Trackball via Ball Transfer Units *3D Mock-up Made*

[Synnöve]

Get samples from Alwayse for the 11MI-05-17 by contacting this fellow:

Paul Hogg
Sales Manager
ALWAYSE Engineering Limited
(t) +44 (0)121 380 4700
(f) +44 (0)121 380 4701
(e) paul.hogg@alwayse.co.uk

Reference this thread/project and my name "Synnöve Vikström". He should be able to arrange something for you and will be pleased to see how far along this project has come. The Alwayse units are made with high quality components in the UK, and should last for a long time (the same cannot be said for the Chinese variants, not to mention they aren't as friendly to do business with).

I am very pleased with the continued development in this thread and hopeful that an excellent product can come from it. My transfer units are still spinning as well as when I originally bought them so that's a good sign as far as longevity is concerned.

Designers: Please do not forget those of us who have an extreme preference for a non-thumb trackball; whenever I try to use one I cannot achieve the movements nor same level of dexterity.

[dorkvader]

Designers: Please do not forget those of us who have an extreme preference for a non-thumb trackball; whenever I try to use one I cannot achieve the movements nor same level of dexterity.

Yes, I prefer non-thumb, larger ball, etc. I'd be willing to try a thumb ball, but ultimately I think the traditional kind is better for me.

[MrJohnK]

So, I'm pretty comfortable using a 1.5 inch ball for my thumb trackball.  It is currently a Delrin precision bearing ball I got from McMaster-Carr.  Would that size also work for the "non-thumb" versions or is a larger ball required?  If a larger ball is needed, where would you suggest one be sourced? 

Thanks.

-John

[jacobolus]

A billiard ball works pretty well for some people. I think there are about 3-4 different reasonable sizes for balls depending on where the ball is, the shape of the housing, and the intended grip.

[SL89]

Idk how i managed to miss this thread until now, but this looks like a very interesting project.

[hanya]

I have never used thumb trackball but I wonder 2" ball is suite for thumb used in:
http://www.cursorcontrols.com/products/e50-desktop-h-blbut-trackball/

[hanya]

I bought some cheap ball transfer units but each ball of the units were not rotating when I put billiards ball.
It seems the large ball slipping on the balls of the units.

[Synnöve]

The friction of the units internal system might be higher than the billiard ball on the units. Where did you purchase them from?

[hanya]

The friction of the units internal system might be higher than the billiard ball on the units. Where did you purchase them from?

This one: http://www2s.biglobe.ne.jp/~quartier/pettit-ball-castor.htm
Size and price ($6/4 units) are good but it does not match our purpose.
I would try others.

[MrJohnK]

I ordered the previously linked bearings from Holo-Pack and got some of all three to see which will work best

Also planning on starting the CAD design work this weekend.   Going to plan for the thumb design to use miniature pool balls at 1.5 inches and the forefinger design to use a standard pool ball at 2.25 inches.

Graphical user interface is complete and working.  It is written as a Java app.   It communicates with the Teensy over the virtual USB serial interface allowing for ten unique profiles that are stored in EEPROM on the Teensy.

-John

[MrJohnK]

Soliciting your opinions.  Since I'm a thumb ball user, I thought I'd tackle the more difficult part of the project (for me at least) by drafting a design that is not a thumb design.  In the design attached, the ball is a standard billiard pool ball (2.25") and the switches are Cherry MX.  The handrest is detachable from the mat (shown in black) so it can be replaced with a user supplied handrest or adjusted per the size of the hand.  The mat is neoprene covered in soft Velcro-loop.  The rubbery-ness of the neoprene grips the desk pretty well.  The ball assembly and handrest will have the hook side of the velcro to stick to the mat.

Electronics-wise, it will be driven by a Teensy 3.1 with either one or two ADNS-9800 sensors.  I have not decided if I want two sensors along the sides of the ball (at 90 degrees orientation of each other) or one under the bottom and optional one along the side of the ball.  Scrolling could be programmed on a pair of the MX switches or be handled by the dual sensors by twisting the ball.  I have used scroll on my thumb-trackball with two MX switches for a couple of years now and like it better that way.  Open to opinions though.

This is my first draft and will likely need many changes.  Let me know what you all think, especially those experienced with non-thumb designs.

Thanks.

-John

[dorkvader]

Well I think a lot of those switches would he hard to hit with your hand in a "normal" position on that. Also, I think at least one of the switches would interfere with a sensor placement.

As far as sensor placement, I think the best is likely to get one on the bottom and one on the side. You can use data from both sensors for one axis if you want one axis to be a little more precise, or just throw it away and use sensor #2 for one axis only.

