[WIP] The REAL Nav Ball Project Thread

[cyberKerb]

4 hours ago, NeoMorph said:

Wile, I was wondering if you had figured out the way to paint your nav ball itself.  I've got the theory but again the problem I came up against was something essential.... i.e. accuracy.

<snip>

But the paint job for the ball itself is going to have to be done here and sent out because of what I need to make to actually make the markings. Also I need to find a source of spray masking... hmmm

Just as a response for how I've so far planned to print of the ball, I was thinking a subassembly like the eggbot, but all manual. With a few gear reduction ratios for the ball to rotate very slowly, I'll be able to get a pretty accurate positioning on when the paint pen would lift or go down. The pen arm would be on the  axis as the rotation of the ball, but I'd have an height adjustable set of 8 "bumps" rotation axis where the ball is connected and sitting under the pen arm. This would set the negative space when I want the pen lifted as the ball is rotated around it's axis. 

(It's at this point I wish I could draw a model of the idea to better communicate what I mean *sigh*) Best I can describe is to take a big dinner plate and stick a small saucer on that plate at the edge so large place circle has a saucer sized "bump" protruding over the edge.

A little bit of trig maths will get the the angles needed for the paint stick to be at a correct angle to the balls surface for the higher latitude lines. The length of each line would just be a function of which latitude I'm at. Thankfully in KSP all the longitude line are mostly solid with 8 gaps for the numbers.

And then there's the hard bit. At the moment, I'd have an accurate ball with all the lines on it. (At least for home use - I'll happily recreate it accurately if I have to personally take it to space myself  ) Best I could come up with to do it cheaply was to use dry transfers. I've used them before, but the shear number of them is the issue I'm facing with trying to place them all  - 25 for latitudes and most of those 3 digits and then 4x32 for the logtitude number and most of those are double digits. So rough guess would be 381 individual numbers.

Still might go with your route with CNC - as I'm getting stepper motors anyway, but I might wait the standard 3 years before I pull the trigger on that

Heh! No prob on the parts - you gets your working first! I'm sure you'll have plenty of sim-pitter asking for help in the future if you set precedent here. At any rate, I'll guess the CNC first job might be making the frame/arm/clamps for a spherebot so you can get a good print on your spray-paint primed spheres.

Edited August 10, 2017 by wile1411

[NeoMorph]

Yes, the way to do it is a modifield Eggbot made specifically for the navball. For example, the Eggbot is HORRIBLY unbalanced... but then again it's only supposed to be to do designs on eggs. My version is actually balanced and instead of the servo motor I've gone full three axis with a GRBL loaded UNO with a Gshield v5 running the servos.

Another thing to deal with is getting as MUCH of the sphere as possible so I will be mounting it on a strong axle and the head will reach to near the red circles.... if not actually MAKE the red circles..

That's why I need the CNC manchine... to get it as accurate as possible so the map gets posted on top as accurate as possible. I was going to engrave it but decided to use black sharpie on the white/light grey side and use a white pen on the black side. That means I will have to do two runs with two different maps which should be interesting. As long as I keep the work mounted securely and the CNC doesnt get reset.... well, it should match up.

I've even named my modified Eggbot... to... The KerbaGRBLizer. - It even sounds like something Jeb would say.... "ENTER THE KERBAGRBLIZER!"

I've got SOOOOOOOO much 3D work to do... turns out all those hours of technical drawing aren't wasted after all. I should be able to make some awesome parts..

 

*Note: GRBL is a GCode interpreter... and it's pronounced "GER-BULL"

[cyberKerb]

Heh - figured it'd be something like that. In my research I saw the "egg-bot", but also ran across it's clone called the "sphere-bot" doing a good job on ping pong balls. It did run through my head that given a secure enough mount you'd get most of the surface area, but I was looking for cheap and accurate "enough". 

