moving camera over Z-axis

thinking about this plan - it might require more than 2 lasers in total such setup. 2 that are low in the case and 2 that are quite a tad higher. one wouldn't want to move the lasers i guess as those are accurately pointed. And the linear movement is not so accurate. Especially the leadscrew or ballscrews are notoriously seriously bad quality. Accurate movement with ball screws is too expensive obviously for a cheap scanner project. So the question is not so much how much lasers and linear movement is needed - the question is whether the software is so clever to fit different point clouds together from different Z-heights taken - and we have a rough guess what the Z-height is - yet ball screw (C7) will be off easily 0.08 mm from the ideal path in absolute error. Please note German C7 quality ball screws use a different errortable so one would guess there absolute path error is 0.04mm roughly. The question and discussion here is of course the software like Meshlab. Does it try to testfit in clever manner point clouds and determine what the position of the camera was? Kind Regards, Vincent Diepeveen

I have been working on a design that uses four lasers, trapizodal screw and Nema 23's to drive the motion. The Turn table has been up graded to be supported by a pin bearing to aid in motion and allow better bearing of heavier objects. The link is here:

https://www.thingiverse.com/thing:4023019

That was the last update I posted as have been working on multiple projects and trying to get business and family items dealt with first. Would be glad for help in the design and also in upgrading the current Fabscan board to being able to handle 4 lasers and only having to motor drives or the ability to off load the drives to a higher voltage driver for smoothness and control.

Yes much appreciated effort. The design seems little bit overcomplex to me. Now of course i have some experience designing machines. Fixed positions for lasers - so to speak also 2 lasers might be possible initially. 1 low position and 1 high position. Only move Camera over Z-axis. The distance i do not understand. If you have such tiny turntable of just 150mm why put the camera at a whopping 360 mm (-70 = 290mm) distance from it? If camera moves over Z-axis why not put the turntable closer say at 100mm distance. Moving camera over Z-axis i see the same weird thing you see in many 3d printers. huge leadscrew and small diameter smooth rods and both are way too long. You only want to move the camera over Z-axis. Then i see 2 smoothroads connected to turntable which doesn't make sense to me. As those smoothrods do not really give stiffness to the construction. Of course i assume the thing itself you want to put in a box like fabscan has one. Yet the real problem of this setup might be the turntable. It seems from plastic and using plastic as a bearing. That will rotate around like a pancake of course. You cannot expect software to correct for pancake-rotating mistakes of the turntable. Many bearings, especially in the stronger motors cause the shafts to have no visible 'runout' -  if i measure it here it's order of less than a micrometer at the stronger stepper motors - you cannot buy those bearings easily in the shops by the way - the internal radial clearances are much better than you canbuy for example from SKF which best radial internal clearance is bearings C2 which is roughly until 0.011mm 'runout'. And if i buy those from SKF they go for 9+ euro a piece here or more. Which is why i suggested using a nema34. Then the turntable platform you cannot use selfprinted 3d plastic for this. That i say as someone who designs professional 3d printers. It's simply not flat like a mirror is - and the way how you connect the turntable to the steppershaft there must be manners of adjusting that in X,Y,Z directions. Any micrometer type mistake of the shaft - in case of nema23 that's usually 6.35mm or 8mm diameter. At R=75mm then any error at a 8mm shaft translates to an error that multiplicates by: 75 / 4 = factor 18.75. Now using default bearings is no good idea either as those have a radial internal clearance by default of 0.03mm roughly. 0.03mm * 18.75 = 0.5623mm deviations just by the turntable. Which is why i suggested taking a Nema34 which is 14mm diameter the ones i got here. As then R=7, big improvement over R=4. A turntable that's very level and adjusted to run precisely - then things you put on there do not fall off easily. Seeing all the folks tying objects to the turntable in order to not fall off doesn't come over as very professional. Cheap solution there is using mirrors - even square ones there of 20x20 cm are very flat and then have some adjustable arms attached to the motors shaft that allow adjusting. As for the linear movement i would keep it simple. As short as possible smoothrods and a simple 8mm leadscrew that's also very short. Just moving the camera at most 200mm or so. What's drawn here is 500mm high that's way too ambitious. Of course the leadscrew, turning it around with a motor like drawn here is possible yet i wouldn't do it like this. Yet that's not so relevant compared to the other points above. 

No metal bearings and my background on the industrial side is CNC and Industrial construction (former Industrial Pipefitter in Detroit area). I have designed and built a few CNC Gantry routers that had the major fault of over build on many of them. The reasons for the height and distance is the intentions of adding additional cameras. As I stated this is a early prototype design and the function is to look at ideas and be able to build from there. The moving lasers is due to the fact that I hope to be able to get near 300-500mm height volume available for the scanning area. the major issue with that is being able to offload the image data effectively during the scan as at that volume the Pi would slow down a good bit. The 10mm smooth rods will add enough stability to the arms to allow the lasers to move with the camera which will be needed if the design is able to get the vertical volume I am hoping for. Though as stated this is a beginning design and things will move from there. If you have any designs ideas you wish to share please do. File wise I work with Solidworks, so most file formats are able to be dealt with. Just prefer not dealing with .stl or .3mf for actual design work (output for parts great, design not so much). Looking at a 1205 lead screw as it will allow very smooth motion, reasonable microstepping without loss of positional accuracy and a small amount of structural rigidity. May be over kill but rigidity that still allows for smooth and controlled motion is never a bad thing.