Of course having both sensors on the diameter interferes with more MX switches, but allows you to place the ball closer to the desk. It may also be hard to get them both there unless you go with 4 ball transfer units. I think with 3 BT units one sensor underneath somewhere and the other on (or near) the diameter between 2 of the BT units is the best configuration physically and with acceptable mathematical properties.

Jacobolus knows a lot about the math of it, so I'll run it by him.

Thanks for all the hard work, John!

[MrJohnK]

The overall goal is to create a trackball for gaming, CAD and digital drawing.  As such, it has many programmable buttons and will have a very good set of bearings to roll freely for extremely high accuracy.

I would prefer to use three bearings as it is very difficult to get more than that to balance out properly.  Not impossible, but three is just so much easier.

I agree with you regarding sensors.  If I have two along the sides, each would be only using one axis under normal operations for X & Y movements.  The other axis would sense the twist.  I think height of the trackball will be challenging, thus sensors on the side is tempting for this reason.

The list of possible values that can be programmed on any key:

Any combination of key modifiers: Shift, Ctrl, Alt
  -Plus-
Any of these:
a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w, x, y, z, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12, SPACE, TAB, PgUp, PgDn, LCTRL, RCTRL, L.ALT, R.ALT, L.SFT, R.SFT, ;, ', [, ], ~, ., , /, \, =, +, -, *, M.1, M.2, M.3, S.UP, S.DN, Up, Down, Left, Right, 200, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000, 2400, 2800, 3200, 3600, 4000, 4400, 4800, 5200, 5600, 6000, 6400, 6800, 7200, 7600, 8200

Most of the list is based upon the Teensy's ability to send keystrokes.  M.1, M.2, M.3 are mouse buttons.  S.UP & S.DN are scroll buttons.

The numbers at the end of the list are sensor CPI values.  Any key (or multiple keys) can be set to toggle to an alternate CPI.  As an example, my CAD profile has a normal CPI of 2000 with toggles for 200, 800 and 1200; each on their own separate buttons.  It sure makes it handy to get the trackball into a very slow mode for pixel crawling on those detailed area with just a click of a button and back fast again by tapping that same button again.  In a gaming environment, such as a FPS using a sniper rifle, it would help with long distance targeting.

[wcass]

Putting the sensors on the sides would also help keep most gunk from collecting on the sensor.

I'm not crazy about twisting a ball for scroll. I would prefer a a thumb wheel (or horizontal axis cylinder) left of the ball. And I would put the mouse buttons on the other side of the keyboard to be used with the other hand. There is only two reasons for one handed computer input; neither applies to me (a two handed man with a great wife).

[FoxWolf1]

Ehh...I've been vaguely contemplating picking up a trackball for a little while now, and if I do, it will be a finger-type rather than a thumb-type, but I don't think I'd buy that...not with that button placement, 9800 sensors, twist scroll, or MX switches for mouse buttons.

[MrJohnK]

Ergonomics aside, I printed a prototype bearing socket today.  It houses a standard billiard ball and uses three load transfer bearings (the aforementioned bearings from Holo Pack).  The bearings were positioned around the socket at 120 degrees from each other and at a 40 degree angle from the top of the socket, which meets the equator of the billiard ball.  In simple terms, the bearings are about half way down around the ball between top and bottom and spread the load evenly among the three bearings.

I purchased three varieties of these bearings.  The are all identical in dimensions and only vary by materials used.  All three have offer low friction movement and the billiard ball can spin freely after release if accelerated quickly or "flipped".

The load transfer bearing varieties are: 
1. steel body with steel ball:
very noisy.  I doubt anyone would be happy with a trackball that noisy.  Motion is free and fluid, however, small movements are a bit chunky or not exactly accurate.

2. steel body with polymer ball:
the noise is less, but still significant.  Accuracy seems good.

3. polymer body with polymer ball:
much less noisy, free flowing and seems accurate and true.

One final tangent feature I added is a sensor mounting on the bottom of the ball and one more along the side.  There is enough room for both using my smaller rectangular "mini" ADNS-9800 boards.  Getting alignment is indeed tricky and in-fact, the ball measured a little on the small size and the side sensor was slightly too far out to function properly.  Since these bearings are in M8 screw bodies, it makes it very nice to dial-in the correct distance (2.4mm +- 0.2mm) between the billiard ball and the sensor, at least on the bottom sensor and, to a certain degree, on the side sensor.

If I plan to move forward, I'd probably choose either bearing 2 or 3.  I'm not sure how long the bearings will last though as the polymer options are probably more suppressible to wear over the long term.  Perhaps offering replacements in the case of failure would be a good option to hedge against premature failure of the bearings.  One concern I have is that most load transfer bearings are made for "ball up" designs and mounting them in an alternate orientation may not be ideal.

