RepRap: Blog

A clean thread on the drive rod helps a lot. Who'd have thought? So combined with the new slide holder (picture below), testing is now possible again. I tested the Y axis out to 0.5mm with the X axis disconnected, and adjusted so that it is spot on at +/- 0.3mmand  it is repeatable. Take it out to 0.5mm and it thinks it has moved 0.53mm, going back in at +/-0.2mm it thinks it is at 0.23 which again is repeatable, so should be able to compensate in software.

 

Connecting the X axis changes Y positioning by less than 0.02mm and that's very close to my ability to actually test against a 0.01mm grid - plus the changes are consistent.

The X axis, however, does not show anything like even movement even with the Y axis disconnected, so that's the next thing to come apart on the bench for inspection.

I was wondering why my probe failed to move when the thread rotated on the drive motor. Turns out there's an f*****g great dent in it! Looks like The Chosen Screw needs to be very, very carefully inspected before installation. This particular divot causes a 0.1mm deviation in the probe movement.

It also shows that the drive nut is making extremely limited contact with the thread. Maybe longer nuts will help? This is a good reason to use longer lever arms - slight variations in thread manufacture will have a less significant effect on probe movement.

I'd like to introduce you to the Micron Accurate Universal System, or "Maus" Prototype. Rather than a monolithic enclosed print, this is a collection of drivers, flexures, and mounting brackets that can be reconfigured. This allows different parts to be easily swapped out, modified, individually tested, or completely rearranged. The use of Metriccano assists the rapid development of parts.

Banana for scale.

For reference, here's the original Meccano RepRap prototype from early this century. We come full circle:

Control is from the same RAMPS/GRBL board as the previous μRepRap prototypes. I haven't fitted endstops because (a) I haven't got a good design yet, and (b) it is relatively easy to determine and manually set the zero point on the drives. The stage is currently set up to hold a calibration slide, but I'll be adapting the previous small slide holder with Z probe ground to fit.

The Z probe holder itself is going to be redesigned and made flatter, which will allow the USB microscope to get closer to the stage and increase the effective magnification. Learned that problem and increased the clearances.

Importantly, the whole design is now GPL'd, so I can share it on file sites and let everyone join in without worrying too much about licencing. This does mean I'll have to significantly update the Wiki page, create a Github project, and tidy the OpenSCAD code so it doesn't look like it was written by a hyperactive squirrel that has OD'd on coffee ice cream. That said, hit me up if you want a current tarball.

Just ran up some tests with a calibrated slide on the XY platform. Supposedly identical motors and drivers on both stages. The X axis is running at about 9,000 steps/mm and the Y at 21,000 steps/mm. The X needs a bit more anti-backlash, as the backlash is measuring in at about 8 microns. Still, not too shabby. I'll add more rubber bands and duct tape...

I made some Metriccano tabs joined by 50mm flexures and couple of spacer plates that allow me to connect the X & Y drivers to the XY table. Over the range of motion, these will need to drift a bit, but calculations show this is less than 0.2mm and should be within the flexibility of the table to contain. The assembly looked like this:

I then took the functional prototype, made a few minor changes, and recreated a model in OpenSCAD using the Metriccano library parts. This was condensed to eliminate most of the fastenings and tunnelled to allow flexures to pass through beams etc. Here is the after and before:

 

The part then drops right into the original assembly like so, still with Metriccano grid holes in it so that I can make modifications during further prototyping. Though I can of course set the hole diameter to zero and leave all beams solid if I wish.

I had an idea for later. This is lending to a flat-pack construction and the use of fasteners is gradually being eliminated. This will make it easier for a RepRapMicron to produce a small model of itself, theoretically with greater resolution. All the parts for a plane of the smaller printer can be printed in a frame, which the RepRapMicron probe can then swivel into position like a combination of an  IKEA set and origami. By dyeing parts of the print during manufacture to absorb light, the idea is that a laser or similar will be able to simultaneously but selectively blast away (thin, dark strands) or weld (chunkier contacting dark blocks) the components in the frame, which can then be dragged away leaving the free standing miniature printer.

I don't quite know what size or shape XY table I need. Rather than print lots of them, I've built a set of flexures for Metriccano, and knocked up an XY table with them. This allows me to pursue the scientific method and change one attribute at a time. Linkage types, attachment points, number of legs etc.

Yes, I know there is IK modelling, but that doesn't take into consideration where you need to put the wires, what fits on the bench, where you can stack the microscopes and so forth. Not without a lot of work anyway. Anybody wants the GPL'd files, go to the comments and tell me.

As "Titch" is out of action, I've put together a new stage driver with anti-backlash on it. It checks out to 10μm as near as I can tell by averaging and eyeballing the 0.2mm layers of 3D print on the stage as it moves. Better accuracy test with calibration slides coming when I make a Metriccano slide holder mount.

The lugs for the bands have been moved above the nut, which drags the nut rather than pushing it and removes the wobble. The bands are "Loom Bands" which stretch quite a lot and are stable over time (unlike rubber bands). They are quite weak, however, and this allows me to add several to alternate sides without stressing any flexures.

