Re-Create the DataHand - Thumb cluster under development. Project 75% done.

[JesusFreke]

Oh, nice!

I loved assembling my 2 prusa's. The assembly instructions are top notch, and I love that you can comment on individual steps, so you can see what kind of problems others have run into, to know what to avoid, etc.

I'll see if I can throw up my big slic3r .3mf plate file for most of the smaller pieces. It has a few non-default print settings to dial-in the printed parts. I think I have an almost-ready plate file for the clusters too, but iirc I had run into an issue when I had tried to print it to verify the settings were correct, and then got busy with something else and hadn't got back around to fixing/re-verifying it.

[JesusFreke]

Here's the .3mf file for the keys, etc. I'll try and get the one up for the clusters in the next day or two.

https://github.com/JesusFreke/lalboard/commit/0b3cd0d5322d0ab87f838aaef26e1346cf8cfd71

[ironfox]

Thanks for the .3mf file, although in the end I had to split it anyway, since it simply would have taken too much time to print in one.
The first partial print appeared to have come out fine:

But it turned out I had the Z-setting too close to the printing surface and everything is a little broader than it should have been. I probably have to reprint most of those.
After some inexplicable disasters

I gradually got better, although I can't seem to be able to get rid of the stringing entirely.
At any rate, after much more printing
,
I started experimenting with the first (thumb) cluster
.
The idea with the ball screws is both ingenious and improvable.
Ingenious because it allows maximum flexibility and makes use of printed screws the way I thought'd never work. The caps with the spherical space in-between work surprisingly well.
Improvable because the bottom screw part

simply does not.
The larger screws I cannot print because they keep falling over. The smaller ones complete but refuse to be inserted into the sockets more than a turn or two.
Using force turned out to not be a good idea. My first attempt to print them using ABS instead of PLA also didn't work particularly well. I got a heap of plastic. I think one of the settings may have been wrong but that was definitely not a screw I got there. So I decided to approach the issue from a different angle.
A nephew of mine knows how to work metal and he quickly agreed that this could be done using metal parts, and, even more awesome, those parts partially already exist:
.
They may need some adjusting but generally should fit. The foundations can also be bought with the right diameter:

Just need some cutting and drilling.
Now, admittedly, this will be a different beast, entirely, but I am reasonably certain it will prove a much more resilient solution than even ABS screws would have been.

I'm also starting to understand how this construct is supposed to work. Quite impressive. Am I still to heat the magnets to 200 degrees Celsius to adjust the force?
They are sometimes quite a lot of fiddling to get into those slots. Any tips and tricks I can use?

Anyway, nice work indeed. Lots of engineering effort there and I can definitely see a lot of experience at work in those designs.

[ironfox]

So, after the first thumb cluster eventually broke trying to plug a magnet in, here is attempt number two. As you can see I made some modifications to the design to strengthen it (some more so than others).

The magnet holes are 10% larger, the sensor holes 30%, which turned out to be overkill. 10-20% is probably good for the sensors. The magnets now hold the thumb-down-key properly up (and return it there when released), same goes for the horizontal key. You are right, oriented correctly the magnets are a little strong there (seems fine for the down-key). Rotating both by 90 degrees seems to have reduced it sufficiently, though. They have less attraction that way. Maybe a little too much. Have to see

[JesusFreke]

Sorry for the late reply, I somehow missed the thread update email. Not sure if it wasn't sent, or it went to spam or something.

In any case, cool stuff!

Yeah, if you need to split it up you can just import the .3mf and then reduce the part count or remove some of the parts. That way it keeps all of the per-part settings.

I used polycarbonate (polymaker PC-Max) for the ball-head screws. They are sized so that, when printed on my printer at least, they screw into the base tightly enough that there's no play, but aren't too difficult to turn. But each printer prints a bit different, so you may need to adjust the size, etc.

I agree that they are difficult to print though. Probably the most difficult part to print of the lalboard. I ended up having to use a raft to make sure they were able to stay adhered to the bed while printing.

Yeah, metal screws would be great! I wasn't able to find any off-the shelf ones that were readily available, although if you look at alibaba, it looks like it's possible to get some made. And I don't have easy access to a metal-working shop to try to make them myself.

For inserting the magnets, I love these parallel jaw pliers. In some cases, there's not enough room to use those, and I used these needle-nose pliers with tiny jaws instead.

In either case, the jaws are steel, so you can attach a magnet to one side of the pliers, maneuver the jaws into place over the hole and use the force of the pliers to insert it into the hole. I prefer using the parallel jaw pliers when possible, because they're less likely to damage the parts.

The holes for the magnets are sized so that it should be a tight friction-fit, although I have had a couple of magnets come loose after a couple of months of use. If/when that happens I just re-insert them the same way, with tiny dab of superglue to make sure they don't come out again.

The only magnets that need to be partially demagnetized are the ones for the central down key in the finger (not thumb) clusters. The corresponding one in the base, and for all the other keys can be full strength. It's easy enough to reprint/replace the center keys, so you can experiment with them much easier than having to replace the magnet in the base. 200C worked well for me, but YMMV . iirc, I got the center key force down from ~40-50g with the "raw" magnets, down to about 25g with the "cooked" magnets. 25g felt about right to my experienced datahand fingers.

I would suggest skipping that and just using the full-strength magnets for now. Once you get everything working well, you can reprint just the central down keys and experiment with the magnet strength to find what feels best.

Also, I finally got up the .3mf file for the finger/thumb clusters. I just printed it last night to verify the settings work well. I'll try and do the one for the polycarbonate screws next.