-John

[do_Og@n]

Has there been any discussion on using ceramic ball bearings over steel and polymer?

http://www.bocabearings.com/bearing-types/balls/ceramic-silicon-nitride-si3n4-series

I'm holding off on buying a CST or another Slimblade. Wanna see how this project plays out.

[dorkvader]

Has there been any discussion on using ceramic ball bearings over steel and polymer?

http://www.bocabearings.com/bearing-types/balls/ceramic-silicon-nitride-si3n4-series

I'm holding off on buying a CST or another Slimblade. Wanna see how this project plays out.

Do you mean using those ceramic bearings instead of the ball transfer units? They would have more friction for sure. Kensington uses "plastic nubs" instead of bearings and they can work but I think they are a bad solution.

I really want to get a ceramic trackball for the ball but they are expensive and hard to find in any size larger than 2"

[OverKill]

I bought ceramic bearings for my Trackball Optical for a replacement to the steel ones. They are much, much, much better.

1) Don't wear out at all compared to steel ball bearings wearing pretty damn fast
2) Even when it gets dirty it still works fine, I have only cleaned it once since I put the ceramic ball bearings in
3) They don't feel scratchy like the steel balls feel

As for ceramic balls in the ball transfer units I think it's probably a wash since there is supposed to be little/no friction with whatever is riding on them. The only thing I could see the ceramic balls being better in the transfer units would be when/if they get dirty there might be less problems. Until a unit is made that has the ball transfer units and is used continuously for at least a month or three then it would be really hard to see the difference between ceramic and steel. If the ceramic units are not that much more expensive I would try them out but if they are considerably more money I would probably stick with the steel until the ceramic can be proven to be better.

[lkong]

Any updates and plans for new years?

[Shmalzbrot]

I found this forum + thread while looking for repair hints and a possible replacement for the Trackball Explorer. I love the idea of a modular ball socket, and I think I can contribute quite a bit (from the time I considered building my own trackball from mouse parts a couple of years back), so here comes:

- Don't disregard ergonomics. It's the single reason why countless people keep repairing their decades old TBEs - there is nothing even remotely as ergonomic available today. I'll go into detail below.
- What would be the effort to send 5 different mouse buttons? Do you need driver support for that? Would it be possible to fool an existing driver?
- Have you calculated the force being applied on the BTUs? A 52mm Ball will weigh around 100grams (tested a Logitech Marble FX ball), my relaxed Fingers laying on the ball add another 150grams. At the angle the BTUs are positioned, I figure you're pretty close to the max rated load of the Polymer ball type, especially if you consider additional forces from moving the ball.
- Have you considered that the "natural" X/Y axis (the directions a user intuitively moves the ball to go straight left/right and up/down ) do not correspond to the physical X/Y axis of the sensor?
Have you considered that the natural X/Y axis might not be perpendicular?
I.e. are you planning to implement a translation from physical coordinates to actual mouse movement? Are you planning to add some kind of calibration to account for the different natural movement directions of different users?
- Is there a specific reason you made the socket cover so much of the ball? I think you can gain a lot of flexibility and ergonomics by making it smaller. Imagine a MarbleFX without the arch - so much freedom for all your fingers including the thumb!
- Wouldn't it be easier to make the sensor assemblies adjustable instead of the BTUs?
- PLEASE don't put a sensor underneath the ball - if you care about ergonomic design, you want the trackball to be as low as possible. Every mm of height adds bend+strain to your wrist and/or reduces the contact area of your arm, increasing pressure on your heel of hand. Also, if you need to translate physical movement to mouse cursor movement anyway, it does not matter where the sensor is placed.
Also, the way I use an ergonomical trackball, the actual center of movement (where physical movement and desired mouse movement would differ the least) is towards my hand - incidentally very close to where the sensor of the TBE sits


Now, my take on ergonomics and what I would expect of a finger-operated trackball. Of course I can only speak for myself here, that's why I separated this paragraph.
The reason why I consider the TBE so ergonomic is that it allows me to rest the heel of my hand completely motionless on the unit. The "hump" of the unit lies in the palm of my hand and supports it (albeit not as good as I would like it to). If I would have to move or even lift my thenar, I might as well use a mouse.
Note that in this position, the center of my movements is not at the top of the ball, but approximately 45° behind the unit.
I can move the ball in any direction indefinitely without stopping by using index and middle finger (although not as fluent as on a MarbleFX using index and middle finger + thumb). I don't like the stop-and-go caused by trackpads or thumb-controlled trackballs.
I can press and hold any button while still moving the ball; left ones with the thumb, right ones with pinkie and ring finger (which rest  on these buttons).
All buttons can be pressed without having to reposition my hand or arm.
Same goes for the scroll wheel.
The left/right tilt of the TBE is a good compromise between the natural relaxed angle of my hand (which would be almost upright) and a horizontal direction that distibutes weight pressure evenly over a large area of the hand.
For ball size, I would prefer the 52mm MarbleFX over the 45mm TBE. Gives a better feeling of gripping the trackball rather than resting your stretched hand on it. So yes, definitely keep the billard size ball.
I can use the 5 buttons without having to remap them for every application - e.g. games recognize them as 5 separate mouse buttons.


