Mini Keyboard with Trackpoint

[mondalaci]

sordna:

You've raised some very good points!

Indeed, although I think our design is novel and superior in many respects it's unlikely to be copied.  I still worry because even though our board is different but not in a weird way so I consider it commercially viable.

As for the straight columns layout you've made me fall in love with this idea but I'm uncertain about its commercial viability because of its unusual look.  We could try that over time, though.  Are there any such boards created or any sketches drawn?  I'd also be curious to hear the experience of people using such a board.

mkawa:

The NDA would only be valid until the board gets released.  The reason is that we're talking about a fully-fledged, professionally designed product (which is very rare in the Open Hardware world) implementing some innovative ideas and having a very unique design.  I wanna empower users by going Open Source but I don't want anybody manufacturing the board before us.

[Icarium]

As for the straight columns layout you've made me fall in love with this idea but I'm uncertain about its commercial viability because of its unusual look.  We could try that over time, though.  Are there any such boards created or any sketches drawn?  I'd also be curious to hear the experience of people using such a board.

The Kinesis Contoured and the Truly Ergonomic, the TypeMatrix is another...
Most people seem to love it.

EDIT: somebody should condense this discussion into the wiki site...it's gotten far too long to follow

[sordna]

The NDA would only be valid until the board gets released.  The reason is that we're talking about a fully-fledged, professionally designed product (which is very rare in the Open Hardware world) implementing some innovative ideas and having a very unique design.  I wanna empower users by going Open Source but I don't want anybody manufacturing the board before us.

Dude, even in the remote chance they will want to copy your design, it takes real companies years to go from idea to shipping product. I honestly think you might be hurting your project if you prevent early ideas / information exchange from the keyboard enthusiast community. That's just my personal opinion of course, it's your project and your choice. As far as the straight columns go, apart from Kinesis / Truly Ergonomic, Typematrix, there is also the Datadesk Smartboard. The Fingerworks Touchstream also had straight columns, and of course many POS (Tipro, Access-IS, etc). Straight columns are really very easy to get used to (1-2 days at most) way way easier than a different layout like Dvorak or Colemak.

[mondalaci]

Icarium: Thanks for the references.  Some very interesting keyboard designs with useful ideas of which I can definitely see myself embracing some eventually.

[sordna]

Some more amazing projects that I WISH companies would copy. But they don't give a sh*t. It's always the same same same unergonomic and outdated shape/layout for the big guys. We need THESE becoming mainstream:

http://geekhack.org/showwiki.php?title=Island:6292     /     http://www.humblehacker.com/keyboard/

http://geekhack.org/showwiki.php?title=Island:26845

http://geekhack.org/showthread.php?22780-Interest-Check-Custom-split-ergo-keyboard

[mondalaci]

sordna:

Thanks for the additional links!  I'll have a talk with my buddy (who's the mechanical engineer behind our project) about a more open information exchange towards the outside world.  Seeing how knowledgable you guys are it'd be certainly foolish to not to have a thorough conversation regarding possible improvements before product release.

Ripster:

Thanks for the advice on the font!

[cactux]

So what exactly is the hold up on this project? Is it the trackpoint? The case? Or just low demand, high cost?

that is a very good question

[daerid]

lowpoly (or somebody else) should draft up that case in some sort of CAD program.

[net2522]

Can, we bring this project up again(even op doesn't interest anymore)?

[Input Nirvana]

Mondalaci:

Look forward to hearing/seeing whatever it is you have planned. There seems to be a fair interest in something other than the "typical range of keyboards" but the answer is not just a single alternative. That's the problem. Because of that you can't really achieve economy of scale, and the alternative costs too much (relative to the "typical range of keyboards"). There are already many alternatives of matrix, split, separated, modified layouts, super compact, etc. with most being well under a $150 price point. Let's face it, in the big picture, there is no way most people will ever spend more than $50 on a keyboard unless it also washes their clothes and feeds the dog.