 

[JesusFreke]

For those needle-nose pliers, it looks like haako makes a pair that might work too. I haven't tried them, although they look pretty similar to the xuron ones I linked above. And I do love the haako flush cutters.

[ironfox]

Hey, thanks for the hints. I slowly and steadily figured a routine to handle those magnets, too. For starters, like you I realized without some form of parallel pliers, you're plain screwed. I got these:

I also started paying very close attention to the magnetic orientation of the magnets. Typically I would insert magnets into new keys such that they would be aligned exactly as those I already had. I let magnets fly towards inserted ones, then mark the off-side with a black marker. If the magnet was to go into another key, the marked side would have to be front, otherwise disappear into the slot. I also started using those thin files I have to navigate the magnets into place. Doing what magnets do best, they would stick to the files until there was enough friction to keep it in place. Then I would finish the job with the parallel pliers. The thing I linked may not look the part, but it gets pretty much everywhere it needs to.
After much fiddling, and also much filing on those center down keys, eventually all parts of the right hand worked magnet-wise:

Packed everything up and sent a package so that the bottom plate and foundations can be made. 
Even printed a quick-to-print flat version of what I had assumed to be the right rest (since it was kinda labeled that way):
.
Except, when the full print finished just now after some 7 hours of printing

it dawned on me that this is, in fact, the left rest:

I also realized I will have to adjust the size a bit. I'm guessing 20% shorter and flatter might do the trick. Width seems right. I appear to have tiny hands....
At any rate, just ordered more filament as I expect there will be much more printing on the road...

[JesusFreke]

Yeah, the center keys are the hardest to get working right. I went through many iterations to get them to the current state. Iirc I usually smoothed out the flat face on the key post, and cleaned up the bottom and top edges of that center hole in the base. The problem I ran into most often with those keys is that when pressed, they would get stuck down and wouldn't pop back up by themselves. It took a bit of tweaking to get them working reliably, but I haven't had any problems with them since -- even after some heavy use during gaming.

And yeah, it's best to pay attention to magnet orientation, and always try to use the same orientation for all the parts, so that they are interchangeable. I use a slightly different technique, but whatever works


Oops! I'll take a look at the handrest files and fix that. I must have gotten them backwards. They're just mirror images of each other, of course. I was curious how well they would fit someone else's hands. Thanks for the feedback I think that may just be something that each person will need to tweak for their own hands.

If you're feeling adventurous, you can do what I did and mold your own handrests from some clay to fit your hands exactly, and then use photogrammetry to scan it in (see, e.g.

). I'm a big fan of "custom" fits like that. e.g. I also have custom-fit in-ear-headphones (so comfortable! I can't go back to normal in-ears now), and I modified my coffee cup and computer mouse with some sugru to fit my hand exactly...

Good progress though! Definitely keep us updated.

[ironfox]

Great idea with the photogrammetry. 
I modeled a hand rest using modelling clay, which does look a little different from yours but seems pretty comfortable in clay. Height and angle will have to be adjusted once I have the base and parts back.

To check how well this works, I used MeshRoom and converted it to a 3d mesh. It rejected half my photos but the other half was apparently enough. Sadly didn't make any screenshots. Will provide them the next days. Suffice to say, the scan seems good enough but I will model a clean mesh such that it matches in the major features. The scanned mesh is too messy to fix it seems. At least, I couldn't find any filter that managed to clear out the noise, unwanted details, and errors but keep the major parts as they are. The geometric resolution appears to be too non-uniform. Nonetheless, I am confident I can model this to match. I'll probably make the rest two-part. Top cover and bottom case to fit electronics into. That way, I can reprint or independently clean the cover it need be  . I can probably also skim on the printing time but won't be able to complete this before I have all the parts to fit.

In the meantime I started reading into how your PCB's are supposed to work. I think I got the gist but have a couple questions:
1) Could it be the central PCB SVG is flipped? That would make somewhat more sense to me.
If so, It seems the IR emitters are halted by providing Vcc on both Vcc and GND thus preventing any current?
It looks as though cluster-Vcc is permanently provided, but cluster-GND is only actually GND if the LED driver is configured to link to GND, otherwise it also links to Vcc.
2) The lines between the two Teensy's puzzle me. You are basically using four of the seven possible lines it would seem. If flipped, D0 and D1 connect to either each other or reflected: D0 of one Teensy to D1 of the other and vice versa. What I don't get is why Vcc and GND are also part of that bundle. Either they each connect symmetrically, which makes limited sense, or flipped Vcc->GND which makes even less sense. What is the intention here? Is one Teensy supposed to provide the other with power if only one is connected?
Also, why are there 2k resistors between the Vcc line and D1/D0? Shouldn't those resistors connect to GND?


I will first attempt to make this run at all, then figure out how to make it smaller. I did eventually dawn on me I would need some sort of automated Vinyl cutter to get the PCBs the way you did, so I'm probably not gonna do that. Making actual PCBs has so far also eluded me, considering I have near zero experience with those programs and don't even know which ones can import those SVGs with exact dimensions. So far it seems ... none
Contrary to everything else, I seem to only be able to find people who cannot do it, do not have the time for it, or both. The odd shaping of the components they are supposed to fit to make this surprisingly hard. The central part should be easier but is also of limited concern right now because even just with wires it should be compact enough to fit under the rest.

At any rate, everything is moving forward. Electronics experimentation parts should get here by tomorrow. Then I'll start plugging some parts in and see what happens

[JesusFreke]

The handrest is looking good! Iirc, I think I mostly used meshmixer to clean up the mesh afterwards. It has some smoothing tools that helped smooth things out.