Sorry for that wall of text, but I hope it helps.
Also, if I can help you with any of the topics I mentioned (or if something was incomprehensible), let me know.

[jacobolus]

Hi Shmalzbrot,

Welcome to geekhack! Sorry about my own lack of progress working on trackballs; I’ve gotten stuck on several other unrelated projects (both keyboard related and otherwise) in the last few months.

I’ll tell you about some of my personal goals, but note those may be different from other people’s goals. My main goal with a trackball project will be to build something that is in between two keyboard halves, part of the same chassis. The goal is to have something very easy to reach by either hand without reaching far from standard typing position. This is a somewhat different goal from a standalone trackball where someone would have their hand on it continuously for hours at a time. I don’t personally really care about the palmrest part at all, because the way I type, my palms/wrists are “floating” above the keyboard surface, not resting on anything. With a trackball, it will be the same. I plan to simply use regular keyboard keys (probably on the other hand from the one manipulating the trackball) as mouse buttons, so button placement, etc. is also not really among my concerns.

Were I to design a standalone trackball, I’d recommend resting the hand tilted at a 40°+ angle if possible (to avoid excessive hand pronation). Any permanent palmrest should mostly support the side of the hand (the part below the pinky finger), rather than the center of the palm (this avoids pressure on the carpal tunnel). I’d personally try to use keyswitches with slightly longer travel than a typical mouse switch (on order 1.5–2mm, with maybe .5mm of pre-actuation travel and 1mm+ of post-actuation travel). I personally think mouse buttons are awful, and mainly get used because they’re what people are used to and expect, rather than any particular merit.

What would be the effort to send 5 different mouse buttons? Do you need driver support for that? Would it be possible to fool an existing driver?

That shouldn’t be much trouble, once the firmware has a reasonable USB HID implementation. Sending mouse buttons, keyboard combinations, game controller inputs, or whatever else should be pretty straightforward. Figuring out the best system for configuring the firmware to do what you want is an interesting non-trivial problem, though. The user interface design part is the difficult problem here though, not actually coding up the mouse button actions.

Have you calculated the force being applied on the BTUs? A 52mm Ball will weigh around 100grams (tested a Logitech Marble FX ball), my relaxed Fingers laying on the ball add another 150grams. At the angle the BTUs are positioned, I figure you're pretty close to the max rated load of the Polymer ball type, especially if you consider additional forces from moving the ball.

Which ball transfer units are you considering? The ones I got some samples of from Synnöve have a metal ball, and seem quite sturdy. I haven’t subjected any to long-term usage, so I can’t tell you for sure how well they’ll hold up, but I have high hopes.

Have you considered that the "natural" X/Y axis (the directions a user intuitively moves the ball to go straight left/right and up/down ) do not correspond to the physical X/Y axis of the sensor? Have you considered that the natural X/Y axis might not be perpendicular? I.e. are you planning to implement a translation from physical coordinates to actual mouse movement? Are you planning to add some kind of calibration to account for the different natural movement directions of different users? [...] Also, the way I use an ergonomical trackball, the actual center of movement (where physical movement and desired mouse movement would differ the least) is towards my hand - incidentally very close to where the sensor of the TBE sits

Absolutely. Personally, my goal is to use two sensors, and as an intermediate step compute the full 3-dimensional rotation of the ball at each timestep (e.g. as a quaternion). Once you have a precise/accurate measure of the 3-d rotation of the ball, it’s fairly easy to transform that to find the rotation relative to any particular set of axes, which means any arbitrary ball directions could be associated with X and Y (even with different sensitivity in each dimension if you want), when you want to use the ball motion to move a 2-d cursor.

Note that on a one-sensor trackball, the center of movement on the ball is the point directly opposite the sensor. The TBE has its sensor a bit behind the bottom center of the ball, which means that on the opposite side, at the top of the ball, the center of motion is actually quite substantially in front of the top of the ball. If you’ve learned to correctly move the ball despite yourself putting your hand on the nearer-than-center part of the ball, that just means you’ve learned to compensate for the specific way the TBE moves. The sensor is actually in a very suboptimal place given your description of your hand’s resting position.

A separate scroll wheel should be entirely unnecessary if the whole ball can be used for scrolling while some button is held down.