The crossbar conceals an M3 nut drilled out to 3mm. As a stop on the thread I'm using an M3 Nylok because dual lock nuts in confined spaces are just too damn hard. The upward flex in the coupling ensures the bottom of the screw does not pop out of the coupling.

The coupling does need to be manually centred by pressing hard on the high side until the unconstrained screw rotates centrally. Seems to work though.

There are a couple of other things going on, but I'll cover those later when I'm a bit more sure of not making an arse of myself. More than usual, anyway.

The Y axis failed during a probe test (yes, the Z axis has a probe fitted, kinda). After a bit of poking around, I found the Y axis drive screw was jamming. I wonder why? Let's pull it out and have a look next to a new one:

Well thars yer problem.

That is what we in technical circles call "well shagged, mate." I'll replace it and resume probing, but I think the issue here is the screw resting on a plastic bearing hole, which is dragging plastic into the thread, filling it up, and causing excessive wear. I hope the drive nut is ok. 

Edit: The Y axis has now crapped out and only has 1mm of travel. At this point I'm tempted to begin the Alpha Prototype XY stage.

Big update. Don't ask about Mod. 5, I got my maths wrong ok? As you can see, Alpha Prototype Mod. 6 Z Axis is much shorter than previous versions. The flexure arm only has a 4:1 advantage (and has a vertical flexure in it), and I've taken the backlash mechanism out because it's a pain in the butt to assemble/disassemble repeatedly. The Z limit switch is not yet installed either. I've substituted a 6mm length of hexagonal brass M3 pillar for the drive nut which is press-fit and it'll do for now.

The screw support beam has an M3 nut in it as a bearing for the top of the screw, drilled out to 3mm. The coupling is again press-fit onto the 5mm NEMA17 shaft, and grips a hex head M3 x 50 by - you guessed - press fit. The coupling itself is a flexure design that gives a bit of play. Behold its labyrinthine internals (GPL'd STL/SCAD on request):

The motor sticks out the bottom a bit, so I've had to prop it up on some Metriccano struts to get it on the stage. I had to move it back a bit to leave room for the probe, so there's a lot of plastic cribbing. TBH still not figured out how to mount the probe to it. Next step.

Things have been moved around on the stand to make room for the Z axis and all the microscopes. I'll probably fit a manual thumbscrew to the OpenFlexure Z axis and run mine off a stepper motor. Comparisons will be entertaining.

So with the short arm, how's the accuracy? Well, promising given I've taken the backlash off. I tested it by pointing a USB microscope at the Z stage, marking the screen on a layer line, and stepping the Z axis up and down. You can certainly step through the 0.2mm layers with a lot more than 10 steps/layer until the lack of a backlash system bites you - the thread wobbles sideways as well as up/down if you don't support it, and that also plays hell with the height. But I have an idea to fix that too, which seemed to work on Mod 5.

On the plus side, dang this Z axis is fast. 20mm/min which sounds like a snail's pace but given the Z axis is only meant to have a maximum range of +/-2mm and an operating range of half that, you can completely clear the work zone in 12 seconds. The coin was probed with a safe working height of 0.15mm, so things are going to speed up a bit. When the probe is fitted I'll test that and compare.

The Mod. 4 axis was a multiple fail. Its improvements were good, there just weren't enough. Actual frame rigidity was not good. The NEMA17 coupling was a press-fit sleeve with the drive screw bolted onto the end. Unfortunately this tended to set an angle to the screw thread and everything wobbled.

After some time I realised that I needed to hold the drive shaft in the collar flexibly. So I put a fixed nut on the drive shaft, and the motor turns the shaft. The nut hole is tapered, allowing it to be firmly grabbed, yet allowing it to move to re-centre rather than eccentricity driving the shaft.

 

The OpenFlexure stages avoid this by having a gear between the motor and the drive shaft, and that absorbs any wiggle.

The nut trap in the drive end of the pivoting arm has a lose cavity for the nut. This allows the nut to rotate partially. Tighten it, and the nut is hard to insert. The solution again is a conical nut trap. The temporary solution: Jam it in with a bit of copper wire to take up the slack.

Finally, lubrication is very important, particularly if you can't find brass nuts.

I may insert a pair of vertical flexures into the Mod.5 lifter bar to see if this will remove any residual sideways wobble and/or try making the other flexures slightly more rigid.

I may need more anti-backlash tensioning but that may need to wait for decent springs.

Assembling Mod. 3 was difficult as the exact length of the motor shaft and placement of the bearing lock nuts is crucial - and they're real swines to access. In the process, I snapped the flexure frame (see pointer).

Mod. 4 has that bit strengthened, and the bearing beam is removable to improve access to the nuts. Printing now.

In white behind the orange coupling you can just see the crossed wire Z endstop, currently held in place with a zip tie and hot glue. It'll do for now but will change that later.

Last century the rapid prototyping tool of choice was a metal child's consruction set of bolt together girders, straps, and gears etc. called Meccano. Many of the original RepRap experimental prototypes - including the first functional extruder and bed - were constructed from it. I've combined the attributes of this with 3D printing to basically use adjustable straps and brackets based on a 10mm pitch, 5mm beam height, and M3 fittings. OpenSCAD parametric library files to follow as I tease them out of the main files. As the original Meccano was in imperial units, I'm calling this Metriccano.

