slow on the rebound to actually almost-don't-rebound level of sticking.

Pictures!  First the pretty:  I tacked the mounting plates to the bottom of the top frame.  I accidentally flipped the patterns when it was time to get these cut out and the textures are on opposite sides on each plate.  In case you can't see it well, the left side is smooth and the right is textured.  I think the texture looks better but I'm pretty sure it will be easier to score the smooth side.  Sometimes accidents lead to education, and I think this is one of those times.  I could be surprised and find the texture takes a more-visible score line.



Then the ugly: These are the worst of the rejected switches.  There were not that many really funky ones; these are the worst of the lot.

More

Corroded switch pin


Corroded interior contact


Corroded diode


Grotty slider - the only thing I can think is this must have been a switch that didn't have a key cap for several years or decades.


...43...

[VoteForDavid]

I sorted through all the loose switches I have.  91 have acceptable action with their new springs.  A couple dozen are somewhat sticky, ranging from slow on the rebound to a couple that wouldn't even pop up again.  I've almost got a full set of good here and I think I'll be able to deal with slightly-sticky switches for less-used spots like ~ and numlock, etc.

Some more time working out the matrix.  I've almost got all the traces outlined and it's going much easier since I made the image larger.  If I could tell myself what to do at the start, I would remind myself to break out the light table and make backlit images, because that would likely have been much faster.  Next time I have to trace out a matrix, I guess that's what I'll probably do.

...45...

...and my laptop died, so that's probably getting attention instead of this project tomorrow

[VoteForDavid]

Thanks God, the heat gun treatment revived the ancient nVidia GPU one more time!  Laptop in business, I fiddled with switches some more.  I have a full set of switches plus a few spares.  I don't have enough leftovers to fill a 104-key board for DW.  That, plus the fact she can use a straight keyboard mean I might just buy her one instead of hand-making another custom board.

...46...

[VoteForDavid]

OK it's time to start in on the main mounting plates.  That means it's time to finalize the layout* and mark switch mounting holes on the plates for cutting.

To that end, I made a couple of custom tools.  One is the size of a switch cap, with a central hole.  This can be laid on the square switch location marker on my paper layout sheet from 4 posts up, and the center of the switch can be marked.  You can see the + of the stem of a cap through the hole on the topright picture.  Drilling this hole on a switch plate will put a hole in the center of where I want my mounting hole.

The second tool is for marking out the to-be-cut-out squares for the holes.  It's approximately 0.55" square, with a central hole.  There are little baby gaps on a few spots when the tool is inserted in a switch hole and backlit,  but these holes probably won't be perfect anyhow.  To use, drill the hole you marked with the first tool, put a rod (drill bit) in the locating hole you just drilled, and put this tool on the rod.  Straighten this tool for proper key alignment.  Mark the square, remove the tool, cut out the mounting hole.  SOUNDS simple enough eh.



*I realized I could have more room for a larger carriage return key if I shifted the whole right hand layout to the left a little.  I realized I have room to do that.  When laying out my switch locations before, I was fixated on keeping the home keys in the same relative positions as the Microsoft layout had them.  I don't need them to be in the same spots, so I can move EVERY key.  I don't know why this was such a revelation but it was kindof a big deal for me.  I need to re-re-draw the paper switch layout picture.  I think, since the plastic frame pieces are mounted to the top frame of the board, I'll cut some paper to fit the mounting plate area exactly, instead of eyeballing it, and draw my layout with the paper in the keyboard frame.  Sounds like a minor hassle, but then - so is this whole project!

...47...

[VoteForDavid]

Holy.  Cow.  I would not have guessed it would take so long.  It took two hours to do a final layout for one hand.  I got halfway done, before I realized that I needed to begin with the home row.  I erased it and started again from the home row.   I was going for a large left shift key.  Left shift ended up a normal-sized key - and I am NOT going to be curling my hand halfway into a fist just to type a z with my pinky.  The z is going to fall under my finger like someone designed it that way and the Shift is relegated to standard-width.  There is enough space when all the rows are populated toward the top, to have a stabilized Space key at a hefty angle. 

It made a surprising difference to have the frame around the key plate.  It was surprisingly helpful, actually.

...49...

[VoteForDavid]

Okay this is either going to be brilliant or a disaster.  I think it will be brilliant.  It will take another man-week of work to even see if I like it   DW and her smaller hands are completely incompatible with this layout.  One more thing to make co-workers hate using my workstation   

The final layout, after much millimeter-scale revision on the right side:



Next up: drilling a hundred holes and making a bunch of little squares on plastic.  Then comes the really hard part.

...52...

[jacobolus]

I don’t really have anything substantive to contribute, but I’m enjoying following this thread. Keep us posted how it goes, and good luck!

[VoteForDavid]

Thanks!  I'm enjoying it too, especially thinking about how potentially good the board could be when it's finally done.  Every time my current MS4K registers key presses out of sequence because of the inconsistency of the membrane switches, it motivates me a little more to get this project done.  I've set aside my side job that brings in actual money, just to have "spare" time for this.

A little more work went into the pictures of the matrix.  I had to increase the resolution again for the hopefully-final bits of work, and the file is in the 300MB range.  I think I'll stop saving it to a USB drive because the loading/saving times must be getting ridiculous compared to how fast it would be on the SSD in the workstation.

I marked the locations to drill holes on the plates, and ran out of time for the night.  Turns out, the smooth plastic is WAY easier to see score marks and dimples compared to textured, but if one is inclined to make mistakes when marking for drilling/cutting the smooth plastic would show fewer of them.