I don't think you'll have too many issues getting the sharpy all the way to the shaft holes at either end. The radius of those holes should be plenty to get a good grip and not have to just rely on the flat foam friction hold that the egg-bot uses.

Good luck with the design and hope to so see your progress pictures!

Separately, I might actually see if I can try the sphere-bot thing too - I've got a spare UNO that I can pull from (another) stalled project), but I'll need to grab a couple of small NEMA 8 steppers (about $15ea), a pair of drivers and a 1:5or6 gear shaft reduction to get somewhere in to 1200 step range for accuracy. I've got a stack of the 9gm servos from my remote control planes laying about and that  most of what I'll need I think. All that's left is a solid mount and maybe getting an optical encoder just so I know I can get a project working with it (wouldn't strictly need it though)

[Dr.Vulpinus]

On 8/10/2017 at 0:51 AM, NeoMorph said:

Regarding the slip ring.... it's a 12 way for a huge reason. I didn't want to be resetting the ball every time I load d up KSP so I figured out that something as simple as a magnet and a Hall effect sensor does the trick. Glue the magnet to the inside of the ball and the sensor mounted on the plate, that's the pitch done... then a magnet on the edge of the plate with the sensor on a mount on the cage did the yaw... and finally a disk on the roll axle with a magnet attatched and a sensor on the case sorted out the roll. Then all you do is something like the homing routine of a cnc machine will reset the ball when you need to.

That is a really great idea.  I have been working on a Navball design for my open source Gemini sim-pit project (KSP Forum link), and I had planned on just simply aligning it manually once, then always making sure that it was turned off in the "home" position and then never touch the ball again.  Not really an ideal solution I realize.  Magnets and hall effect sensors are far more elegant.

I have opted to completely 3D print mine because oh lawd trying to engrave a sphere with a CNC gives me stomach cramps just thinking about it, plus since 3D printing is one of my areas of expertise it vastly simplifies everything for me (the markings can be printed right into the surface of the ball) and majorly reduces cost.

One other issue I ran into was with the needles.  Trying to use simple rotary needles ended up being really complicated and finicky to get right to avoid any possible situation where the needles collide.  I was wondering how you planned to deal with that?  I spent more time than I'd like to admit testing the alignment and shape of the needles in CAD to no avail.

In the end I decided that I wanted full length needles to make visualizing the nodes much easier, and ended up just making the needles slide in slots on the side of the frame so that I could avoid the problem entirely (see the pictures below).  Also because I realized that the rotary needles couldn't be driven directly from the servos anyways due to the small range of motion, so instead I used the servos arms to push the needles one way, and springs to push it back, thereby eliminating any hysteresis error.

Here are a couple pictures of my design so far:

Basically all of the mechanics are complete, I just need to finish the frame design and panel mounting.

Also, on somewhat of a tangent, @wile1411 if you want to learn CAD I would highly recommend learning with Autodesk Fusion360.  As someone that has tried just about every CAD package out there, it is probably the easiest and most versatile CAD program available today, and plus it is available for free for students and makers. (and just to be clear, no I don't work for Autodesk, just my honest opinion here).  In my experience, trying to learn Sketchup is a great way to scare people away from CAD by making it excessively awkward and obtuse.  If you really want to use a totally free CAD program, then I would recommend FreeCAD.

[NeoMorph]

Doc Vulpinus I'm actually learning Fusion 360 at the moment as I realised it pretty much free... and for Autodesk that's awesome. 

Regarding the needles I'm using small gearboxes out of a couple of gears to reduce the 180 degree of movement down to a small arc. The hardest part wasn't so much alignment as the needles bouncing all over the place being so long... but then I smacked myself upside my head when I realised NASA had already sorted it out.

You take a small metal pipe for the majority of the length and the feed the display wire into the end and solder it in place to stop it moving. Worked perfectly when you realise the outer needles are actually also raised slightly so they don't interfere with inner needles. They don't bob about too much and don't interfere with each other. I think I'm using 1mm brass pipe with a 0.5mm stiffened steel wire. Like the picture below.