Here you can see the Mod. 3 prototype with a little thumbwheel on top, supported on the OpenFlexure Block Stage with Metriccano. This runs in conjunction with the existing Z stage so I was able to use the height probe to measure performance. The result: I don't trust my Block Stage calibration. However, I am getting smooth movement and an averaged mechanical advantage of approximately 4:1 - the probe moves in an arc, so it's not an easy thing to measure.

Next step is to motorise the Z drive screw and perhaps add a limit switch. The motor will go underneath the mechanism, and the thumbscrew will become the top bearing.

As to overall accuracy, it is theoretically overkill. Range of movement and permanent deformation of the flexures is the limiting factor. For example, the motor has 400 steps/rev. and an M5 threaded rod has a 0.5mm pitch. So that moves at 800 steps/mm by itself. With a 4:1 advantage that gives 3200 steps/mm or 0.3μm per step. And I can reliably quarter step the motor.

I've added a drive screw to test the Z axis movement. Yes, I know, it uses 'O' rings. I have a box full and have yet to purchase extension springs but that will happen later.

The height for the drive screw bracket is wrong, the underside of the crossbeam prints ugly, I don't like the hooks, and many other things. Also, need to do new probe tip mounts and get away fron the old design that was based on hypodermic needles. Busy, busy.

The OpenFlexure microscope is excellent, and I highly recommend it. Its axes print in one piece, all the moving parts are encased in a sleek outer shell, it uses inexpensive steppers, and pans a microscope slide around nicely.

Unfortunately, it is a microscope, not a developmental printer. Those things do not fit well with a prototype system where the parts need to be observed, fiddled with, modified, swapped out, stuck underneath, and poked in from various angles. There is also a problem with the licence in that most common 3D model web platforms don't support it and it is incompatible with the GPL.

So, I have started development of a more modular 3-axis (possibly more...) system that should be more suitable for incremental development. This is the first crack at a Z axis for the Alpha Prototype. This has a 5:1 mechanical advantage, which with a 1/4-stepped 200 step motor directly driving a 0.5mm pitch M3 screw has a theoretical resolution of 0.5/(200*4*5) 100nm - in my dreams. It is designed to have an operational range of +/- 2mm, but will likely only be used for probing in +/-1mm of that to minimise lateral motion due to curvature. To this end it has a 14mm radius on the platform lifting arms, giving an overall sideways drift over the 1mm probing range of approx 36μm or about a 1 in 28 slope.

 

The probe itself will mount on the right-hand face, which I have studded with mounting holes.

Next step is to design a manually operated anti-backlash base to test it out with a thumbscrew. It should mount on the Titch probe platform, so I can compare it directly with the Z motion on the OpenFlexure Block Stage.

With the same resolution as before (50μm) I scanned the smallest details I could find on the NZ 10c coin: the letters 'IRB' found just above the year numbers and so small most people don't know they are there. I misplaced the probe a bit and only got the first part of the 'I' but the results are clear enough. It's reaching the limits of the robust probe tip:

This is what the original looks like, shown on the stage with my horribly clunky probe below it. The letters approximately protrude 40μm and are 0.7mm high, or for those used to colonial measurements, just less than half the width of a piece of printer filament:

With the previous scan of the 'A', the xy slope adjustment was (0.2,-0.06) and for the 'IRB' scan, (-0.018,0.033) so it is fairly clear that the slope changes drastically depending on where on the stage the probe is. Unfortunately I did a demo to my brother-in-law just before making the scan and lost my zero, so I don't know what point the probe started from.

During the demo scanning process, the coiled copper grounding probe lost its spring and would not maintain contact, so I affixed a little PET plastic arm with a nut to ensure downward pressure. I kept nudging the coin around while putting the ground probe on, and I'm afraid some duct tape was deployed for stabilization...

Here's the little perisher. I know a weak point when I see it so I purchased 3 spares. Now 2 spares.

Fitting the new one requires a 3D printed tool and a lot of brute force. They get hung up inside 2 out of 3 times, requiring fishing around with tweezers before trying again. Still, better case than fixing the Y limit switch.

UPDATE: I thought the Y-axis limit switch (actually a couple of sprung, crossed wires) had failed. This is, of course, right in the guts of the thing and I'd have to unscrew every single component of the base and remove all the 'O' rings that are under tension to get at it.

The update, once I'd taken the Y motor off, is that the switch is fine but the 'O' ring in the Block Stage has failed. These are starting to be a royal pain, and when I design my own stage they will be the first thing to go! But, once I unbolt the stage, I should be able to replace a broken 'O' ring without having to dismantle it. Of course, an attachment point could have broken off, in which case I'll have to print a whole new stage and do all that tedious disassembly/assembly.

Also, the probe with the jumper pin sticking out the side won't fit in the nice storage pot I made for it. Damn. Bigger storage pot needed. Fortunately, this is a very tough probe that I use for test-fitting things so it can rattle around until I'm done.