I did also experiment with remodeling the handrest using fusion 360's sculpting functionality, but I ended up using the cleaned up mesh instead.

> 1) Could it be the central PCB SVG is flipped? That would make somewhat more sense to me.

The svg has the correct orientation, when viewing the central pcb from the bottom.
See, e.g. https://photos.app.goo.gl/RYPwcSMWUfKyigzVA

> If so, It seems the IR emitters are halted by providing Vcc on both Vcc and GND thus preventing any current?
> It looks as though cluster-Vcc is permanently provided, but cluster-GND is only actually GND if the LED driver is configured to link to GND, otherwise it also links to Vcc.

Yes, mostly. The TLC59211IN is a sink driver, which means that it goes on the "ground side" of an LED. When a channel is "turned on", it connects the output pin to ground. When that channel is "turned off", the output pin is left floating. It is never connected to Vcc. If you look at the datasheet, the logic table says input low corresponds to output high, but with the caveat that that's with a pull-up resistor. There is no pull up resistor here, so the pin is just left floating, as if it weren't connected to anything.  At least, that's the way I understand it. I'm no EE though

>The lines between the two Teensy's puzzle me. You are basically using four of the seven possible lines it would seem. If flipped, D0 and D1 connect to either each other or >reflected: D0 of one Teensy to D1 of the other and vice versa. What I don't get is why Vcc and GND are also part of that bundle. Either they each connect symmetrically, which >makes limited sense, or flipped Vcc->GND which makes even less sense. What is the intention here? Is one Teensy supposed to provide the other with power if only one is >connected?
>Also, why are there 2k resistors between the Vcc line and D1/D0? Shouldn't those resistors connect to GND?

The connector for the cable between the 2 sides should be wired so that the same pin on each side is connected. So one connector needs to be wired "opposite". I should probably mention something about that on the BOM. The other wires, between the central pcb and clusters, are all wired normally.

And yes, it only uses 4 of the 7 wires. I decided to stick with the 7-pin connector to keep things simple, i.e. 1 or 2 less types of parts to source .

Only the right side needs to be connected to the computer via USB. The left side is essentially the "slave" side, and is powered by the right side. It has no USB functionality itself, and doesn't need to be plugged into USB.

The 2 sides communicate via i2c, which requires a pull-up resistor on the data and clock lines. That's what those 2k resistors on the 2 data lines are for. They should only be populated on one side, but it doesn't matter which.

[JesusFreke]

Once you start doing the electronics, I did discover one tip you may find useful. Since the LEDs are IR, you can't see whether they're turning on or not... but your phone camera likely can! The IR light shows up as a pale purple glow. That's useful for testing, although in actual use the LEDs have such a low duty cycle that it's hard to see anything even on a phone.

[ironfox]

Just a smallish update
Lacking an automated vinyl cutter or matching PCB, I had to come up with some alternate solution, as I have stated before. My current workaround for the clusters looks a bit like this:

The bottom ABS plate is completely covered by copper tape and leaves out pins not connected to the VCC line. I experimented with different shapes to see which is easier to cut out. Straight lines seems better.
The second ABS plate has openings for the solder points of the first one and allows soldering individual pins to wires without affecting the first one.
To my total surprise that idea actually worked.
A small resistor network in each cluster reduces the soldering effort considerably. The next generation will have extra spacing for that chip ... and maybe proper PCBs.

So, with all ten clusters properly soldered it was time to move on.

The breadboard mess almost worked. Sadly, one of the cluster keys doesn't work, so it has to be resoldered. And in the process of that, I started reconsidering how to do the central pieces. Figuring that neither the original vinyl cut concept nor this breadboard solution were gonna fly, I revisited the PCB option, and after much fiddling around with EAGLE, and requesting component libraries left and right, I eventually, maybe, hopefully succeeded:

Ordered a couple of those online, so let's see how that works out. I'll upload the designs somewhere as soon as I know they work. Much like the cluster, I opted for resistor networks here. I hope I wired it all correctly. The pins for the connectors are different, but that should be easy enough to fix on the plugs. The boards are 40x52mm each.
Knowing the size of those boards, I started working on a concept to hold them in the hand rests. I figured since I already have so many magnets, I might as well use those:

It's rather firmly held in place. So much so that I added more finger holes to get it out with less effort in the next design iteration.
So, yeah, lotta work still left, but some parts started actually working...

[JesusFreke]

Nice!! Awesome progress, and interesting to see all the tweaks

It looks like you found and sourced some actual metal ball-head screws? And you machined the supports they screw into out of aluminum?

You're probably going to want to add magnets on the bottom of the hand rests. If one of my hand rests accidentally slides or moves, it's a bit annoying getting it back into *exactly* the right spot. For me, a combination of the magnets on the bottom, and a little bit of liquid rosin on the base plate to increase the friction does the trick.

[Lord of Narwhals]

Just stumbled upon the Azeron gaming keypad which seems to be inspired by the DataHand. It doesn't have the east and west clicks for each finger but it has an extra row of keys above the north ones. And since it's a gaming keypad it's not designed for typing and only has 26 keys per hand.
Has anyone here tried it? Even if it's not a perfect substitute for the DataHand I'm glad that that kind of design is starting to gain some traction.

[wolfv]

Thanks for posting Lord of Narwhals.
azeron looks interesting.
I asked them what kind of switches are under the keys.  I will post if they respond.

[wolfv]

Wow, azeron replied fast.  They are micro switches, similar like in a mouse.