 

A lot of the problems are just a cumulation of smaller ones... fix them and it solves others too. But equally if you mess up one part you can mess up others too. 

Now for a bit of light hearted stuf after those tough little problems. I buy some really weird stuff on eBay at times and today I received some small Stepper motors heh. I guess I should have looked more closely at the scale. 

Edited August 15, 2017 by NeoMorph

[NeoMorph]

LOL... another of my random purchases arrived today and these really powerful steppers finally arrived. 

 

They are actually real Stepper motors. I can't believe how small they are. I knew they were going to be small as a pack of 10 only cost 99p (UK pence). But they are less than the top of my index finger. 

Edited August 17, 2017 by NeoMorph

[cyberKerb]

Are they for the needle indicators?

[NeoMorph]

On 17/08/2017 at 1:35 PM, wile1411 said:

Are they for the needle indicators?

I was thinking of that function but I'm not sure they would have enough oomph. I will have to try when I have 5 minutes spare... unfortunately I have soooooooooo much on my plate at the moment that I doubt I will have time for ages. I am learning Fusion 360, learning CAD/CAM and G-Code, getting ready to build my first CNC machine, build a noise-reducing enclosure and then putting it all together... and I can only do a little bit each day due to my damned messed up joints. I will try to document it though. Cool part is that every day for the last few weeks I have received parts in the post for various things I have bought online. I gave up putting it off due to Brexit... I didn't want to get hit by high import taxes.

[NeoMorph]

Disaster Time... LOL...

In a typical Kerbal event, the laptop I was thinking of using to drive my CNC machine decided to have a meltdown... Quite literally. I had used hot glue to keep some parts in place due to the plastic breaking into lots of little bits and the laptop got sooooooo darned hot... THE GLUE ALL MELTED.

So I decided that I needed to trash the thing... It is 12 years old after all... but I do like to keep old PCs... I still had an old Windows 98 PC running on my network as just a storage server until I finally retired it about a month or two ago. All the stuff on it is over 15 years old or there about. I hoarded data as well as other physical things.

I might keep the case and refurb the innards. That way if I ever get robbed they thieves will go "Ewwwww... what a load of crap" and ignore it heh.

So anyway, I couldn't afford much and the CNC machine will be arriving next week so I wanted something to drive it (without having to relocated my main gaming PC from my bedroom to the living room)... so I spend a whopping £95 on a refubished Core 2 Duo Desktop. All it's got to do is run the control routines for GRBL after all. If it's powerful enough I might do some CAM on it too. But it doesn't have to be really uber powerful.

So the next task will be learning CAD/CAM/CNC parts production. I've been spending my time learning Fusion 360 and banging my head on the desk heh. I could use Sketchup and SketchUCam but Fusion is sooooo much more powerful.

[cyberKerb]

bugger eh? Sounds like you're not far from the CNC machine. I can image you being excited to the point of 2nd Xmas when that arrives

[NeoMorph]

On 24/08/2017 at 0:22 PM, wile1411 said:

bugger eh? Sounds like you're not far from the CNC machine. I can image you being excited to the point of 2nd Xmas when that arrives

Nahhh... totally chill here... I'm not that bothered really.... WHO AM I KIDDING...  I'M REALLY PUMPED, ESPECIALLY WHEN THIS JUST ARRIVED!!!!

I'm going to be taking it slow as I have so real custom stuff for it on the way (from the other side of the planet in fact). Saving up for this thing has been something I knew I would buy IF I figured out how to 100% make the navball... but not only that, I can now make my own custom flight sim panels (or I will when I get it put together.

I'm going to be running it in a dust proof, sound reducing cabinet so I've actually got to build that first. I just hope I can get it all together before the winter pain levels skyrocket again.

But first thing though is itemising EVERYTHING in this kit and there is a ton of stuff in there... it's like being a kid with his first Meccano set (and yeah, I had one of those when I was a kid).