I have no idea what it would take for a manufacturer (Dell, Apple, blah blah) to just make a simple matrix, angled (but flat) keyboard as the new "standard". Just about the easiest and most cost efficient "change" to the conventional keyboards of the day, yet keeping it as small as possible. But I guess that would be too radical. The Microsoft Natural is probably the closest mass-produced and well-known but it isn't even matrix, and it isn't flat. Dunno how many of those are sold, but I know huge companies supply them to employees with RSI issues so they don't get sued.

I would have a strong desire to market directly to cities, corporations, and similar to promote ergo typing/input devices. I have a background with this type of marketing, but not with keyboards of course! If there was a reasonable product/price/track record, I would recognize it in an instant and be forging an added chapter on my career! LOL

Anyhow, please be sure to use GH as a sounding board, or as a place to release 'teasers'!

P.S.- I saw your blog site and some of your keyboard requirements....click my sig and you'll see where I integrated a Trackpoint into the keywell of a Kinesis

[Loligagger]

How difficult would it be to make a programmable 60% board without the trackpoint? Neither the pure or poker have a perfect fn layer layout but having it programmable would solve pretty much all problems.

[bpiphany]

How difficult would it be to make a programmable 60% board without the trackpoint? Neither the pure or poker have a perfect fn layer layout but having it programmable would solve pretty much all problems.

Not particularly =) PCB designing, and controller programming, is pretty well understood around here by now. The less developed part is the case. Litster has been doing multilayer acrylic cases though, and a few very small runs of shapeway cases and the likes have been made as well.

[mSSM]

Everytime this thread gets a bump, I get a heartattack.

I wonder - what's actually missing to get this project the final shove over the finish line? Designs are done, prototypes are produced; what's missing? Heck, I guess it should be even easy to have all those people producing custom aluminum cases for Pokers change the design a bit and make something fit the miniguru.

The miniguru is the very first mechanical keyboard I read about and the actual reason I came to Geekhack/got interested in mechanical keyboards.

[Loligagger]

Not particularly =) PCB designing, and controller programming, is pretty well understood around here by now. The less developed part is the case. Litster has been doing multilayer acrylic cases though, and a few very small runs of shapeway cases and the likes have been made as well.

Well the next question is what about making a pcb that would fit in current poker/pure cases? I'm sure a group buy for such a pcb would get a lot of interest (especially with how pokers have become harder to get nowadays).

[bpiphany]

Well the next question is what about making a pcb that would fit in current poker/pure cases? I'm sure a group buy for such a pcb would get a lot of interest (especially with how pokers have become harder to get nowadays).

I'm slowly working on PCB that is meant to fit the Poker case as well as both the full size and tenkeyless Filcos. All controller parts will be located so that the unwanted parts of the PCB can be cut off. Not knowing how to design and program everything with surface mounted loose parts instead of a Teensy is what has been holding me back. Now I think I've got all that covered though.

[Loligagger]

That's good to hear. What about things like the usb port location or dip switches? A programmable pcb that could drop into trebles poker case would get my money as soon as it would be available.

[mSSM]

That's good to hear. What about things like the usb port location or dip switches? A programmable pcb that could drop into trebles poker case would get my money as soon as it would be available.

Same here; that above idea sounds amazing. Maybe one should open a different thread for this though (PrinsValium), so that people are aware of it? I don't think this here is the correct thread for this.

[bpiphany]

That's good to hear. What about things like the usb port location or dip switches? A programmable pcb that could drop into trebles poker case would get my money as soon as it would be available.

The USB port for the Poker "section" would of course (be intended to) end up in the correct location. DIP switches are for losers who are not allowed to program their own keyboard to their liking. (And there are no I/O-pins left for them in my current design...) This project is still far away from being finished. For the Filco variants there needs to be mounting plates as well. And if it ends up in some sort of group buy I should probably find a partner over in the US for tax-technical issues =P Switches and other components might be held outside such a group buy to keep it lighter, but plates I feel should be included.

Same here; that above idea sounds amazing. Maybe one should open a different thread for this though (PrinsValium), so that people are aware of it? I don't think this here is the correct thread for this.

There will be a thread when I get closer to some sort of final design.

[mkawa]

can we move this discussion to its own thread?

ps, i would like to add a trackpoint to your design, prins. i have heard that vortex has sourced the strain gauge element and associated encoder/controller, and can start discussions with them to obtain enough for a GB.