...53...

[Zustiur]

I'm following this with great interest. Mushy keys was one of two reasons for not using the MS keyboard that I was given at work. The other was that it was wireless.

Please tell or show how you plan to cut the square holes for the keys. My mind boggles whenever I try to imagine doing that.

Regarding USB vs SSD I have to say save it to the SSD and (this is important) BACK-UP YOUR FILE. Not implying that ssds are unreliable, just saying that you don't want to risk losing all that effort!

[VoteForDavid]

Thanks for that, it's very good advice!  I currently have files within a day of each other in physically separate locations, and the computer copies the SSD to a platter drive every week also.  And I have a not-terribly ancient clone of my home PC sitting on a shelf at work, as a last-ditch backup for the local, airgapped backup drive for the home PC.  Backups: important.

Loading the file into Photoshop still takes several seconds, but saving is much faster to the SSD!.

The squares will be a challenge, for sure.  Currently I consider that the hardest part of this whole affair, and I will have to experiment a bit to figure out a good way, before I lay into the main plates.

I drilled out the center-locating holes and started marking the square mounting holes.  Pictures to follow.  This took surprisingly longer than I thought.  I suppose by now I should have stopped being surprised that everything is taking. so. long.

...54...

[VoteForDavid]

OK so now it's time to figure out what to do  )  I need to come up with a strategy for cutting these holes out.  The current plan is drill out the centers and xacto the corners and edges, which took 5-6 minutes each when I tried it before.  That puts me at around 7-8 hours of making holes.  Maybe a chisel could speed things along.  Then I need to mill (or otherwise thin) the backsides, because this plastic is too thick to let the switches snap in place.  Then I want to notch the sides of the holes to allow opening the switches without removing them.

The frame for the tenkey area was only tacked in place, so when I accidentally dropped the board it popped loose.  Making lemonade: I turned the plate into an alignment tool.  Mount the second custom tool and put the frame on it, and rotate the frame so the drilled hole next door is in the middle of the empty square next door.



This would have worked beautifully if the rows were straight or the keys were aligned regularly.  I got about 2/3 of the board done this way, and eyeballed the rest.  At least, with such an oddball layout, nobody is likely to notice a few degrees of misalignment

I used a very sharp awl to scratch around the square, to mark my hole locations.  The smooth side took a line much more easily.  The textured side looks a LOT better IMO.  In progress:



Done!

[VoteForDavid]

Using pilot-tipped drill bits of two sizes, I drilled the holes out to 1/2".  The square marks were disturbingly close to the edges of these holes, but 1/2" is smaller than 0.55" so it was ok.  I roughed one out square and it was still too small to fit a switch, so no worries.  I tried one of my smaller end mills but it doesn't have much in the way of side-cutting ability so I used a rotozip-like bit to rough out the corners of the holes on one of the plates.  I also tried a manual nibbler, but it's designed to cut material a hair too thin, so it could only take a partial bite.  It worked, and a slightly-larger nibbler would have been useful - except that it also marred the scored-square lines  so it looks like a bit of handheld x-acto trimming will be required for each hole but much less than otherwise.

It's scary how not-square the holes are, but they are also undersize.



in-the-weeds details about the drilling follow:

More

These two plates were cut from the same piece of plastic, so I was pretty surprised at how differently they reacted to drilling.  The smooth-up side drilled super easy.  It made fluffy shavings that had me stopping to clear the bit every 4 holes.



The texture-up side was WAY different.  It took three times of it catching on the tip of one of the flutes and rising fast, slapping my hand a little on the way by, before I figured out it needed a much-lower rate of feed.  Until that realization came, at the same feed rate, it kicked up these funky tight spirals of rigid plastic that kicked off the bit.



The voices tell me this pile could come in handy as a source of fine ABS, in the near future.  Note the tenkey plate included for scale.

...56...

[VoteForDavid]

The top layer of the matrix is all photoshopped up and the traces are different colors, ready to be sorted out.  The bottom layer is well underway. 

The second plate looks much like the first, with the holes roughed out square.

I think I have a method for the switch holes!  The baby end mill may not be much for cutting on its sides, but the end is sharp enough.  With the holes roughly square, mill the back of the plate to about the right thickness so there is less to cut, then cut the square to fit a switch nice and snug.  Mill the edges of the newly-square hole to something more like the right thickness, and the switch will fit and clip onto the back of the plate.  I got a couple of these done, and it seems to work well but with one caveat: the plastic is not a uniform thickness, so the depth of the mill has to be adjusted per hole.  That's a bit of a hassle, but it leaves the rest of the plate between the holes thicker, which will help with strength of the overall part.  It is much easier to cut the holes square with a knife, with the backside of the plate thinned a bit.

So the making of the switch holes may have just gone from the hardest part to only tedious.

...58...

[Zustiur]

I'm not sure that I followed all of that but the picture showing the drill bit with the pilot tip helped my understanding a lot! Thanks. That gives me options for my own custom board (thread coming soon). I wonder if foam board will be strong enough for prototyping. I think it will be perfect for such placement testing even if not for wiring up.

[VoteForDavid]

Foam board seems like more work than paper, but also more rigid if that's what you're going for.  Lay your fingers on a flat piece of paper in a position which feels natural.  Draw a circle around each fingertip.  That's your home row.  That's what I ended up doing anyhow.

The milling blockoftext will make sense with a picture of it done, which will be coming in the next few days.  Maybe some, maybe none work in the short term, I have a day trip to Houston and back and it might take enough energy out of me to not want to work on this at all!