It's Xmas... in September... and Destiny 2 is out in 3 days as well lol. WHAT DO I DO?

 

 

[IAmNotCreative_21]

Awesome project.

Edited September 3, 2017 by sumghai
Removed unnecessary string of numbers

[NeoMorph]

Well I Kerballed it yet again... So the router bits I bought don't fit the Dewalt collets. The 6mm shaft bits I have just float in the 6.35mm collet. The 4mm bits don't fit in the 3.125mm collet... and when I looked up replacement collets I realised that I was double stuffed as they are rarer than dodo doodoo. 

I've put the 4mm bits to one side but have bought a 6mm collet to fit the bits I have... and the damn thing came to £26 (that's nearly $34) for a collet and the router itself only cost £159.  That's bonkers. 

I think eventually I will switch to a spindle with a standard collet and chalk this up to learning steps.

I do know that I really should have done more research on the collets ahead of buying the bits but I have a 1/2" and 1/4" router that each came with an entire collet set. I just never thought it would be a problem. 

[NeoMorph]

Holy crap... I have to admit that I took the pee out of people who took sooooooo long to put their CNC machine together... I mean, it's just grown up Meccano after all. It's simple.

NOPE!

I've had one nut go spanging off into the stratosphere already and I've only just put one side of the gantry together. What takes the time is tuning it so there is no play in any direction at all... it's either too tight or too loose... took ages to get it in goldilocks.

Oh and I've realised that I should have resorted the packets of parts so that I only have the bits I need. Before I go any further I think that's what I will do. You live and learn.

Also my friend asked me how much I've spent on this project since 2013 and after doing some rough ballpark figures I was shocked to realise it comes to nearly £3,500. Granted a lot of that has been getting the tools to make the parts and the CNC came to £1,500 on its own without any bits or even the spoilboard. Add onto that are some mods and a unit that will dampen out the sound of it in my one bedroom flat and that's another £500-£600 (the foam alone came to £70).

But I did save for 4 years so I had enough to not go into debt. Oh and the tools are going to be used on a bunch of other projects too. I will be able to make my own custom panels that I had to Order from New Zealand last time.

Anyway, I am just doing the CNC subassemblies for now as I need to finish the housing unit first. It's just that assembly is quiet in the evenings (well other than me cursing when I drop a part).

Peace, out!

Edited September 5, 2017 by NeoMorph

[NeoMorph]

Well it's slow coming (I'll show you why when I finish it... but the bits are coming together... this is the 80% completed gantry. It's got to get the brains attached and wired up.. then the base completed and attached... then I will be able to "make some chips" as they say in the industry... probably why machinists are chubby    Hmmmm  

(Please excuse the crappy hallway... I'm doing umpteen things at once... including decorating and getting rid of the horrible 90's wallpaper).

[h0yer]

Soooo, the NavBall is now turning into a fully fledged industrial scale 3 axis CNC milling machine?

 

That's the spirit

 

Don't forget to make it versatile, like adding a hotend/extruder from ordinary 3d printers/fused depositioning molding machines, and maybe even a UV projector for resin based 3d printing, and if you are seriously serious about it, add a CO2 laser for nice laser cut panels and parts and even the ability to mold iron dust, probably as interchangable modules or all mounted on one tool-magazine-spindle-cartridge thingy^^

[TheKurgan]

I did electronics engineering in school... many years ago... I love to build stuff (mainly long distance drones, both air and ground)

I'm not normally immature in my posts, but...

THIS THREAD IS ABSOLUTELY AMAZEBALLS!!

[NeoMorph]

Dammit... was sure I posted...

Anyway I AM making progress but I had to go ahead and make my own version of this build so the wiring was tricky... but here is where it is so far. Hell, I even made the units it sits on (if you call a couple of tables “units”).