[uberben]

can we move this discussion to its own thread?

ps, i would like to add a trackpoint to your design, prins. i have heard that vortex has sourced the strain gauge element and associated encoder/controller, and can start discussions with them to obtain enough for a GB.

I would be very interested in getting in on that GB, or at least to know the supplier of the parts.

[bpiphany]

This thread is way beyond the point where it is possible to browse through anyways...

I don't remember if I kept the track point hole on the Phantom design. The guy who wanted there to be one never came around with the exact location or size. Track points have been the slow and painful death of some other projects already, no? One thing that is for sure is that it is not going to fit into the PCB design nor are there any pins free on the controller. So in any case it would need to be a separate unit that fits through the PCB somewhere. I dont know if there would be any room in any of the intended keyboard cases either.

I was doing some work cramming the controller in earlier today. There isn't that many empty surfaces where it fits easily. There is like an infinite amount of free area on a keyboard PCB. It's just very spread out =P

[mkawa]

This thread is way beyond the point where it is possible to browse through anyways...

I don't remember if I kept the track point hole on the Phantom design. The guy who wanted there to be one never came around with the exact location or size. Track points have been the slow and painful death of some other projects already, no? One thing that is for sure is that it is not going to fit into the PCB design nor are there any pins free on the controller. So in any case it would need to be a separate unit that fits through the PCB somewhere. I dont know if there would be any room in any of the intended keyboard cases either.

we can put it on a daughterboard and have treble do the case design to fit the extra depth once the pcb sandwich is designed. the current cases on the market kind of suck anyway for other reasons.

anyway, prins, can you start a new thread for this project?

[bpiphany]

anyway, prins, can you start a new thread for this project?

No, you do it

Seriously, I will when I have any material to put there... *off to work*

[mkawa]

this thread is long dead. i'm going to close it and (eventually) archive it so that it doesn't keep getting newbies' hopes up.

[lowpoly]

(Thanks for moving the thread back. Pics in first post repaired, grrr, imageshack.)

Update: working on it again and planning to finish it this time. 

I tried two different stainless steel stick shapes, here the first, 2.8mm head diameter:



ZIF cable connection, 0.5mm pitch:



Fake photo with Filco caps:



And the current design (left). 4mm head diameter. Stick is a little bit too thin at 1.1mm:



The idea is that the stick is so thin that it fits between GHB without having to cut the caps at all. This will depend on the keycap set and how pointy the corners are. The smaller circle is 1.32mm in diameter, the bigger 1.5:



J-pointer, suggested by bhtooefr:





Will not make it to the 'board but was fun to try out. You obiously lose press-to-scroll and tapping. Requires a deep spherical cap IMO.

Also, we're currently trying to put capacitive sensors under the spacebar so you can detect where it was pressed (or which thumb). For example right tumb = spacebar, left thumb = Function, both thumbs = backspace.

[mkawa]

looks capacitive? how's the responsiveness of this mechanism vs the standard strain gauge units?

[lowpoly]

Depends on the code but it can be quite soft. My original TP IV from an IBM 8923 (?) requires much more force.

Edit: yes, capacitive.

[mkawa]

the original TP IVs are ancient compared to the units that alps now vends for thinkpads and HP units and others. would love if you could do a comparison/potentially even tune for similar response to a standalone thinkpad board: http://shop.lenovo.com/us/en/itemdetails/0B47189/460/60AC6A0372B14F5BA7B12F1FF88E33C7

still the gold standard in eraser nubs as far as i'm concerned

[uberben]

What flex sensor are you using? Is that a custom PCB? (It looks purple; OSHPark by any chance?)

[jacobolus]

the original TP IVs are ancient compared to the units that alps now vends for thinkpads and HP units and others. would love if you could do a comparison/potentially even tune for similar response to a standalone thinkpad board: http://shop.lenovo.com/us/en/itemdetails/0B47189/460/60AC6A0372B14F5BA7B12F1FF88E33C7

still the gold standard in eraser nubs as far as i'm concerned

I would be very interested to see a comprehensive review of various generations of thinkpad trackpoints, ideally also including comparisons to other OEMs’ trackpoint-style pointers.