[Zustiur]

Given the current state of my fingers I suspect teasing on paper won't be enough testing, though I'll definitely start with that.

[VoteForDavid]

Wait  what's the matter with your fingers, mate? 

So here's the back of one plate with examples.  These holes were drilled out, then roughed out with a side-cutting bit (a few days ago).  The holes are square-ish but all undersized and the corners especially need some attention.  In the top of this picture, the backs of the switch holes are all roughly thinned out with the mill.  The bottom-right has the square for the switch cut into the top (which is the reverse, in this view).  The hole will accept a switch but the edges are very close in places to the edges of the milled-out area.  The bottom-left holes have been treated to another pass on the mill, and there is "lots" of space around the switch hole.  There is room here and to spare, to cut out notches for opening switches without dismounting, in the future.



I went through and rough-milled the edges of the backs of all the holes on both plates.  Then I used a handy-dandy 1/2" wide, flat, fine toothed file to shape the holes.  I cut with the file to the square lines on top of the plates.  It's a file for steel, so this happens pretty fast (and leaves gouges in the corners of the holes which speaks to how far they are undersize! )  When the hole gets big enough to squeeze a switch in, the switch polishes the high points in the hole so they are easy to spot and knock down.  The switches are not-quite square, but if the holes are only cut enough to fit the switch one way, the sides may be bound too tightly and prevent the stem rising when the switch is pressed.  Make sure to cut the holes big enough to fit the switch when rotated 90 degrees for proper operation.

I also spent some quality time with the image of the matrix.  That's coming along faster now, since the image is over 300MB and I can use hard-edged 8 pixel-wide brushes.

...59...

[Zustiur]

I'm not sure if it is RSI or arthritis. My left pinky is the most annoying. I've noticed that with my other fingers curled to sit on a home row my pinkies have to be straight in order to reach. Column stagger helps, and I can test that on paper but I think key height will be a major help. That is where rapid prototyping on foam board comes in.

[VoteForDavid]

Seems like you might want to try a layout like the TEKB that has the pinky keys closer to the wrist.  If you can lay hands on a keyboard like the Microsoft Natural series or their new Sculpt, that will put the keys rotated on both sides to have pinkies closer to the wrists.  If you have worse problems with the left side, maybe rotate the right side of your current board away from you and see how that goes?  You sound like a prime candidate for an "ergonomic" spread layout.  Also maybe for
http://www.amazon.com/Pain-Free-at-Your-PC/dp/0553380524

[VoteForDavid]

I spent some more time on the mounting plate holes.  I timed it, it’s about 3 minutes or less per hole, using the file after the holes are drilled and rough-cut square.  Call it maybe 5 minutes total to get a finished switch hole.

The switch matrix image is finally finished!  I am in process of converting it into a wiring diagram.  It’s a pretty odd ball compared to some other matrices I’ve seen.  The traces wander all over the place, to avoid rollover and/or ghosting.  To get all over the place, they loop around and around each other.  The sections of the clamp-on PCB connector I thought were overlapping, actually don’t.  They used some of the pads on the bottom layer to provide contacts for the top.  Clamp it all tight, and you don’t need ANY other connectors so it’s pretty clever.

I added an image layer per trace (35 new layers) to show me where to wire what.  Because they’re all going to be on wires and not constrained to go around each other on a plastic film like the original traces, the routes are amusingly different from what Microsoft used.  This is going to be a rat's nest, though.  The image file is at 340 megabytes and will be growing quite a bit yet.  If I remember I’ll make a (much) smaller version just so you peeps can see how all-over-the-map this matrix goes.

...61, this is crazy...

[heedpantsnow]

Love this thread and always love the updates.  keep up the good work!

[VoteForDavid]

Thanks! 

***

Ok so here is what I'm dealing with: the switch matrix, both layers overlaid, much-smaller than the working file (which is currently sitting at 9000 x 3852 pixels, with over 40 layers, 353MB)


Now you get an idea what I mean when I say this matrix is all over the place!

If I am going to implement per-switch diodes, I need to figure out which side of these traces goes to which side of each switch.  Then there needs to be enough voltage to overcome the voltage drops of the diodes.  That would involve dozens of test points on the controller which I am disinclined to install, or re-wiring the whole thing AFTER it’s done.  This board has never ghosted or rolled over on me, so I think diodes are a complexity I may leave for version 2.  I’ll leave the diodes in the switches though, because it’s more work to remove them.

I need to finish the switch holes, re-install the plates in the frame, install switches, then wire it up. I can't really wire up the switches with the plates out of the frame, because of the way the circuit runs all over the place.  I am considering leaving the holes plain squares.  Since it's going to all be flying leads, there should be enough slack to pop a switch from its mounting hole without desoldering the switch, if I ever do decide to go back in there.The first-draft wiring diagram looks like this:



It would be REALLY helpful if I had a transparency printer, so I could make color overlays of all the traces and all the wire paths, and turn it over and have a 100% visual from the backside.  That’s not happening, but it would be cool.  I guess I could use the actual matrix sheets on clear Plexiglas, with a blank sheet on which I could draw one wire route at a time, but it wouldn’t have all the fancy colors the digital version has.  It WOULD be reversible, which is kinda huge because I made this matrix looking at the top, and wiring and soldering is from the BOTTOM  

I'm about 3/4 done making the square holes in one plate, then there are the holes in the other plate to finish which will take a few days more.  The astute observer has noticed by now the space bar switches don't have their holes drilled yet.  After the plates are mounted and the other switches populated and caps placed, THEN I'll be ready to decide on final space bar placement.  I may leave the plates and switches mounted to cut the holes for the space switches and stabilizers, but I'll have to remember to mask off the switches to keep ABS dust out of them!