[cyberKerb]

Yikes that desk! Looks like you have a few things that are impeding your X & Y directions.  *jk*

great progress so far. I can only imagine how chuffed you'll be once it starts to move on its own.

 

[NeoMorph]

On 27/10/2017 at 4:23 AM, wile1411 said:

Yikes that desk! Looks like you have a few things that are impeding your X & Y directions.  *jk*

great progress so far. I can only imagine how chuffed you'll be once it starts to move on its own.

 

Heh... I just have tooooooooo much crap and the sign of a clean desk is a crazy mind...but the sign of a chaotic desk(like mine) is 10,000 times worse.

Well after tracking down some problems with my build (loose grub screws, intermittent contacts in connectors etc, etc) and then designing some custom interface bodges (isolating systems so if one goes boom then the others don’t) I’m now just doing the finicky bits to get it working... like wiring up the multi core wires that DOES MY HEAD IN. I think I wired and unwired this one connector FOUR FLIPPING TIMES this morning because of some reason or other.

There are also some mods I’m adding as well that should help increase the rigidity, stop gear slippage and it’s all to make it as accurate as possible. Problem is, I should have bought this thing last Spring but I was faffing about... and now we are heading for winter... FAST.

The problem is that I’m having to up my goddamn pain meds again plus my nerve spasms are rising. Normal winter stuff for my disability but disastrous when I need to solder tiny, tiny wires to connectors and make sure there are no shorts. It’s the curse of making it modular but the benefit is that if I need to fix or remove something in future I just unscrew the box and can take it to my desk rather than trying to do soldering upside down in a cabinet.

The one cool thing is I got my XPro v3 GRBL board to talk to the SuperPID2 and can start and stop it, set the speed from 5000rpm to 30000rpm. I’ve also got the system set so it can start and stop the vacuum dust extractor too. I guess I’m just a geek who refuses to go small lol. For a home machine it’s going to kick ass.

Once I’ve finished wiring all the connectors into their final place it will be testing, aligning, testing, tramming and more testing and the first proper job will be using the machine to make a Control Panel that will give me shortcuts to some of the more current functions. Once that is done... I can start working on the first CNC cut parts of my NavBall. Ultimately it will me progressing and making the ball engraver CNC from scratch but I think that is next years problem.

[Ericwi]

Did you still need Krpc code?

[cyberKerb]

Hey @NeoMorph

Just wanted your thoughts on how to connect the Non-fixed hemisphere that lets you repeatedly get back into the guts of the ball? So far, my plan involves the dual shaft NEMA8 stepper (used for pitch) with 1 hemisphere permanently glued to a wooden disc that the shaft will connected to via a flange screwed onto it.

However, I'm wondering if there another way than securing the other side.with the 4 x rod Neodymium magnets bought for that purpose to act as a screwless flange. I'll need to ensure the connection doesn't move when under rapid rotation, but I'm just worried it'll separate when the y axis is spinning.

Also - how's your learning curve going with the CNC?  

Edited February 2, 2018 by wile1411

[NeoMorph]

On 02/02/2018 at 10:02 AM, wile1411 said:

Hey @NeoMorph

Just wanted your thoughts on how to connect the Non-fixed hemisphere that lets you repeatedly get back into the guts of the ball? So far, my plan involves the dual shaft NEMA8 stepper (used for pitch) with 1 hemisphere permanently glued to a wooden disc that the shaft will connected to via a flange screwed onto it.

However, I'm wondering if there another way than securing the other side.with the 4 x rod Neodymium magnets bought for that purpose to act as a screwless flange. I'll need to ensure the connection doesn't move when under rapid rotation, but I'm just worried it'll separate when the y axis is spinning.

Also - how's your learning curve going with the CNC?  

As is normal, during winter I have a TON more pain... so progress slows... but I wanted you to see the current state of my baby..

I've had the stepper motors for inside the ball since the beginning of last year (and yeah, I have a dual axis one). When I have finalised and tested the mechanics of the system I will be going onto the ball itself... and yeah, the extrusion at the right hand side is actually to make the dedicated ball engraver.