[Chemoletter]

amazing!!! i really like it. the soldering and cablework looks complicated.

[mkawa]

i think alps is the only OEM of the strain gauge pointers, but i'm happy to be proven wrong. cap sense is like, the new thing in small micros due to the proliferation of touchscreen devices, so it's a good direction to go. that said, the assembly there looks a little finicky, both to build and to source. you should be able to do cap sense with just an elastomer disc above the lower electrode array and then a flanged conductive stick (doesn't have to be metal, just conductive).

the bigger thing is that you're going to have to run a lot of autocalibration routines to deal with the relatively large variations in absolute cap values between users elastomer tolerances, etc. the strain gauge solution is so nice because it's purely resistive, and yet pretty precise. (unicomp has a compressive pressure resistance mechanism that is easy to encode but not very precise at all..)

[FoxWolf1]

Ouch, those narrow metal sticks look like they'd be painful to use for an extended period or in high-intensity situations. You're at 4mm of metal, compared to 6mm of rubber for the narrowest real TrackPoint cap, the "classic dome" style. And the classic dome already leaves a crater in your finger, which feels rather weird until you get used to it.

I wish people would get over this obsession with changing keycaps already...it very much gets in the way of functionality-oriented innovation.

[lowpoly]

Thanks for the replies.

the original TP IVs are ancient compared to the units that alps now vends for thinkpads and HP units and others. would love if you could do a comparison/potentially even tune for similar response to a standalone thinkpad board: http://shop.lenovo.com/us/en/itemdetails/0B47189/460/60AC6A0372B14F5BA7B12F1FF88E33C7

Feels similar to my T41 Thinkpad as far as activation force is concerned. It is hard to compare a 5mm stick to a 27mm one though. The software is different too. It does need tiny pressure from above or it won't work. I think you apply this naturally when the stick is between the keycaps, just to get some friction. With the stick sticking out, you can theoretically apply force from the side alone or even from below, but it is not made for that.

What flex sensor are you using? Is that a custom PCB? (It looks purple; OSHPark by any chance?)

No, not OSHPark. It's a complete unit. Let me get my orders in first before I disclose the stick info. Shouldn't be long.

Ouch, those narrow metal sticks look like they'd be painful to use for an extended period or in high-intensity situations. You're at 4mm of metal, compared to 6mm of rubber for the narrowest real TrackPoint cap, the "classic dome" style. And the classic dome already leaves a crater in your finger, which feels rather weird until you get used to it.

It's possible to exchange the stick. The 2.8mm one is too thin and not the way to go. 4mm is fine for me. Bigger, like the traditional square, is no problem too but not necessary IMO. I was never happy with the traditional caps. The cat tongues wear off quickly and the rubber ones get slippery. And you really only need little force when pointing. More for press-to-scroll and tapping.

i think alps is the only OEM of the strain gauge pointers, but i'm happy to be proven wrong. cap sense is like, the new thing in small micros due to the proliferation of touchscreen devices, so it's a good direction to go. that said, the assembly there looks a little finicky, both to build and to source. you should be able to do cap sense with just an elastomer disc above the lower electrode array and then a flanged conductive stick (doesn't have to be metal, just conductive).

the bigger thing is that you're going to have to run a lot of autocalibration routines to deal with the relatively large variations in absolute cap values between users elastomer tolerances, etc. the strain gauge solution is so nice because it's purely resistive, and yet pretty precise. (unicomp has a compressive pressure resistance mechanism that is easy to encode but not very precise at all..)

The 1.1mm stick is too thin. I could bend it easily while in the lathe. It works though. 1.3mm should be fine. Speaking of autocalibration, I never had problems with crawling so far. Always a problem with my TP IV, sometimes on my Thinkpad (but rarely).

ION, I installed the LUFA HID bootloader to this breakout board and then hasu's tmk software through the command line interface. Nothing spectacular but a necessary step nonetheless.

[mkawa]

crawling was the autocalibration routine and triggered periodically without any feedback from the circuit (except for a watchdog timer). i'm suggesting something similar with your unit's firmware.