PLEASE don't put a sensor underneath the ball - if you care about ergonomic design, you want the trackball to be as low as possible. Every mm of height adds bend+strain to your wrist and/or reduces the contact area of your arm, increasing pressure on your heel of hand. Also, if you need to translate physical movement to mouse cursor movement anyway, it does not matter where the sensor is placed.

My personal advice would be to avoid flexing or extending your wrist at all if possible, keeping it in a neutral position, and also avoid putting pressure on the heel of your hand to the extent possible. Having your palm and wrist entirely in the air is the ideal, but resting it lightly on some surface is probably acceptable, as long as you’re not pressing down hard. Of critical importance is the height of your desk and chair, and the distance to the trackball.

[ECA]

Iv thought about this for awhile, and Iv used trackballs since the 80's, from a C64, Amiga, PC, even an old mac..

1. it does not need to be built to handicapped people...Does not need HUGE BUTTONS..
they take up to much space

Ok, fun part.
Thumb side can have a 4 position button pad, like the 4point control on a game controller..
Finger side can have 3-4 buttons..
trackball..do we really need a ball, or something that Acts like a ball.  Just a Domed movement sensor would be nice.

Now for the buttons..
It would be interesting if 1-2 buttons would Shift/Alt the other buttons..
like a button that changes the Trackball into a scroll.
it would be nice if all the buttons were programmable..but thats asking alot.

Being able to program them makes it so, we can config WHICH buttons are best for our use.
but that is asking alot also..as then the trackball might need memory.

Something interesting to think about.  And I have been trying to get others to think this also..replaceable cable..
IF POSSIBLE..  the OLD phone cables (handset/Base set)  cables were very nice and durable, and are Cheap at most stores.  They are 2 and 4 wire..  Even network cables are the same design, and generally 4-8 wires..  Even an adaptor from Net cable to USB should be a simple idea..

Min needs..
4 buttons
1 tracking mech
Cable..

more buttons would give this item More uses, then just being a mouse.  ranges from being a 3d controller to what ever a person needs.
Not all buttons need to be Quick access.  You could even have a 10key pad on the side for other functions.  Keep the basic ones with themselves, and the functions to the other side.

My problem is parts and the chip to control all this.  Also I dont have current knowledge to do this.

Any takers.

PS..
these hints/verifications kinda suck..."popular Keyboard connection"?  PS1 PS2 USB Serial...  Im to old for this.

[ECA]

no one wants to comments....

Ok..

[jacobolus]

Hi ECA, welcome to geekhack.

Note that your post is somewhat off topic for this particular thread. It’s a fine thing to discuss, but it’s probably not what most of the people involved in this particular discussion are specifically interested in.

You might want to make a new thread, and I highly recommend including a diagram or sketch.

[ursulache]

Hello,
has there been any progress made on this subject?

LE:
Flush with confidence from building a left hand ergodox spin-off for gaming I decided a trackball would be an interesting followup project. While doing research for it I came across this thread.

My CAD skills are not great but they sufficed me in modeling the case for the aforementioned keyboard and having access to a few 3d printers an possibly other "maker" tools would allow me to rapidly prototype something.

Recently, contact with ALWAYSE has been resumed in the hope they would agree to send some more samples, as those are the final pieces of HW needed for concluding the first and most importnat phase of the project: designing an ergonomic and practical housing for the trackball and putting the design to the test in real life.

Design goals for phase1:
1) thumb operated trackball that would allow for left/right hand use and with as much as possible of the ball exposed;
2) at least 4 rapidly accessible (directly under fingers) buttons. Cherry MX compatible switches (gateron) will be used as that is what I have around and allows for great key cap diversity without the hassle of designing and building the key+switch assembly from 0;
3) large ball. A Carom aramith billiards ball was chosen as it was the largest diameter (61,5 mm) ball that was also easily and cheaply available;
4) a mouse's guts will power the whole thing. The trackball housing should allow for a wide range of mouse pcbs to be used. In my case an old A4TECH X-718BK. I decided to use an exiting mouse as that greatly reduces the complexity of the project;
5) use ALWAYSE ball transfer units as ball interface or which ever units are readily available. As I don't have these units yet and have no idea if the company will agree to send a few samples my way, if anyone knows of cheaply but of good quality alternatives, do not hesitate to post the information.

Current state:
-CAD model of ball housing unit is in the final stages and going well (for 11MI-05-17 units) - POK;
-CAD model of hand rest unit is in the intial stage - NOK; Without testing the ball housing with the ball and working sensor I can't decide which exact hand position and hence which finger+hand movements would pose problems. In the current phase my prototype has a lot of friction on the ball and thus all finger movements are forced.
-Ball - OK;
-switches - OK;
-1u key caps - OK, larger keycaps (2u) which might be better - NOK;
-mouse  - OK;
-ball transfer units - NOK.