[user 18]

It's quite admirable that you're going to all this effort to match up the original traces. Unfortunately, unless I'm missing something, you don't have enough information in your pink layer. You're representing each switch as a single node, but you need to separate which wires go to which pin on the switch. The exact pin is arbitrary, but each switch will have at least one trace going to each pin.

I would suggest you reconsider your desire to reuse the original controller. Building a Teensy 2.0 (or 2.0++) into the board would cost $20-30 or so, allow you to greatly simplify the matrix, put a full round of diodes in, and make your keyboard programmable, allowing macros and function layers. It's probably also possible that you could wire your trackpoint and buttons up to the teensy as well, unifying the whole device as a single USB peripheral.

If you're trying to get a bottom view of the traces, there's a couple things I would suggest:

1. Mirror the image left to right and flip the plate left over right when you turn it upside-down to solder.
2. Print it normally on thin paper, and put it face down on a lightbox. You can rig up a lightbox fairly easily for this application with a piece of glass (a tabletop or desktop will work well) and a desk lamp underneath facing upwards.

I've really been enjoying your progress log, keep up the great work!

[VoteForDavid]

As I understand it, each switch IS a single node, at the intersection of two traces.  The magenta lines are just to show which traces go to which switch, which is all I need.  Once they get to the switch, it's flip-a-coin time to decide which side gets which wire.  I remembered today that I have a BIG piece of glass that just needs to be cut to be used as a mounting surface for the traces and as a sketchpad for the wiring.  I think it would be a good idea to use opaque tape or take a piece of paper or something so I can clearly label each pane as TOP or BOTTOM view.  I'd hate to get THAT mixed up!  I also have a light box plenty big enough for this (a good suggestion BTW) and if I ever build a time machine I'll go tell myself to use that when making the images of the matrix sheets

Currently this is right at $0 +time, and think I like that as performance art for its own sake.  Anyone with this board and some patience should be able to follow in my footsteps, if the controllers are close-enough (and I don't see why Microsoft would ever feel like changing the controller pinout once they got a working unit).  Aside from reusing the original controller, yes a Teensy is a great way to go.  For one thing the matrix could be not-insane and laid out very neatly.  I don't need function layers or macros on this board for the forseeable future, and so programmability is nice in an academic sort of way.

Thanks!

p.s. forward this to the Unpopular Opinions thread in your imagination: I think function layers are silly when there is room for keys.  With all the web and favorites keys I intend not to install, there are EIGHT key locations just sitting around doing nothing on top of the board, plus the UTTERLY USELESS FLock and Insert keys. 

... now that I think of it, I guess it wouldn't hurt to include a DIP switch array that would turn on those keys, install and wire them, but just leave them useless.  But I won't.

[user 18]

Here's my concern with the way you've set things up, assuming I understand it properly.

When a controller scans a matrix, it will use a pin as an output, with whatever it's outputting (5V, 3V3 are pretty common). It will also use pins as input, and check for current coming into those pins. If current is coming in, it's clearly connected to an output pin.

In this picture, pin A is the output pin, and pins B and C are input pins that are intended to register the presence or absence of current flow. In the top drawing, the switch connected to pin C will only register properly if the switch connected to pin B is also closed, due to the connection of the red wire to the wrong pin. In the bottom drawing, both pins will register normally.


(sorry for poor photo quality, I hope you can tell what I mean).

Now, you've posted pretty low resolution images of your matrix, and it's pretty complicated. My eyes aren't good enough to tell if there is any point on the matrix of your board where this might be an issue, but I'd say it's fairly likely. If there are only two leads going to a particular switch position, then you're absolutely right -- orientation of the switch, and which lead connects to which pin, is completely arbitrary.

If you've taken the above into account and I've missed it, my apologies.

[VoteForDavid]

I am superduper tired (3x <5hrs sleep/night in a row) but I think you have missed something fundamental.  Either you missed that this board has a conventional matrix style controller with a crazy looking layout, or you have missed how a switch matrix works.  If the former, please understand that all the switches on this board do have only two wires going to each of them, and most have the same two wires continuing on to other switches in the matrix.  If the latter, Please see:
http://pcbheaven.com/wikipages/How_Key_Matrices_Works/

I spent some more quality time with the mounting plates and am 2/3 done with the other plate.  As I got more comfortable working with the material and developed a feel for what is required it started taking more like 1 or 2 minutes per hole, and I like that a LOT better than 3 or 4.

...63...

[user 18]

Okay, at this point I'm 99% certain we're just miscommunicating. I'm going to give explaining myself just one more go.

On your wiring diagram, you have a couple of nodes (which I assume are representing switches) which have 3 or more lines running from them. I've indicated a few on the diagram (circled in red).


Of those 4 wires, it's almost certain that there are some which are required to be on the same pin of the switch as each other, and some that must be on opposite pins from each other. Take a look at switch A1 on this matrix. It has 4 wires running from it, 2 red and 2 blue. One red wire runs to pin 1, the other runs to switch B1. One blue wire runs to pin A, the other runs to switch A2. The two blue wires would need to be connected to the same pin as each other, and the two red wires would need to be connected to the other pin.


If, for example, the blue wire connecting to pin A and the red wire connecting to pin 1 were connected to the same terminal, that switch would always read as though it were pressed, even if it is not.