[OldDataHands]

No, not OSHPark. It's a complete unit. Let me get my orders
in first before I disclose the stick info. Shouldn't be long.

You, sir, have my rapt attention.

This could be the answer to the question: "What pointing stick
solution will work for the DodoHand?", or even for retrofit projects
like that pulled-off so nicely by gator456. Won't you please provide
some details? If these are available without tearing them out of
a donor keyboard it would be fantastic! In any case, I love your
longish, skinny metal extension, too. It could really ease the
integration challenge...

[lowpoly]

crawling was the autocalibration routine and triggered periodically without any feedback from the circuit (except for a watchdog timer). i'm suggesting something similar with your unit's firmware.

With my Trackpoints (IV and Thinkpad) the calibration got confused easily. Especially with longer usage I suspect them to set a new zero in the meantime but wrong of course. Then, when I stop moving it would interpret the real zero as movement resulting in crawl. This would stop after a couple of seconds.

If these are available without tearing them out of a donor keyboard it would be fantastic! In any case, I love your longish, skinny metal extension, too. It could really ease the integration challenge...

New units, not harvested.

[mkawa]

strain gauges, like clearances in capacitive devices, are not stable with time, especially when those gauges are mounted on a small diameter abs pillar. strain gauges are precise but not accurate and require constant recalibration to determine dynamic range. the trackpoint circuits had two methods of dealing with this:

1st, they were closed looped feedback systems so that harder pushes would cause the system to remap the dynamic range along the axes pushed on thus calibrating the upper bounds of the dynamic range. second, zero would be reset periodically, which caused the drift. the instructions given to the user were to stop using the pointer during the drift. you can then see how the closed loop feedback system, assuming no input, could set a zero (the lower bound of the dynamic range).

this is required for any system with tolerances as close as the trackpoints. and you can feel the difference in the precision of the algorithms in the various generations of alps circuits.

also, if you tear down a unicomp resistive circuit, what you can see is the alternative, which is that you just build the thing with gigantic mechanical and electrical tolerances and don't have any fancy control systems. you set a calibration once at manufacturing or even for an entire board series, you just write the calibration to the mask, but the result is really really loose feeling joysticks.

in fact, your little stick harness looks a heck of a lot like the unicomp mechanism, and you can pretty much assumed that the screw in it is at some point going to start backing out over time. you _cannot_ assume any static thresholds for an electromechanical device like this..

[lowpoly]

in fact, your little stick harness looks a heck of a lot like the unicomp mechanism,

The stick and clamp are just a mechanical extension.

and you can pretty much assumed that the screw in it is at some point going to start backing out over time.

This would be a mechanical failure? I know I could use Weicon Lock or Loctite on the thread but these can be a pain if you want to change something.

you _cannot_ assume any static thresholds for an electromechanical device like this..

I don't have to worry about calibration. Not sure if I understand you?

[mkawa]

thread locker will prevent you from repeatable torquing of the machine screw, as the thread locker causes large stochastic error in the torque values read by any torque control device.

all analog sense circuits require calibration, period. if the static or dynamic range of the sensed value ever changes, that calibration should ideally be a constant process, hence the existence of feedback and controls. feedback circuits allow you to maintain accuracy with low tolerances. lack of feedback requires very very high error margins to maintain output accuracy. in electromechanical circuits that sense a mechanically derived value, this means large mechanical margins. in your case this means a very loose mechanism and calibration at factory to this loose mechanism. otherwise, expect large and increasing sense error over time.

[lowpoly]

in your case this means a very loose mechanism and calibration at factory to this loose mechanism.

It's not loose. And the ASIC does the calibration.

I think I misunderstood you somewhere along the way. Thought you were saying I had to do the calibration myself.

[daerid]

Nice to see the Guru come back to life once again. Hopefully for good this time.

[mkawa]

in your case this means a very loose mechanism and calibration at factory to this loose mechanism.

It's not loose. And the ASIC does the calibration.

I think I misunderstood you somewhere along the way. Thought you were saying I had to do the calibration myself.

ah, gotcha. yep, i was just saying that one or the other has to happen, but that the asic-based closed loop calibration is by far the better option. good deal. i'm pretty excited this is coming back to life by the way.