Any and all feedback is welcomed.

[ursulache]

If the gods of 3d printing will smile upon my project, all the parts should be ready by tomorrow.

Sadly, I still have not received an affirmative answer from the local distributor of Alwayse. I am looking into Si3N4 Bearing Balls, does anyone have any experience with them?

LE:
I just orderd 2packs of 5mm balls, one from si3n4 and the other delrin.

[livingspeedbump]

you have my attention 

I agree, there definitely are not enough trackballs on the market.

[ursulache]

I received some parts from the 3d printer.


This is the ball+sensor test frame.
It's purpose is twofold:
1) to allow x,z adjustments so I can find the position of the sensor relative to the ball in which it tracks the best;
2) to allow -20 deg and -10 deg mounting options for the ball holders. I already tried -30 deg on a different test frame and found that you can to easily push the  ball out of it's socket.






The metal legs are placeholders and are to short.

I played around with it yesterday evening but couldn't find a position in which the sensor would consistently tracks the ball movement.

[Oobly]

Okay, I know I'm really late to the party here, but I've had an idea relating to this.

Why not duplicate the functionality of a single ball transfer unit as the trackball itself. You'd need to make an accurate housing for it, though. The basic idea is part spherical housing with internal diameter of ball + bearing size, with a rolled off edge. You put a bunch of smaller bearings in there and then rest the big one on them, with a number resting on the rolled off edge. As the big one rotates, the small ones unload at the rolled off edge and are free to move. With an enclosed space they will migrate to the other side where they are rolled under the big ball again. The sensors will have to be poisitioned above the ball rolling area, so it means you'd need two sensors to track all rotations. A pair of optical mouse sensors should work just fine.

Disadvantages:
1. requires accurate "bowl"
2. larger mass of small bearings with more surface area contact means more friction and more transfered momentum
3. sensors need to be mounted out of the bottom "well", so needs more than one to track normal motions. For simplest tracking you want a single sensor mounted opposite the ball / finger contact point so movements there translate correctly.

Advantages:
1. Simple, easy to clean
2. more compact
3. operates in the most efficient orientation for ball transfer units, vertically. Angled ball transfer units have less efficient recirculation.
4. Can use ceramic balls or any other material that can bear the load as long as they're accurately made.

Hmmm... probably still best to go with commercial ball transfer units... Thought I'd share the idea anyway, in case anyone's interested.

<SKF also make ball transfer units, so almost any bearing shop should be able to at least order a few from a catalogue for you.>

[ursulache]

Great to have feedback!

If I understood correctly you would have to cover a lot of the target ball with your "bowl" which for the moment I hope I can avoid. I would like to have as much as the ball exposed as possible.

Cleaning could also be a problem as you would have to extract all the bearing balls (easy), clean them individually (less easy depending on their diameter and number), clean the "bowl" (easy) and not lose any bearing in the process (hard!! ).

When the bearing balls will arrive (delrin balls should arrive next week) I plan on trying the design from the pictures below.




I contacted ALWAYSE because their units had been tried by someone from the community with good feedback and they had already sent samples. As I don't live in the UK I have to deal with the local distributor whom has not yet responded to my request for samples. Anyway, If they are not willing to send samples, I am not willing to buy any at the moment. I would rather try to build my own first. If that works, the cost would be diminished and, as you noted, we gain in term of size  and also flexibility of design.

*SIDE NOTE: how long do I have to keep typing these verification strings? Is there a number of posts after which you are no longer required to do it?

[CPTBadAss]

How long do I have to keep typing these verification strings? Is there a number of posts after which you are no longer required to do it?

I think they stop after 10 posts.

[ursulache]

I just got the legs of the frame from the printer:



Tonight is "find the optimal tracking spot night" .

[ursulache]

Bad news:
I couldn't find a good position for the currently used mouse sensor (red led). Maybe it's tracking envelope is very tight and I wasn't able to find it , maybe it just doesn't track the aramith ball very well or maybe, due to the open stand, stray light is confusing the sensor (is that possible?).

Good news:
I used a tiny logitech wireless mouse (invisible optical). It seems to track like a champ! Its problem is that it's wireless and has just 3 buttons+scroll.

I have a laser mouse which has 9 programmable buttons and is corded that I could try but I have to find it first...

LE:

There was a project for using a ADNS 9800 sensor and a teensy 3.0 (I think) for building a mouse, does anybody know what happened to it? Did they manage to finish it?


*Yay! I no longer have to type verification strings, probably the limit at which it stops asking is 5 posts.