The weakness of representing switches as nodes rather than as a conventional circuit diagram switch is that you cannot differentiate which pin a lead is meant to be connected to. The second image above (which I've snagged from your link on matrices), differentiates this by using blue and red wires to indicate which wire should be connected to each pin -- all the red lines go to one pin, all the blue go to another. However, your matrix diagram is all one colour, and does not have any method of differentiation between pins as far as I can tell.

[VoteForDavid]

Yes, I think we are talking past each other.  That is good!  Better than being factually incorrect - when we get on the same page we have both clarity AND agreement!   Otherwise one of us is still missing something   I think what you have missed is either or both the insanity and/or the method in the madness.  Here I have selected only three layers per image, to show the case of the Enter key, which has FOUR magenta lines to it when all layers are showing.

The magenta line is a wire path.  The gray line with dots on its length is a trace of the matrix circuit.  I have offset the magenta line down/left for clarity, as it partly overlays the gray line.



Same deal for the other sheet of the matrix, except that the trace in question is the orange line and the magenta is offset down/right.



Most switches have wires to and from, on the same leg of one circuit in the matrix.  In these two cases, the bottom-left diagonal line is headed for the controller.  For switches like this one which are "in the middle" of a trace, the switch has a wire going to-and-from each leg.  Some keys have traces that wind around so strangely that I chose to use the "wrong" end of a trace to connect the leg to the controller, and this is one.  The Enter key is a dead-end on the orange path, but the other end was probably screwy so I connected the dead-end to the controller for the sake of simplicity in making the diagram.  A few keys (off the top of my head, E is one) have only a wire from the controller to just the one switch on one sheet of the matrix, and the other side of the switch is connected to a string of other switches on the other sheet of the matrix.  Those keys practically can not be part of any ghosting, because they have their own dedicated circuitry.   

It is the equivalent of Microsoft laying out their matrix like so, with B1 and C1 on different circuits altogether:

[user 18]

Okay, I see what you've done now, and it makes perfect sense. It wasn't clear to me that you had two different wire path layers, as they were both rendered in similar or identical colours (at least on my screen).

Again, it's wonderful all the effort you've put in, and I'm really looking forward to seeing the finished product!

[njbair]

That thing is insane. 18, if that diagram had made sense to you on the first try that would just confirm that you're crazy. So you've got that going for you, which is nice.

It reminds me of a transmission valve body. Every time I see one I think the engineer must have been certifiably crazy.

[VoteForDavid]

The solenoids on the valve body are your Fn and layer shift keys

OK so it took an hour and a half, but I went through and doublechecked my traces-to-buttons colored lines, and here is the wiring of the stock board.  Items in (parenthesis) are going to be omitted, mostly the Favorite and Web keys.  Keys on the TenKey area are preceded by "TK".  Please feel FREE to point it out if there is any key that isn't listed on both a TOP line and a BOTTOM line.  The TOP lines are on the top sheet of the matrix, the BOTTOM lines are on the bottom sheet of the matrix, and together they should have all keys duplicated exactly once.  The two digit number after Top or Bottom is the number of the contact on the controller, as numbered on my image.

Bottom  01 W
Bottom  02 LWin  V  1  F3
Bottom  03 Q
Bottom  04 Ctrl  Esc  (Mail)  4  Fav5  F6  Flock
Bottom  05 LAlt  2  (Search)  F4  3  Mute
Top 06  LAlt  RAlt  TK+  TK9  TK8  TK7  (TK( )  Home  End  Play  P  [  ]  \
Top 07  B  Caps  (Web)  (Search)  N  M  <  LeftArrow  DownArrow  RightArrow  TK0  TK.  FLock
Top 08  V  C  X  Z  Esc  F7  F8  F9  F10   F11  F12  CarriageRet  (TK<-)  Menu
Top 09  LWin  (Fav1, Fav2, Fav3, Fav4, Fav5, Fav*)  Y  U  I  oO  PrintScrn  ScroLock  NumLock TK/  TK*  TK-
Top 10  LShift  (Mail)  7  8  9  0)  RShift  Delete  PageUp  TK4  TK5  TK6  (TK) )
Top 11  Space  A  S  D  F  Q  W  E  R  T  Tab  F1  F2  F3  F4  F5  F6  Pause
Top 12  ~  1  2  3  4  5  6  J  K  L  >  ;  ?  -  +  Backspace  Insert
Top 13  LCtrl  G  H  RCtrl  "  UpArrow  TK1  TK2  TK3  TKEnter  (TK= )  Calc  PageDown  Volume+  Volume-  Mute
Top 14, Top 15, Top 16 (to zoom control)
Bottom  17  E
Bottom  18  R
Bottom  19  T
Bottom  20  ~  A  G  B  Y  7  F7  P
Bottom  21  S  N  H  U  8  F8  [  -
Bottom  22  D  M  I  9  "  ]  +  F9
Bottom  23  Space  Z  (Fav*)  Volume+  >  RAlt
Bottom  24  C  (Web)  F2  2  RCtrl
Bottom  25  F  5  0)  oO  <  F10  \
Bottom  26  LShift  3  F5  (Fav4)  Calc  CarriageRet
Bottom  27  X  F1  (Fav1)  Volume-  Play  RShift  ?
Bottom  28  6  LeftArrow  UpArrow  Delete  Home PrntScrn  F11  Tab
Bottom  29  J  DownArrow  End  PageDown  PageUp  ScrLock  F12 
Bottom  30  BackSpace  TK-  TK+  TKEnter  Caps
Bottom  31  Insert  Pause  (TK= )  (TK( ) (TK) )  (TK<- )
Bottom  32  ;  TK*  TK9  TK6  TK3  TK.
whoopsy I skipped 33 in the numbering!  I left it out here for consistency with my image
Bottom  34  K  Menu  RightArrow  TK1  TK4  TK7  NumLock
Bottom 35  L  TK/  TK8  TK5  TK2  TK0

...64...