[hemflit]

I originally discovered GH through following news about this awesome keyboard. Just the other day something reminded me of it, and checking back here I'm super happy to see that it's actually going to happen after all!


I have one little idea to propose, not so much for Lowpoly who's been working on this vision for years, but more for the next person who decides to make a different trackpoint keyboard (and who will inevitably look at this thread for ideas).

Some years back I had a thought about how cool it would be to have two trackpoints, one for each index finger, and I imagined we would use one of them for moving the cursor around and the other for scrolling. That thought came back just a couple of days ago, with two changes:

1) The two trackpoints on this hypothetical keyboard don't have to peek out between the keys. They can be on the front side of the keyboard, where the Miniguru has mouse buttons. They'd be used with the thumbs. This sounds like it could make some of the engineering easier.

2) In the last couple of years I've become a heavy user of touchpad gestures, and I see now how wasteful my old idea was, the one that one trackpoint would simply scroll and the other move the mouse. Humans have great agility and coordination in their thumbs, and combo "gestures" for two trackpoints can be way richer than that:

←→ zoom
→← unzoom/pinch
↑↓ rotate clockwise
↓↑ rotate ccw

There's more things that could do something:
- ↑↑, ↓↓, ←←, →→ - pulling both trackpoints the same way
- tapping one and then pulling it
- tapping+holding one and pulling the other
- pulling one ↕︎ and the other ↔︎

One of those would work for dragging (and at the same time text/rectangle selection) which is a hugely useful operation, and often needlessly hard with many mouse-alternatives.

The others could be used for whatever a user wants, like min/max/restore windows, back and forward in a browser, invoking some common but tiring key combo (Fn+Alt+0). Even the four I listed first could have different meanings, and scrolling could be moved to one of those new "gestures" if needed. I don't need to say that one of the points would serve as a right-click when tapped, and tapping both could mean a middle-click.

Yay, high-five myself, I'm really happy with that idea

[sordna]

Very cool ideas!

[lowpoly]

who's been working on this vision for years

Well, there were some breaks... 

Some years back I had a thought about how cool it would be to have two trackpoints

Right now the plan is to provide two additional stick positions in the case (for JKM and DFC) and two zif cable connectors on the pcb with connections to different mcu pins. So it should be possible. I don't have quotes for the case yet so we may have to simplify. Hopefully not.

1) The two trackpoints on this hypothetical keyboard don't have to peek out between the keys. They can be on the front side of the keyboard, where the Miniguru has mouse buttons. They'd be used with the thumbs. This sounds like it could make some of the engineering easier.

Maybe with another stick unit. The one I have is integrated with the pcb and it looks like it will be too wide for that.

←→ zoom
→← unzoom/pinch
↑↓ rotate clockwise
↓↑ rotate ccw

There's more things that could do something:
- ↑↑, ↓↓, ←←, →→ - pulling both trackpoints the same way
- tapping one and then pulling it
- tapping+holding one and pulling the other
- pulling one ↕︎ and the other ↔︎

Nice. You can also put more than one function on a single stick by reading pressure (which is how press-to-scroll works, edit:I think this is what you mean by holding). Precise tapping isn't easy and if done sloppy the firmware can confuse it with short press-to-scroll or movement. Pulling doesn't work with the units I have. I think the stick firmware filters it.

[kriminal]

well well well....

[lowpoly]

It's not what you think.

We've been working on it since. 3+ guys (until end of May at least, less after). Daily full time except weekends. There's just not much to update because we're in the middle of development testing circuits, doing changes to the case. I finished the designs for the host software and development has been started on that too. I'll order the first test pcbs in early June unless we don't run into some unforeseen problem (which can always happen).

[lowpoly]

First pcbs (controller board) came today:



Not likely that these will work out of the box.

Got a lot more because of over production, 14 instead of 5.

[vivalarevolución]

First pcbs (controller board) came today:

Show Image

Not likely that these will work out of the box.

Got a lot more because of over production, 14 instead of 5.

Sexy.  Great to see some progress.  I hope this becomes a thing. 

And thanks for changing your avatar.  That screen shot Grand Theft Auto or whatever was kind of distracting.