[jacobolus]

Why not duplicate the functionality of a single ball transfer unit as the trackball itself. You'd need to make an accurate housing for it, though. The basic idea is part spherical housing with internal diameter of ball + bearing size, with a rolled off edge. You put a bunch of smaller bearings in there and then rest the big one on them, with a number resting on the rolled off edge. As the big one rotates, the small ones unload at the rolled off edge and are free to move. With an enclosed space they will migrate to the other side where they are rolled under the big ball again. The sensors will have to be poisitioned above the ball rolling area, so it means you'd need two sensors to track all rotations. A pair of optical mouse sensors should work just fine.

I don’t think this has any particular advantage compared to a standard trackball ball perched on three small ball-transfer units. Overall it takes up a lot more space around the outside of the ball, and reduces a lot of the available flexibility w/r/t sensor placement, etc.

In general, I at least want as much of the ball surface available for my fingers as possible.

ursalache: try one of those “laser” mouse sensors. Some of them are pretty impressive about tracking on all kinds of surfaces.

* * *

I bought a few of the Avago sensor boards from https://www.tindie.com/products/jkicklighter/adns-9800-optical-laser-sensor/ and some Teensies, got some sample ball transfer units, and had the goal of building a 3d trackball using two sensors to measure the full 3d rotation. But I never got the housing design/construction off the ground, and I have gotten horribly side-tracked and not worked on input device projects at all in the past year or so.

[ursulache]

The mouse I can't find at home has one of those adns9800 laser sensors.

I got the delrin balls over the weekend so today I will print the poor man's ball transfer unit I linked in a previous post and will try them this evening.

LE1:

The laser mouse has been found!

LE2:

The poor man's ball transfer units didn't print so well. I'll make some changes to the design and try again.

[ursulache]

The avago sensor tracks the aramith ball very well and the board is rather small (the ruler is metric).

More

The printed transfer units turned out an unusable mess (badly printed) but I found out that a 5mm delrin ball fits nicely in the inbus head of a m6 screw.

More

It's not bad, the ball spins about ~1 sec if you flick it with average/low force. The printed transfer unit should improve that by minimizing the contact surface and PLA also has an inferior friction coefficient as compared to steel (0.16 vs 0.5?).

LE: The new donor mouse is an Anker 98ANDS2368-BA

[MaNiFeX]

SO cool.  Can't wait to see more on this project.   

[ursulache]

I redesigned and printed the ball holders:

More

They still need some work.

[CPTBadAss]

Really enjoying your progress ursulache. Those delrin balls in the little arms is such a clever design.

[sek1ne]

There aren't enough trackballs on the market.

Also, forefinger operated trackballs are evil. I feel sad that there are only two thumb-operated models on the market, and they're both mediocre.

Somebody said Logitech hates thumb-trackballs, and sics their lawyers on anyone who markets one. If true, the entire patent system and anything tangentially related needs to die a fiery, painful death.

I'll agree to the need for more thumb-trackballs. I love my m570 but I would really like more options.

[ursulache]

Thank You!

At this point the problem with them is that the printer I am using is not very precise in it's execution. Although the dimensions I specified should allow for the ball to roll freely inside it still doesn't. So I am reprinting them with greater and greater tolerances until it does. The first design was mechanically weak so it broke. It seems you sometimes forget the real world physical constraints when you're nose deep in CAD .

I hope I will have a first printing draft ready this weekend so I can print and iterate over the next week (weeks? ).

What I want in this first draft: a housing that supports the mouse board, the ball and 3-4 switches (LMB/RMB/MMB/DPI). I will leave the other 6 switches and scroll wheel for later.

The use of cherry switches might pose a slight problem to my intended design due to their total height. A cherry switch+housing+cap needs around 20mm of space whereas a mouse microswitch only needs 13 or so. I hope those ~7 mm will not pose a problem.

[ursulache]

I am of to a bad start with my plan:

I didn't manage to finalize the draft design. I kept getting hung up on unimportant details and that combined with limited time ensured that I didn't finish the draft yesterday. Also designing something from the ground up is hard, shocking I know!

Not really a bad thing, the 4th button (dpi switch) has been dropped from the draft requirements. It will remain accessible on the mouse motherboard which is good enough for a draft.

On the plus side:

I think it will be possible to adapt the trackball for left hand operation just by changing the "grip" position. This was/is not a design goal, just a happy coincidence.
I got a few 1.5U caps I will use for the 3 buttons (LMB/MMB/RMB) instead of the 1U keys.

[ursulache]

Nothing much beside some cad work happened this week as I've been very busy with other things.

I made two draft designs because the button grip on the one I like the most (visually) could pose some problems to the printer as it has many curved surfaces and holes. As the printer is not available for the moment I'll see next week how it will turn out.