[wolfv]

azeron creators have a thread on https://geekhack.org/index.php?topic=93039.0

[JIH]

been watching this for years. will start pitching in.
I was looking at the "Ball screw" and thought this might work out
https://www.aliexpress.com/item/10pieces-lot-M3-M4-M5-M6-M10-M12-Brass-female-ball-head-hanging-board-ball-nut/32905344632.html?spm=2114.10010108.1000014.1.599240afOcmp7U&gps-id=pcDetailBottomMoreOtherSeller&scm=1007.13338.136569.000000000000000&scm_id=1007.13338.136569.000000000000000&scm-url=1007.13338.136569.000000000000000&pvid=82c6e091-14db-4bdb-a14a-1c181b1b26d9
something like this anyway with some drill rod and coupler nut https://www.mcmaster.com/hex-coupling-nuts.

[wolfv]

Nice find JIH.

Also the ball end of ""ball link" or "rod end bearing" used in radio-controlled aircraft and cars.

[JIH]

Couldnt find one with Hex. dont know the thread size
https://www.banggood.com/Ball-Short-Screw-For-PRC-118-Crawler-QX-4-Remote-Control-RC-Car-Parts-p-1270445.html?rmmds=detail-left-viewhistory__2&cur_warehouse=CN

[ironfox]

Okies, time for another update.
Also, sorry for the longer gaps between updates. I sometimes feel like waiting until more stuff works out. So here goes.

First off:

Once you start doing the electronics, I did discover one tip you may find useful. Since the LEDs are IR, you can't see whether they're turning on or not... but your phone camera likely can! The IR light shows up as a pale purple glow. That's useful for testing, although in actual use the LEDs have such a low duty cycle that it's hard to see anything even on a phone.

Awesome. That worked. At least in proving my point. And checking the soldered board LED connections.

It looks like you found and sourced some actual metal ball-head screws? And you machined the supports they screw into out of aluminum?

Well, credit for that goes to my nephew, Gereon, who now also ended up doing the soldering. But yeah, that part worked pretty neadly.

So, right. As I stated earlier, I tried my luck with PCBs, send the plans to the internet and, lo and behold, got some stuff back:

After soldering, the result looks rather nifty:


In the meantime, I continued my quest for the design of some useful hand rest. It took several iterations (skipping a few here):

That last (current) one now uses round magnets on the bottom of the board clamp, and those rectangular 1/8"*1/16" magnets of the original BOM in the clamp itself.
With all the experimentally oversized cables, it now looks like this:

I also use a whole lot of round magnets on the bottom, coupled with a near-complete plate of rubber:

Each of those round ones is about as strong as 2-3 of the rectangular ones, so that's a whole lot of magnetic power. So much so, I had to buy stronger glue.
Still, without the rubber it will still slide, as you mentioned. Polished steel on neodymium apparently has little traction.
Generally I would recommend the rubber solution. Only, it kinda stinks... :S


So far I managed to test the bus between two teensies and it seems they talk to each other just fine . The LEDs of at least that one hand also glow as they should but nothing more was tested. I managed to fry my multimeter. As it turns out the 2 something Ampere the Teensy puts out is a little too much for the mA setting. But I need the mA thingy for everything that is not the Vcc line.
So, anyway, I intend to now move on to the software side of things.
Speaking of which, where is the code I can install on these devices? I went to some length as to not change the connectors on the Teensy itself, so whatever you're using should work out of the box. With some luck. Maybe...

Show Image

Just stumbled upon the Azeron gaming keypad which seems to be inspired by the DataHand. It doesn't have the east and west clicks for each finger but it has an extra row of keys above the north ones. And since it's a gaming keypad it's not designed for typing and only has 26 keys per hand.
Has anyone here tried it? Even if it's not a perfect substitute for the DataHand I'm glad that that kind of design is starting to gain some traction.

Man, I gotta get me one of those. If they ever sell again. Who'd have thought people'd buy that in droves. Funky

[JIH]

Everyone blows a fuse in a ameter/dmm at least once. Take the back off and replace the smaller of the two fuses. Remeber not to measure across a voltage source in the amperage setting, always break the circuit and complete the circuit with the leads. In contradiction, an interesting way to find a shorted component is to place one lead at least one component from the voltage source and short each component in the chain. When you find the component that doesnt cause an increase in current you have found the short. You probably knew all this but didn't want you to trash a meter for a bad fuse.
I'm loving the use of magnets!

[iso]

Rough modelling and quick proof of concept testing, my contraption for the finger cluster.
Uses 4x5mm SMDs that also provides the bounce-back-to-position and one central round tactile. Travel distance is little over 1mm.

[JesusFreke]

Gah, geekhack stopped sending me email notifications for this thread for some reason.

The threaded balls + threaded rod + coupler nut looks like a great solution.

Those PCBs are looking great! For the software, I've been a bit lazy and haven't gotten around to polishing it up and publishing it - mostly because no-one has needed it (yet).

You've put some fire under my pants though, so I'll try and get that done here pretty quick

[JesusFreke]

Regarding the magnets slipping on the steel, I dissolved some rosin powder in isopropyl alcohol and applied some of that on the steel under the clusters/handrest. Once dry, the thin film of rosin on the steel provides plenty of friction to keep things from sliding around inadvertently.

[HeuristicBishop]

I bought a 3D printer for the sole purpose of kiting of few of these keyboards out for programmer friends. They've got some really unfortunate wrist and joint issues and I think these will go a long way towards offering some relief. Thanks for everybody's effort in making this a real thing.