[VoteForDavid]

Big time checkpoint!  All the holes on both main plates are squared and hold switches, and the backsides are thinned enough to allow switches to snap on.



It took a lot of adjusting on the drill press, but finally I just settled on "a hair too thin" and hit all the holes with the end mill, so they are all a good thickness on the mounting surgface.  I know this without installing switches, because I made another tool:



Its a scrap of plastic with a notch filed into it, wide enough to slip over a switch plate that's thin enough for a switch to snap on.  Checking holes with this is WAY faster than checking with switches.  If I'd made the plates out of metal (or plastic) the proper thickness, this wouldn't be required but ... oh well.

...66..

[heedpantsnow]

Big time checkpoint!  All the holes on both main plates are squared and hold switches, and the backsides are thinned enough to allow switches to snap on.

Show Image

It took a lot of adjusting on the drill press, but finally I just settled on "a hair too thin" and hit all the holes with the end mill, so they are all a good thickness on the mounting surgface.  I know this without installing switches, because I made another tool:

Show Image

Its a scrap of plastic with a notch filed into it, wide enough to slip over a switch plate that's thin enough for a switch to snap on.  Checking holes with this is WAY faster than checking with switches.  If I'd made the plates out of metal (or plastic) the proper thickness, this wouldn't be required but ... oh well.

...66..

Nice, dude!  So, switches and wiring are imminent?

[VoteForDavid]

For a certain value of "imminent" yes. )  Our lead programmer at my job has a saying, when anyone asks for some change or feature in the software we use.  "oh, it'll take about two weeks" and then everyone laughs.  He's been building this software for a decade at least, with constant revisions to suit company requirements.

So, this keyboard is almost done!  It'll take about two weeks 

Need to temporarily re-mount the plates
   populate with switches and caps
   and decide how and where to place the space bars. 
Then remove it all
   and cut out the holes for the space bars. 
Then re-mount the plates and finish the joints.  I am pretty sure acetone will be involved here, because I don't trust hot-welding the plastic on the backsides of the joints. 
Then I need to finish out the six key holes
   and arrow key holes; I think I might cut these out of fresh plastic too, seeing how well the main plates have gone.  Or maybe I'll trim them out of aluminum like I did the tenkey area, and no further trimming would be required but mounting becomes the challenge.
When ALL the switch mounting plates are done and mounted
   then populated
   then wiring
   then attaching the controller
   testing
   hopefully no troubleshooting!
Deciding what else to use (trackpoint, trackball, touchpad, scroll wheel, whatnot)
   Implementing that
Figuring out WTH to do about mounting the bottom of the chassis
   doing that
aaaaaand done!  So, a couple of minutes' work left still, at least - and all that's just off the top of my head!

[VoteForDavid]

Good news/bad news time!  I tacked the plates back onto the frame and populated them with switches and same-profile caps, and found that I need to fine-trim the mounting holes.  The curvature of the plates makes the holes a hair narrower (VERY tight), and some will need to be opened up with the plates installed, which stinks.  Also, a couple of caps were touching and interfering with each other, and they may need to be shaved just a little.  Maybe not, depending on which caps I end up using.

It turns out I don't have room for the superlong caps I was going to use for the space bars, but regular-long caps (tenkey zero caps) fit well at the angles I wanted.  I made some cardboard spacers and taped them to the undersides of the caps, and put a little piece of sticky-out rolled tape on the bottom, to hold the space bars temporarily in place.  Tweak, tweak, tweak.  After settling on a position, I scribed around the edges well enough to position the caps.  Then I did some stuff with a compass and straight edge and ended up with the outline for stabilized switches about where they should be.  These will be hard to cut, but should be do-able.  That's all the main plate switches located and all but two cut out.  Less than a dozen switch holes to go and the whole board will be ready for switches!

Anyway, because it was all put together, here's what it looked like for about a minute before I blew it all apart again (again):



Yes, the space switches are at drastically-different angles. 

[VoteForDavid]

I didn't like how the switches I was using all felt a bit different to each other.  Lack of consistency is one of the reasons for this project, after all. My unfounded, unresearched theory is that I put too much heat into some of the switches when I was removing the solder, prior to getting them off their previous PCB.  I spent a couple of hours changing these clear stems into the bases of the switches in another MX board that came through, with the nice side benefit of having a bunch of vintage black pieces I can switch in if it turns out I hate the ergo clears.  I also took the time to carefully wipe down every single stem, because some of them had visible "stuff" on them and some didn't.  I wiped the sides as well as the actuating/tactile bumps.  To do the whole board worth of switches took two hours.

I came across a board with PCB-mounted switches, and just left the switch bases soldered to the board, using the board as a switch holder.  Later, I'll do the ol' heat-and-push-sideways switch removal method, which I hope will allow me to put much less heat into each switch.  It seems like a good idea at the time, anyway.

...68...

[heedpantsnow]

I didn't like how the switches I was using all felt a bit different to each other.  Lack of consistency is one of the reasons for this project, after all. My unfounded, unresearched theory is that I put too much heat into some of the switches when I was removing the solder, prior to getting them off their previous PCB.  I spent a couple of hours changing these clear stems into the bases of the switches in another MX board that came through, with the nice side benefit of having a bunch of vintage black pieces I can switch in if it turns out I hate the ergo clears.  I also took the time to carefully wipe down every single stem, because some of them had visible "stuff" on them and some didn't.  I wiped the sides as well as the actuating/tactile bumps.  To do the whole board worth of switches took two hours.