Regarding the ball outer sphere I have designed a mount that attaches to the shaft and then used keyways and three screws to hold them in place. You don't want magnets in there because it will mess up the hall effect sensors as the ball spins and the zeroing system would malfunction... and yeah, I originally designed it with magnets holding the ball in place too... good idea, but not with the overall plan with the sphere rotating in all directions. The hall effect sensor would pick it up and then think it was at the zero point and lock the ball in place.

As you can see in the picture, the majority of the CNC is done. All the electronics work great. The computer can even control the speed of the router and turn the vacuum on and off. Talking about the vac, you can see a cyclone extractor so that no dust reaches the filters and blocks them. The dust and bits get dropped through the top board and go into a sealed box inside the lower unit. When the box is half full I will empty it by just disconnecting the lower flexy pipe from the box and pulling the whole box out. Simples. The routing of the upper pipe was screwed up because I didn't leave myself enough of an overhead so I had to use knicker elastic to create a way to force it to retract and not get tangled up. Works fine so far with the test cuts I have made but will have to keep an eye out. I did have to add those extra pieces of extrusion under the spoil board because it was warping with only two supports underneath. Now I have 5 and it's rock solid when I do up the bolts. I can also remove it to fit a vice directl to the support bars. Oh and as everything is bolted down securely, it's as solid as a rock. The CNC machine doesnt move, the whole darn structure moves (which is why I am going to be adding extra feet and sand bags.

To finish the unit I need to make the front frame carcass and put in the doors. Don't know yet if I am going to use sliding or not... but I am leaning towards sliding on the top unit main door and normal hinges for the rest. The lower left opening will have drawers that I have got to make (got the parts at least) that will hold all my router bits.

Ultimately this will do the job that I want nicely... and let me make stuff even when I am feeling crappy. I do want to ditch the SuperPID eventually though as I think a Spindle would be better and allow me to do tapping. I MIGHT be able to do it with the router but being able to reverse and run at much lower speeds is a big bonus.

If I manage to actually make a real working Nav-Ball I will be really happy because this build was a poodle because I have had to do a whole load of custom stuff due to making sure it is safe (ie no static and you can see the earth lead that connects to the copper wire in the extractor lol... before I did that the dust was lining the pipe due to static).

Bit of an info dump but I hate winter (or precisely my body hates winter). I've mainly been doing 3D training in Fusion 360 when I can and when I can think clearly I do some coding. Just finished part of the remote control board code today that will let me run it like a larger machine (as in Manual Pulse Generator to jog the head precisely). Ironically it was done in a game utility (autohotkey rocks for how simple it is). Just add a Leo Bodnar Universal USB Joystick Interface Controller and you are rocking.

Will probably finish this when there is some warmer weather and my shoulder doesn't keep falling out of its damned socket. Sides and roof pieces just need to be screwed on (after adding the foam soundproofing) and then the doors... Oh and the temperature control system (basically fans that kick in when the temp rises too much and feed them through serpentine tunnels lined with foam to reduce noise getting out) and then I will start cutting the Nav Ball framework. There are loads of differnt layers to get the data across on the front of the unit that you don't realise just how many bits of data are actually being linked in.

I really should type up the system specification and put it on here and maybe someone will be able to beat me. That would be okay as I am going to be giving the plans away for free anyway... I would get a kick out of seeing someone build a Nav Ball that fully works. Building the hardware is only one stage of the problem. Then comes the coding to receive and interpret the data from the systems and then align the ball... the needles will be the feedback from the navigation system and the rate of change (as in if the craft is spinning) means there are 9 channels of basic data to work with. Then there are sensitivity settings so you can fine tune your course, systems needed to zero and then align the ball to a fixed angle. And finally adding in random failures... including that red pop out flag that says the unit has lost power.