Sidenote: If you're looking for a nice way to get these things to stick then check out microsuction tape: Sewell Direct AirStick Microsuction Tape by Sewell, 0.8mm, 250mmx300mm Sheet

One side of the tape is a normal adhesive for sticking to the bottom of the handrest, the other side of the tape uses tons of microscopic suction cups to "stick" things down without requiring any adhesives.
It holds really well against movement in x and y planes but you can peel it right up with some force in the z.
It's cheap, effective, reusable, repositionable, and leaves no residue. Some really wild stuff.

[ironfox]

Okies, first version not even done yet and I'm already planning out the next generation. I've been thinking.
What I would like to see:

• Portability: Get rid of the metal plate and create something that at least behaves like a single object
• Assembly: Provide PCBs for all clusters
• Adjustability: Maintain/restore enough flexibility for a large range of hand sizes

My first goal would be to get rid of the feet holding the regular clusters up. Even with the metal feet I managed to source (and a lot of people won't be able to) and the stronger magnets, it still feels unstable/messy to adjust. This solution introduces flexibility I found the normal clusters just don't need. Some height and position changes may be necessary but those would be along two axes at most. What we have right now is more like 6-DOF.
I have some rough ideas that just might work:
I intend to create three to four rails stretching out from the hand rest with magnets as close as possible to snap the clusters into horizontal position. Three might be enough if ring and middle finger can be mapped using just one joined cluster. Not sure if that fits everyone but I found those have to be super-close in my case. The finger clusters have counter-magnets but also allow putting a spacer part vertically between the rail and the actual finger cluster. The spacer part is fixed height but there would be a multitude of different heights to choose from.
Just an idea. It needs a lot of trial-and-error to get right.

The PCBs for the fingers are much more work than meets the eye but I think I know enough about EAGLE to now get it done. But that whole process also made me think that maybe the magnet-solution is not so awesome for the central down-key. Even filed to perfection it feels kinda off to me. I was thinking maybe a regular keyboard switch would be better there? Kinda similar to iso's idea.
Anyone know any good ones?

[JesusFreke]

Yeah, 6-DOF was an explicit goal for me. I was never quite happy with the adjustability of the datahand - I could never get things quite where I wanted them. I do agree the adjustment is a bit of pain though. It took me a while to get them adjusted just right. But once I did, I haven't had to mess with it.

It definitely doesn't feel unstable to me though. The magnets I use on the posts have plenty of pull force to keep them down, and after applying rosin on the steel, they don't slide at all - except maybe the thumb cluster sometimes. It's easy to press too hard on the outward keys, especially the lower one that you hit with your knuckle.

PCBs would be great. I was kinda thinking of trying my hand at designing some, but I decided to use the vinyl-cut copper while prototyping, because I knew I would be iterating on the design quite a bit. And then.. well, it worked well enough for me and I never got around to trying to figure out how to design the PCBs .

The central down keys are probably the thing I'm least happy with, in terms of assembly, and to a lesser extent in terms of performance. In terms of assembly, both the hole in the cluster and the post on the key usually require a bit of cleanup, which can be a bit.. fiddly. Although I was able to get them feeling right, at least to my fingers (which have had many many years of experience on a datahand ).

And in terms of performance, I do get the occasional stuck key, although it's rare enough that it's not too much of an issue.

I think the design for central keys was probably the aspect of the design that I iterated on the most. Although I don't particularly want to use normal keyboard switches, so I didn't explore that avenue at all. Many many iterations with magnets though. Still room for improvement, but I'm... eh, 95% happy with them


Looking forward to seeing what you come up with!

[ironfox]

Yeah, 6-DOF was an explicit goal for me. I was never quite happy with the adjustability of the datahand - I could never get things quite where I wanted them. I do agree the adjustment is a bit of pain though. It took me a while to get them adjusted just right. But once I did, I haven't had to mess with it.

It definitely doesn't feel unstable to me though. The magnets I use on the posts have plenty of pull force to keep them down, and after applying rosin on the steel, they don't slide at all - except maybe the thumb cluster sometimes. It's easy to press too hard on the outward keys, especially the lower one that you hit with your knuckle.

The legs are my biggest issue. When I move the clusters, the legs always go into some other direction and then the clusters move as a whole and I have a whole lot of legs to place upwards again. I am somewhat hesitant to screw them tighter as I fear the plastic will not like that.
It seems like if the clusters were already pre-aligned and flat on the surface with a single foot, it'd go way easier.
But I'm gonna postpone replacing those until I'm happy with how the rest behaves.

Speaking of which, it seems the thing now works:

Just ignore the cables for now 
Seems I flipped the finger slots on the PCB, so if you're about to use those, just flip the order of the non-thumb keys.

I'm training like crazy right now. For some reason, even using the original QWERTY-like layout, my finger memory does not kick in. Ah well, is only like the third time I'm (re)training some layout.

The central down keys are probably the thing I'm least happy with, in terms of assembly, and to a lesser extent in terms of performance. In terms of assembly, both the hole in the cluster and the post on the key usually require a bit of cleanup, which can be a bit.. fiddly. Although I was able to get them feeling right, at least to my fingers (which have had many many years of experience on a datahand ).

And in terms of performance, I do get the occasional stuck key, although it's rare enough that it's not too much of an issue.

I think the design for central keys was probably the aspect of the design that I iterated on the most. Although I don't particularly want to use normal keyboard switches, so I didn't explore that avenue at all. Many many iterations with magnets though. Still room for improvement, but I'm... eh, 95% happy with them

I get a very different haptic feeling comparing the central down key with the other keys.
However, I compared it to regular flat keyboard and it has much less force, still, so replacing it with a regular switch probably worsens the issue.
I'm not entirely sure how to proceed.