I came across a board with PCB-mounted switches, and just left the switch bases soldered to the board, using the board as a switch holder.  Later, I'll do the ol' heat-and-push-sideways switch removal method, which I hope will allow me to put much less heat into each switch.  It seems like a good idea at the time, anyway.

...68...

Rock on brother!  Keep at it and you'll get there. Thanks for always being faithful with the updates.

[VoteForDavid]

Thanks, and you're welcome.  Encouragement during a long, tedious process is always welcome.

I think I may try left-spacebar as a backspace key.  That sounds like a very good idea.  I basically never use my left thumb for Space. 

I had the faint-but-horrible feeling that I might get the board done and just hate the custom layout.  The Microsoft board I'm using at work is so close to good, it makes me a little afraid that a radical change will be too much.  There's one way to find out! It might be a good idea to tack it all together instead of finishing the joints, so if I need to go to back with some more-conventional spacing the plates won't be too hard to remove.

Speaking of removing, it is WWWAAAAAAAAAAAYYYYYYYYY faster and easier to disassemble a board with PCB mount switches versus plate-mounted.  My.  Goodness.  I depopulated the entire board and had time left over to cut the larger shapes for the space key and stabilizer holes in an hour of work tonight.  Those stabilizers will be challenging and I am glad I've only got two to do (for now).  I'm not glad I was slinging a soldering iron while tired, because I grabbed a hot switch the wrong way and burned my palm but oh well.  That's what you get sometimes.

...69...

[VoteForDavid]

As it turns out, these switches are much more finicky about their mounting holes than I thought.  The holes I had made were too tight to varying degrees.  With the previous inconsistent switches, it was hard to tell what was going on.  With these different switch bases, it is easier to know when the hole is wrong.  The switches rub inside and feel different, when the switch holes are just a thousandth or two small.  The right size seems to be small enough to catch the retaining clips, but large enough I don't have to press hard at all to get them all the way in the plate.  I haven't measured, but something tells me this is approximately 0.55" square

I finished one of the stabilizer holes, then mounted a switch and cap and found that when all three pegs are in the switch cap it becomes very important that the dimensions and relative spacing are the way they should be.  The least little misalignment, and the stabilizers will rub.  With slightly more misalignment, the stabilizers would actually hang up at the bottom of travel.  I got the switch operating smoothly and punted on the other stabilizer.  I started fitting holes to switches and got about half of the plate populated with switches for (hopefully) the last time.

ETA: ...70...

[VoteForDavid]

I finished populating the first plate with switches, cut the second space stabilizer and switch in, and fitted a few more switches to the second plate.

...71...

[heedpantsnow]

I finished populating the first plate with switches, cut the second space stabilizer and switch in, and fitted a few more switches to the second plate.

...71...

Sweet, that's major accomplishment!

[VoteForDavid]

Hey thanks!  It really was.  I was surprised how touchy the stabilized switches were, but I think I shouldn't have been.

Both main plates fully populated



Tenkey populated



The nine last key holes are roughed-out and await finishing.  The arrow key holes will accept switches but the plastic is too thick so they don't clip in position yet.  I'll need to trim the backs of these a little.  Then the nine remaining key holes are all that's left to cut square, thin out, and populate.  That will be all the keys in place!



I clicketa-clackety-clicketed on the arrow keys and it was soooooo nice.  It made me want to have a keyboard full of these switches . . . heeeey wait a minute . . .

...72...

[VoteForDavid]

Moral for the night: freehand on the end mill is something one ought to practice, but with care a passable job can be done.

In related news: all the MX switch holes on all the mounting plates have been finished and populated!  A lot of plastic work was required for this part.  The separator between the six switches and the three switches on top was too fat, leaving too small a place for the top three switches (just like the Fn keys had).  I had to rework the area with my hot knife, and joined the pieces here.  I also had to do some adjustment on the up-arrow key area, because I cut the switch hole too far up and the cap was hitting the frame.  Oh well.



Some of the switches are sideways because I got carried away with the file and the holes were cut a thousandth or two oversize, and switches couldn't catch the plate to latch on - so I cut the holes to accept the switches sideways.

If I can figure out how to get the calculator and media buttons working with switches attached to the top frame, I can fasten the switch mounting plates to the top frame and begin wiring the matrix.  If not, I'll have to figure out some sort of connector that can be disconnected to allow me to separate the top and bottom frames.

...74...

[VoteForDavid]

The top three switches from last night have the same problem as the up-arrow: I'll need to work on the plastic of the frame to allow the key caps to move freely.  Boo.

Tonight was media button time.  The switches I picked out for these are a bit hard to handle because they're so small.  To allow the buttons  to sit at the right height for these, I had to put them at a certain angle resting on the frame, instead of down in the button hole.  The plastic of these switches melts much hotter than ABS, so I decided to go with hot glue instead of heat welding.  Lots of hot glue.  I used the hot knife to sting around the holes so the hot glue would have something to bite onto, and filled the area around the button holes with hot glue.  Then I trimmed the glue down flush with the edges of the frames.



A little dot of hot glue to hold the switch in place, with fine adjustments to position and angle.  This was tricky.



Then hot glue was added a very little at a time, building it up to hold the switch *very* firmly in position.  This is still incomplete, but you can see the progress here from right to left.