The ceramic ball bearings arrived earlier this week. Compared to the delrin balls the friction is reduced at the cost of a slight increase in noise.

[GuilleAcoustic]

I bought an ADNS9800 sensor from Tindie in may in order to build a trackball from it. Still haven't found time to work on it as I had other projects to finish first.

The Idea grew on me when I worked on converting my BUSMOUSE CH Products trackball. I'm considering both roller with ball-bearing or ball transfer unit. Speaking of the later, here is a gorgeous 3 axis trackball using ball transfer units:

Not a valid vimeo URLNot a valid vimeo URL

[jacobolus]

That’s awesome!

[sinusoid]

Hi guys,

Great work so far. I've been following this concept for a while now, here and on Deskthority.

I kinda ditched the idea of getting Kensington Slimblade, and ordered some Rexroth BTUs instead, so I may soon join in on the fun 

Just a question though: I see that most people use pretty heavy balls for their trackballs, which is kinda counter-intuitive for me. Wouldn't it be better to have a lightweight ball, with little inertia?

[dorkvader]

Hi guys,

Great work so far. I've been following this concept for a while now, here and on Deskthority.

I kinda ditched the idea of getting Kensington Slimblade, and ordered some Rexroth BTUs instead, so I may soon join in on the fun 

Just a question though: I see that most people use pretty heavy balls for their trackballs, which is kinda counter-intuitive for me. Wouldn't it be better to have a lightweight ball, with little inertia?

these sensors are very precise, and some feel that a heavy ball is better for extremely fine control.

I still have my 2 sensors from mr kicklighter and have a great design in my head for a 3-axis MSTBE-type design. once I can get it drawn and then CAD'd Ill print out prototypes on the local 3D printer and see if I can get a decent one made.

Will also have to try the 5mm delrin ball / m6 bolt trick. Maybe get some PTFE balls and krytox too

[ursulache]

No luck with the printed transfer units. The small balls don't always spin in the housing thus generating different friction values which do not allow for a smooth motion. It might be related to the interior finish of the housing. Being 3d printed the walls are ridged and not smooth as they are with other manufacturing techniques. The hardness of PLA does probably play a role too.

From my tests, placing the small ball in a m6 inbus screw head is the best option.

Sadly, even with that solution, when you want to perform a very short movement from a stand still, there is a small (quite small, but noticeable) resistance to overcome. Unless one presses on the ball to stabilize one's hand, the resulting movement will be imprecise, a bit jerky. Do commercial trackballs have this problem?

Once the ball is in motion you encounter very low resistance and are able to have a smooth motion.

I will try some more redesigns, maybe 2-3 with different interior shapes for the transfer unit , before I give up, byte the bullet and buy those commercial transfer units. I hope they are god's gift to trackball users at 50 EUR for the lot...

LE:
I just tried a Logitech Cordless Optical TrackMan and there the ball encounters almost no resistance when you try to spin it from a stand still but it stops rapidly if you flick it. With my transfer units the ball encounters some resistance when you spin it from a stand still (results in a jerky start) but spins much longer then the logitech. I don't understand...

[sinusoid]

No luck with the printed transfer units. The small balls don't always spin in the housing thus generating different friction values which do not allow for a smooth motion. It might be related to the interior finish of the housing. Being 3d printed the walls are ridged and not smooth as they are with other manufacturing techniques. The hardness of PLA does probably play a role too.

Two suggestions if you don't intend to give up on this yet:
1.
Minimize contact area between the socket and the ball bearing.
Either push in some needles into the socket so their tips touch the bearing ball (instead of plastic), or get some small bearing balls (think hard drive bearings), heat them up a little, and push into the pla, so they make contact points.
You can also try lubricating the socket (sic).

2.
Use tribological polymers.
There are special low-friction plastics that are recently seeing a wider use in place of metal parts. They have low friction even on large contact surfaces, and are manufactured in different versions, for contact with steel, chrome, nickel, or some slippery plastic on plastic action (sic, again...).
I know Igus is making such parts, and they're keen on sample giveaways (just look them up, you'll probably find a local dealer).
You can buy chunks of these things and machine them yourself too if you need custom stuff.

The pro of this approach is that plastic bearings naturally dampen vibrations, and have a slightly higher friction coefficient than steel parts, so they feel a bit smoother and give a little bit of feedback.
They are also maintenance free.

I just tried a Logitech Cordless Optical TrackMan and there the ball encounters almost no resistance when you try to spin it from a stand still but it stops rapidly if you flick it. With my transfer units the ball encounters some resistance when you spin it from a stand still (results in a jerky start) but spins much longer then the logitech. I don't understand...

Did you try comparing the weight and diameter of the balls? Might be due to inertia.