Btw, how come the down-key has this weird, hard to file layout?
Why the thinner part in the middle, and using an extra part to block the light rather than having the IR tube go through the block above or below the magnet?
There has to be some reason behind it?

[JesusFreke]

Oh nice! I think that's the first working version anyone else has made, at least that I've seen.

The legs are my biggest issue. When I move the clusters, the legs always go into some other direction and then the clusters move as a whole and I have a whole lot of legs to place upwards again. I am somewhat hesitant to screw them tighter as I fear the plastic will not like that.
It seems like if the clusters were already pre-aligned and flat on the surface with a single foot, it'd go way easier.
But I'm gonna postpone replacing those until I'm happy with how the rest behaves.

I have the nuts for the ball socket joint clamped down pretty tight, except when I'm adjusting the leg heights/cluster angle. Tight enough that they're pretty well locked into place. The plastic threads on the nut are pretty beefy, and should be able to handle a decent amount of torquing down on the nut. You might try printing a spare cluster and nut, and really crank down on it until either failure or you can't tighten it any more, just to get an idea of how tight they can go.

Although, I'm using the printed ballscrews, so it's a plastic<->plastic interface, which probably has more friction than a plastic<->polished metal interface with the metal ballscrews.

Speaking of which, it seems the thing now works:

Show Image

Just ignore the cables for now  ;)
Seems I flipped the finger slots on the PCB, so if you're about to use those, just flip the order of the non-thumb keys.

I'm training like crazy right now. For some reason, even using the original QWERTY-like layout, my finger memory does not kick in. Ah well, is only like the third time I'm (re)training some layout.

I get a very different haptic feeling comparing the central down key with the other keys.
However, I compared it to regular flat keyboard and it has much less force, still, so replacing it with a regular switch probably worsens the issue.
I'm not entirely sure how to proceed.

Btw, how come the down-key has this weird, hard to file layout?
Why the thinner part in the middle, and using an extra part to block the light rather than having the IR tube go through the block above or below the magnet?
There has to be some reason behind it?

"hard to file" - you mean the second little nub and the corresponding hole, where the LEDs for the central key are? I don't think I've needed to file that. The nub shouldn't really be touching the smaller "well", and shouldn't really affect the feel or mechanics of the key.

The central hole does usually need some cleanup, it's a square through-hole, so it's not that difficult to get in there to clean up. And iirc, I mostly just slightly chamfer the top and bottom edges of that hole with a utility knife, I don't think I do much else to it. Although it's been a while since I assembled them, so I can't recall for sure.

I didn't want to put a hole in the sides of the central hole/in the side of the central post. Any extra features like that would likely make it even harder to get a clean fit.

I'm happy with how the extra nub and optical path blocking works, it's mostly the central hole and central post that are the tricky parts to get right.


It's been many many years since I had to try to learn the datahand, but iirc my muscle memory transferred relatively well - other than the few keys I ended up remapping, hah! The thing that took the longest for me to learn was all the extra stuff. Numbers, function keys, cursor movement (arrow keys, home, end, and all the various modified combinations, etc.).

[JesusFreke]

Oh, I just realized what you meant by "thinner part in middle". The thinner part on the key post? Yeah, I did that to help make sure that part of the key post doesn't touch the sides of the hole. If I remember right, I was getting some surface irregularities there in the middle area, even on the sides and back, due to the way the slicer handled the hole for the magnet. So I made that area a bit more narrow to ensure that it doesn't touch, so that any surface irregularities don't affect the how the key moves/feels.

That's partly why I didn't want to add another though-hole through the central post for the LED optical path - more opportunity for surface irregularities in the printed surface of the post and the sides of the hole.

[ironfox]

Oh, I just realized what you meant by "thinner part in middle". The thinner part on the key post? Yeah, I did that to help make sure that part of the key post doesn't touch the sides of the hole. If I remember right, I was getting some surface irregularities there in the middle area, even on the sides and back, due to the way the slicer handled the hole for the magnet. So I made that area a bit more narrow to ensure that it doesn't touch, so that any surface irregularities don't affect the how the key moves/feels.

That's partly why I didn't want to add another though-hole through the central post for the LED optical path - more opportunity for surface irregularities in the printed surface of the post and the sides of the hole.

Ah, that makes sense.
I don't much see this in my prints. What I do see is that my printer seems to make everything a little wider and so the unmodified design was so tight you'd have needed a hammer to get the center key back out without first filing.

Speaking of which, I had some ideas the better half of last night and put them into action:

Put together it makes a decent switch:


The general idea is that rather than two, it uses three magnets.
Here's the schematic:  switch.pdf (7.31 kB - downloaded 139 times.)
Red=magnets, blue=key, yellow=spacer, green=cluster.
The spacer is held in place by the cluster magnet and the bottom ridge. It can only ever move up.
The key is held in place by the spacer. Since the only effective pull comes from the spacer, it cannot move any farther up than the ridge between them allows.
Still, you can pull both out if you need. They will not pop out on their own, however, even when upside down. Without the third magnet in the cluster (or glue I guess), this solution can create flying keys, though. Been there, done that .


(this one wasn't filed yet, so it stuck like that)

By displacing the two magnets I can now use your idea from virtually every other magnet to also reduce the force of this one without having to resort to baking them.
This design also has extra tubes so I can get the magnets back out when I'm done.