Fun times ahead for this year... because I also want to make a cockpit for Elite Dangerous... can you tell I love space games heh.

 

[cyberKerb]

 

13 hours ago, NeoMorph said:

Regarding the ball outer sphere I have designed a mount that attaches to the shaft and then used keyways and three screws to hold them in place. You don't want magnets in there because it will mess up the hall effect sensors as the ball spins and the zeroing system would malfunction... and yeah, I originally designed it with magnets holding the ball in place too... good idea, but not with the overall plan with the sphere rotating in all directions. The hall effect sensor would pick it up and then think it was at the zero point and lock the ball in place.

Wow - that's some awesome progress! Are you still having closed loops for all three steppers?

Here hoping you got some warmer weather soon, although not too soon  I'm enjoying summer here at the moment.

 

With your dome mounts, does that mean you'll have visible screws that go from outside and screw into the mount inside the ball?

My other options were to either have a nut at the top of the dome and screw it on to the shaft. That idea is me least favorite due alignment issues trying to get screws on either side to line up, also a risk of the screws loosing, but it would be the easiest solution. The other one that ran through my head the other day when changing the bulb in an oyster ceiling light (but haven't thought it through yet) was using a pair of ball pins to fix the alignment issue, but still need to think of not too happy with the possibility of it flying off as I'm not sure of the holding force of ball pins.

Looking for inspiration, currently trying to work out how the hemispheres are connected on this:

 

But failing that, the last idea I had was to notch the end of the center shaft and slide the hemisphere into place, secured with an e-clip. still need to work out something to stop the dome rotating

Edited February 4, 2018 by wile1411

[NeoMorph]

8 hours ago, wile1411 said:

 

Wow - that's some awesome progress! 

With your dome mounts, does that mean you'll have visible screws that go from outside and screw into the mount inside the ball?

My other options were to either have a nut at the top of the dome and screw it on to the shaft. That idea is me least favorite due alignment issues trying to get screws on either side to line up, also a risk of the screws loosing, but it would be the easiest solution. The other one that ran through my head the other day when changing the bulb in an oyster ceiling light (but haven't thought it through yet) was using a pair of ball pins to fix the alignment issue, but still need to think of not too happy with the possibility of it flying off as I'm not sure of the holding force of ball pins.

 

Looking for inspiration, Current trying to work out how the hemispheres are connected on this:

 

The reason the screws wouldn't be visible is that there is a cover that goes over them on the real 8-Ball. In fact you have probably seen it multiple times and never realised.

Can you spot the end cap cover in this image (hint: red marks the spot heh)?

 

You can see how the shuttle 8-ball attatched in this image... now recess it slightly and think of one hemisphere as a wheel bent over the axle and fixed like an normal wheel would. The red cap has a pin in the centre on the inside which fits into the end of the axle with a keyway to hold it aligned. Little bit of locktite blue would hold it in place nicely and when you need to access the screws (also held in place with loctite blue) would still allow access to the internals.

This was one of the big reasons I went ahead and got the CNC. Making parts like these manually would need more expensive tools and more skill than I have to build to fit to a tolerance that would hold together and not rub against the inside of the frame. With a CNC built sturdy enough I can make slightly domed axle mounts with perfectly centred mounting points that would be a right pain to do on a lathe. Ideally it would be done on a small CNC lathe (and yeah, I would love to get one of those as well).

There has been a lot of thought gone into this build over the years and to think that they built the original by hand by expert machinists to a tolerance so small is mind blowing. Mind you, when I first started doing the CNC research I needed I was regularly bouncing my jaw off the floor at how fast and accurate the machines are. I've had to go without painkillers just so my brain could learn loads of different things that will all come together to make this final machine. I think I might even get a bit depressed when I complete it finally as this has been so many years in the making, from the early research to just figuring out how to make one, to the design of the multiple failed prototypes I made. But with each one I learned something new which led to the next one. It's hard to realise the above picture is showing NINE axes of movement.