Anyhow thanks for the feed back.

Most interesting!

if you want to scan at 0.3 - 0.4mm accuracy that'll work ok.

To me interesting is a build where the scanner that is as accurate as the human eye can see - so 0.01 - 0.02mm we speak about.

blindfolded i would guess then that one would want to limit the Z-movement as much as possible and lift as little as possible weight over the Z-axis. In that case i do not see a problem in 10mm smoothrods over a distance of 200mm and add a little - bearings are 29mm high and we take then 4 bearings at 2 smoothrods which means we need 58mm extra. So that's say 260mm smoothrod exposed in the open and to also attach it that's say 270mm smoothrod (tapped hole at 1 side and fastener at the other side for adjustment).

Now getting to 0.1mm is a big struggle already. A hard requirement i would guess is a more accurate turntable then if i consider the math needed to determine how the object looks like. Just 3d printing that will give you a pancake that doesn't turn around very accurately.

Stiffness of a structure is a directfunction of distance to the power 3. So if you move the camera over a distance of 500mm then you need more like 16mm diameter rods and a very stiff structure. I would call that an over the top effort going from zero Z movement to 500mm.

Now i do not know whether it is an advantage or disadvantage to also move lasers over the z-axis. I would put the smoothrods at a small as possible distance from each other and use 2 and 4 LM10UU bearings. The LM10UU bearings are in itself not so accurate so you need 4 of them in a square to get some sort of accurate smooth movement with them.

The leadscrew - and what you drew there looks like a German company called ISEL that delivers ballscrews with square nut that are notoriously inaccurate (c7 quality is the highest quality that ISEL can deliver - i have that on paper from ISEL at questions mine) - so a good leadscrew from Germany for handful of euro's is just as accurate there. Yet there is a huge price difference between the 2 and the leadscrew nuts are much smaller and easier to fit into a design making the design smaller. Backlash is not an issue either as during the scan the camera moves in a single direction - it's not going backwards.

So that saves big cash.

In short i would make it all much smaller and not go for 570 mm length smoothrods of 10mm - as that's too long for 10mm smoothrods. Pure theoretic spoken it would only be possible with an extremely stiff structure yet at such huge distances such structure would simply cost too much. Think of a cube made out of 40x40mm aluminium extrusions then that's encapsulating such leadscrews.

What object would you want to fit on a 150mm diameter turntable that's 500mm high? 

As for scanning statues from artists - forget about that - it's not gonna be accurate enough. You would need to scan those at 0.01  - 0.03 mm type accuracy.

Would get a very expensive project and build. 

Better is first build a first careful version that has a very limited Z-range and see how accurate you manage to scan that - and whether artists are happy with that accuracy.

At a distance of 500mm then using 10mm rods which kind of are free in the air is asking for troubles from a stiffness viewpoint. From a stiffness viewpoint seen ballscrews that are free floating in the air kind of or hardly attached to any sort of structure one would be dividing by the constant '3' at a crabble paper calculation whereas a stiff strong structure where they would be attached to you use the constant 192. 

A relative weak structure where the smoothrods get attached to at both sides, we could see as a bridge where the bridge is free laying onto 2 points - so there stiffness constant is 48.

Interesting would be a movement of 150mm to 200mm vertical and then build a very accurate turntable and then see whether it's possible to get accurate scans out of it. The time the software needs is less relevant as long as it's less than a week - after all this is free software made in a high level languae that's factor 10k+ - 1M+ slower than the sort of software i program for prime number crunching. We cannot expect miracles there on a r-pi there and be happy and grateful that this great software is there for free and simply give it enough time to do its job!

We also would need to train the software to compensate for leadscrew errors.  Now if those were tiny i wouldn't even mention is yet C7 we soon could see errors that relative to the absolute correct path can be 0.07mm off easily or slighly more.

Now in itself it doesn't hurt the smoothness of the object, as the error happens very slowly and gradual, just hurts the absolute positioning of the object scanned. So it's possible some just would want to ignore that error. Not me :)

Please note i do not have a legal license of solidworks. I'm using a combination of design software.

OpenSCAD, FreeCAD and TurboCAD (is autocad compatible) i have a legal license. mainly that means that i mainly design currently in FreeCAD and only when something is far too complicated or slow to do in FreeCAD, i'm using usually TurboCAD to finetune objects.

Especially take complicated fillets and such there or correct complicated forms.

Factories i have objects produce from different sorts of metals i usually communicate with using STEP format. That imports, exports and modifies everywhere very easily.

As for the mount of each laser i would suggest something very simplistic. That's U-shaped or tube shaped aluminium and only the last few millimeters you attach a 3d printed object onto it to attach and adjust the laser with. Yet ideally you would want to hang them not too far away from the superstructure (from aluminium) of the  scanner. So the main idea is to use aluminium everywhere for the scanner structure and to attach things use as little as plastic as possible so that thermal expansion difference of plastic versus aluminium isn't a big deal.

Now of course steel would be cheaper and even better there - yet that's difficult for most here to machine. That my machines can do it is not a good excuse to build it from steel.

As the hat  supports 2 lasers i would suggest 1 laser at  the bottom and 1 laser at the top of the scanner.  Exactly how to position the lasers i do not know yet. No experience at all with scanners there.

Using 10mm smoothrods to connect the turntable to the laser part - get rid of that. Simply an aluminium cube or similar made out of extrusions which connects to everything. Away with smoothrod nonsense.