Still needs some fine-tuning, but generally it seems to work

I did some experiments with different positions of the magnet in the key itself:

From left to right the activation force obviously goes up. How much? Let's see.
By counting the most coin weight I could stack on top before it gave way, the weight measures somewhere around:
1) 18g
2) 25g
3) 55g
4) 80g
That first one still gets stuck from time to time, but that's some seriously low force. Geometrically speaking, it is not the least possible but the friction is already a big issue here.
Also, take these numbers with a grain of salt as this is the most I managed to stack, possibly far from the average or even the actual maximum, and I cannot say how much of that is actually due to friction differences between the printed shapes (and the filing).

[iso]

PLA on PLA and pretty much any 3d printable non exotic plastic will always stick, fatigue or wear out fast, and depending on what temperature you print at, the part will always be little big/small, printing hot will make the part shrink more when cools off. Also printing thin or thick walls will influence how much is gonna expand/shrink

[ironfox]

PLA on PLA and pretty much any 3d printable non exotic plastic will always stick, fatigue or wear out fast, and depending on what temperature you print at, the part will always be little big/small, printing hot will make the part shrink more when cools off. Also printing thin or thick walls will influence how much is gonna expand/shrink

Makes sense. Know any printable exotic plastics with less stickiness?
On another note, what force is needed for your buttons? I checked out SMD tactile switches but they had forces in excess of 100g, and those were the low-force ones. Is the primary reason, I came up with my weird design there

[iso]

Carbon fiber infused got very good stiffness and wear resistance, I believe you can find samples (5~10 ft) on eBay/Amazon for fairly cheap, that should be enough to give you an idea of material you want to use without spending $50+ right away. Next down would be PETG

For you to get some decent results, no matter what material you gonna print with, re-design as needed to have at least 3mm of wall everywhere, Polycarbonate filament beats both carbon pla and PETG but is kind of bendy,  same with nylon, you cant use these with thin walls, so, up to you how you want to approach this.

The nozzle will have to press down on material so the hot new layer of plastic will make contact with the already in place colder layer, meaning the external wall will never have a smooth surface, even when you print at 0.05 which is pretty much the limit of cartesian/delta printers. You could sand off the interior/exterior walls where the  parts make contact and add graphite/lithium lube to make it smooth but everything is gettin` complicated very quick.

There are other approaches.

1. Pre-tension the switch, while inserting it, the lever is already exerting some down-force (20% ?) and will take very little force to press it/close circuit.
2. Add 1mm magnets to push/pull/compensate for the pushing power required.
3. Leverage based, moving the switch toward/away from the lever joint/axis you can play with how much force is needed. This will also affect travel distance.
4. SMD Hall sensor/switches, with one magnet you can control all 5 positions by measuring the magnetic field power
5. Optical/IR sensor to sense when the transmitter is being blocked/exposed to light and only use magnets to control the force required to close circuit
6. Re-do your model using SLA/Resin, not filament, you might get better results, smoother non-sticky/movement

#4 and#5 Will require extra electronics/board redesign and will not provide much click/clack feedback

[ironfox]

Carbon fiber infused got very good stiffness and wear resistance, I believe you can find samples (5~10 ft) on eBay/Amazon for fairly cheap, that should be enough to give you an idea of material you want to use without spending $50+ right away. Next down would be PETG

Ah, I see. Yes, there are some materials out there that might also do the trick. Like steelfill and such.
There even seems to be a special low-wear printing material called tribo-plastic filament.
Might try that at a later stage.

For you to get some decent results, no matter what material you gonna print with, re-design as needed to have at least 3mm of wall everywhere, Polycarbonate filament beats both carbon pla and PETG but is kind of bendy,  same with nylon, you cant use these with thin walls, so, up to you how you want to approach this.

Yeah, that might be a problem. The clusters are extremely low on extra space. 3mm won't be possible everywhere.
In fact, looking at the cluster as is, 3mm thick walls seem to be virtually nowhere.
Hrmpf...

There are other approaches.

1. Pre-tension the switch, while inserting it, the lever is already exerting some down-force (20% ?) and will take very little force to press it/close circuit.

Yeah, okay, that works for your arms, but I'm not seeing how you can do that for the center down key.

2. Add 1mm magnets to push/pull/compensate for the pushing power required.

I have a suspicion this won't work. At least not for the down key. For pushing, the magnets would have to be above something to push down for which you probably have no space.
And pulling has the inverse force curve of what you want. Low pull at first, strong pull further down. The key won't be coming back up.

5. Optical/IR sensor to sense when the transmitter is being blocked/exposed to light and only use magnets to control the force required to close circuit

Yeah, that's where we are right now.
All of the existing design is based on that.

6. Re-do your model using SLA/Resin, not filament, you might get better results, smoother non-sticky/movement

Uh, but I just got this printer...

#4 and#5 Will require extra electronics/board redesign and will not provide much click/clack feedback

Well, to be fair, the click feedback is there thanks to JesusFreke's modeling Genius. The idea to use the magnets for pulling and starting off from where they are in contact creates a sensation not unlike tactile switches.
True on the electronics design though. I just finished a cluster PCB. Still working out if the printing will allow holes for the down key or if I have to work around this limitation.
Thumb PCBs will come tomorrow.

I'm not sure if your design will get past the electronics part though. At some point you have to translate the signals into keys after all.
Anyhow, keep posting how that works out. Looks interesting.

[iso]

There is no point on trying to do anything with regular switches if the electronics to handle the IR`s are already done, unless there is too much lag or something like that.

There is how I`d do the center button.

The radius of the whole asembly is little over 9mm, my pinky is about 13mm wide at the tip, so there is plenty of space to resize some of the elements, like the guides to make it sturdier/easy to print.