The buttons don't really have anything to hold them straight, and they wobble a little.  I hope this won't bother me.  I don't look at the keyboard much when I'm working, so maybe it will be alright.  To make them more stable would be a challenge.  They click the switches and that's the point.



...75...

[VoteForDavid]

The media buttons are all hot-glued in place.  DW pointed out that hot glue is rework-able, and so I went with hot glue instead of acetone to join the switch plates to the top frame.  I also selected a set of key caps which, although it looks pretty terrible, feels pretty good.  I need to adjust the plastic of the top frame in a few more places, but that will have to wait for another day.  Pictures will have to wait for another day too - it's tomorrow already!

...78...

[VoteForDavid]

Plastic work on the top plate is done.  I changed a couple of key caps, but now all of the keys actuate easily as they should.  I need to figure out what to do with a trackpoint and maaaaybe a scroll wheel from a mouse, if I can cram it in there.  Another late night.  Photos will have to wait again.

...80... two solid man-weeks.  I bill well into three figures for this kind of time, when I have a client paying bills.  I've lost that maybe much again on the side job I laid aside to get this project done.  This better be one heckuva keyboard.

[VoteForDavid]

Okay here is the board fully populated with keys, and all the plastic work done to the top frame for key fitment.  DW asked why there is such a wide gap between the keys on the left side, and I reminded her it's so the keys will fit the natural spread of my fingers.  My fingers.



By the space and zero keys, I changed the slope of the edge of the frame, because the frame slopes up to the keys and this made a sharp ridge that bites into a thumb as the keys are pressed.  This being uncomfortable, I knocked down the edge.  I melted and shaped it down with the hot knife, and sanded it smooth.  It's ugly.  That means it matches the rest of the board I guess.   I may end up having to paint this thing when I'm all done, and maybe also get a color-matching set of key caps.  Maybe.  The Microsoft board looks really cool in stock form, but I intend on never having to market this board and I'm not really sure it has to look nice.  I *may* get some kind of filler in the edges, so the bottom row of keys aren't just adrift in a sea of emptiness between the frame and the keys.  :shrug: we'll see.

I think it would be cool to have a scroll wheel in the keyboard.  I have a scroll ring on my ExpertMouse and I love it.  I plan on having a trackpoint in the keyboard, and I don't like to lose functionality.  If I can get in a scroll wheel to go with the trackpoint, it will duplicate the functions on the trackball.  Here is a much-altered Logitech scroll wheel from a mouse someone recycled, in its new home in my keyboard:



This scroll wheel is a little different than in most of the mice I've taken apart.  This model has a pushbutton to allow the (metal) wheel to spin freely, or to put detents into the rotation.  It's pretty slick, but the mechanism is large and my space is limited here so this was difficult to fit.  The scroll wheel has its own daughter board inside the mouse, and that board is small - but not small enough to fit with no trouble in my keyboard.  As I don't need to use the switches on the scroll wheel, I hot glued the side switches to hold the plastic mount for the wheel steady, and cut off the end of the circuit board because it was hitting the space key.  Note that the daughter board hangs way down inside the keyboard:



There is a strong reinforced section in the bottom frame in the Microsoft keyboard bottom frame, right under this location.  Well, there was.



I think it fits now, but it is hard to know.  Not only because it's impossible to see in there . . .



. . . but also because the frames don't mate perfectly.  All the modifications I've made to the top frame prevent it laying down on the bottom frame like it used to do.  I am keen on not making a whole new custom bottom frame, so a lot of cutting-out of old reinforcements and mounting points is going to be required.  Here you can see the new mounting plate for the tenkey area sitting on top of structure that was useful for the original Microsoft hardware.  This mounting stuff is all going to have to go, to fit the new parts I've fitted on the top plate.



There!  Finally, a solid update with pictures for you   There are some nights when I'm modding that I just don't have an update in me.  I workandwork unti I realize I'm getting muzzy, and just stop working.  My brain barely has enough juice to get a single-sentence update posted, but I know I like seeing even quickie progress reports in other peoples' build threads so I put one for you people.

...81...

[njbair]

This is really coming together! Looks really good, too.

[neverused]

I see that you have already progressed quite far, but this may be of interest to you:

http://web.archive.org/web/20140703132149/http://geekhack.org/index.php?topic=12439.0

[VoteForDavid]

@njbair thanks a lot!

@neverused thanks, I have seen that thread.  I was *very* glad that the Wayback Machine saved the pictures.

&

but now: ugh.  I got the trackpoint module wired for switches and tacked in with hot glue.  It looks great.



I'll need to toss in some black foam or felt or something, and the aesthetics are pretty well spot-on for me.  The stem protrudes a good amount in a good location.  It's great.  But then I went to wire the switches to the Microsoft controller. 

I hadn't looked very closely at the controller to date.  It's a continuation of the matrix from the two sheets, and the circuitry of the keyboard matrix on the PCB is not entirely straightforward.  The switches for Forward and Back, which I would have to isolate from their traces to use as Left and Right click with the trackopint, are in the middle of strings of key switches in the matrix.  I'd have to make clean slices of the trace *around* the solder fillets in a couple of places.



I'll probably put some jumpers in there to reroute the traces, which is no big deal.  I'll also have to integrate the just-taken pictures of the PCB into my big file of the keyboard matrix.  I tried just now, and my laptop sort of sat there with a blank look on its screen.  It tried to open the file, it really did.  But I'm not sure it can handle that much data.  I'll try again at lunch at work.  Using an actual Workstation can make a big difference