Yellow = Guides slot
Orange = Magnet container
Blue = Switch container

Magnets can be positioned to add to the the force that it takes to press the button or make it easier, some super glue can help if the magnets are smaller than 2mm radius
To turn this model to work with IR/Optical is fairly easy, the top part of the button should not have a hole in the center but a shaft coming down with a horizontal hole, allowing the IR light to come trough when the button is pressed.
I havent made the slots for the IR sensors because I didnt had the sizes.

[JesusFreke]

It looks like your design uses sets of two opposing magnets to push on each other, to support the button in the unpressed state?

That's going to give a horrible, squishy button feel, with the force increasing as the button is pressed. To get the best feel, you need to have the magnets touching or close to touching in the unpressed state, and then the pull force of the separated magnets should be used to return the key to its resting state. This results in the highest force when the key is just starting to be pressed, with a quick falloff as it is pressed further, giving that nice, clicky feel.

[iso]

Travel distance for the switch I`m using is 0.20 mm, ... there is no... "toward the end", I`m gonna get rid off magnets altogether and print everything very tight just for the fun of it this weekend. 

[JesusFreke]

I guess I skimmed over your post too quickly. You say in the very first sentence that there's no point in using regular switches, so I assumed you were posting a design based on that premise.

[iso]

No, was my bad, I didnt specified that I was refering to the 4 way cluster, not including the center, from the previous posts I got the idea the IR electronics are already done and I dont want to be the disruptive one that is "trying to push" different switches idea becasue I was too lazy to ask or read previuous posts.

I like the IR idea because is less prone to failing and will require less maintenance, the mechanical switches, sooner or later will need replacing.
On the other hand, require extra electronics, cant just hook it up teensy and upload some firmware.

Monday I`ll have some molding silicone arriving and I`ll try resin casting.
I have the new center switch printing right now and I`m curious about what kind of precision/stiffness I can get using molding method because the way I see it,  that s pretty much the problem now, fitting all components on a rigid shell without having everything rattle or being too fragile.

[JesusFreke]

Sounds cool. I look forward to seeing what you come up with!

[iso]

Well... turns out I dont have to resin cast it.

I had to fine tune my 3d printer, change some belts, some bearings, change the effector, fans casing and my prints turned out very solid/stiff at 0.1
Thinnest wall on my model is approx 1.2 mm, its hard to say precisely because it all depends on printer settings, teperature and material, but arround there.

Round central button - Mouser ID: SKRGARD010
Side switches - https://www.amazon.com/FidgetKute-20pcs-Tactile-Button-Switch/dp/B07T2CXMD8

The way the main body was designed to allow the side switches to be positioned up/down to increase or decrease the amount  power that takes to close circuit, additionally, you can add a ring spring surrounding the levers at the base to make it even easier to push those.

The central plastic bit pushing down on the switch can easily be made out of brass to minimise short lived plastic on plastic friction, I was about to make metal bushing for lever too but I`m already working on 2 projects and I cant take on another one. I`ll still do little thing here and there as I will need to detach from main projects but I wont put as much time on this as I did lately. 

[iso]

Is anyone aware of/or working on firmware that will allow the use of PMW3360 (for trackball) and capable of 200+ keys on teensy 2++  ?

Thats pretty much what I need to finish my keyboard project and I can help with this project too, integrating the trackball on any of the  consoles.

[legopowa]

The DataHand patents were issued in 1998-1999, and patents in the US only last about 14-20 years. I think this whole project is fair game to commercialize, now or very soon.

[JesusFreke]

Sounds like a failed kickstarter waiting to happen!

[iso]

Almost there, got center button left to do.

First two attachments are GIFs

[RSanders]

As a long time user of a DataHand Pro II, I am so, very, very happy to see this project.   I have been able to keep my ~20-year-old specimen functional with judicious cleaning and occasional cyanoacrylate but have doubts to whether or not it will last much longer.  Careful viewing of the image will reveal the slightly off kilter "N" key and the outright missing top half of the left thumb up key, both victims of a careless nighttime cleaning crew, apparently determined to hasten my DataHand's demise. And yes, that is duct tape seen extending beyond the top of the left thumb upper right key. Has anyone here had an opportunity to evaluate the product line available from keymouse.com?   I was seriously considering ordering a pair for mounting on my chair similarly to how I have my DataHand currently configured until I stumbled onto this thread. I may hold off now that a viable re-created DataHand is a realistic possibility.

[iso]

It appears to me that Datahand is not for people with medium/small hands, I`m saying that because the thickness of the switches and distance between the clusters/switch assembly will force the user to uncomfortably spread fingers to touch the center/down switch or pinky/left and index/right.

For example my hand is 7 inch long, measuring from wrist joint to tip of the middle finger, according to the pictures and measurements I found about DataHand, I would not be able to use the device comfortably.

Part of the challenge is to make the 5 way switch no bigger than 1 inch width/height and keep it serviceable. People with extra small hands... sorry, having Michael Jordan`s hand size would solve lots of design problems

@RSanders or anyone that own a DataHand, do you happen to have large hands/long fingers ? And thank you all for posting pictures of the original device.

The keymouse... the trackball version is $500+..., the mouse version is not even worth talking about, you cant destroy your shoulder constantly moving half pound piece of plastic on the desk...

For that money you can get a cheap 3d printer for $150, micro switches max $20, one roll of filament whatever color you want $15, soldering iron, wires/solder/flux $40, teensy $20

And you can customize it, modify the 3d file and print parts as you please, make modifications to fit your need, hang in there, we`ll get this

[iso]

Also, can anyone with a device measure this for me ? Distance between the center buttons, from center to center. I`d like to confirm my calculations, thanks.