RepRap: Blog

This looks terrible, but there was no anticipation of a great success so that's OK. Many variables were changed. The print of a 500μm diameter Micromug was attempted using the new - much faster - segmented plotting algorithm, breaking into 15μm segments with 200 segments per dip. That was ample to complete one perimeter. A little Vivid marker was accidentally added which improved contrast a lot. A simpler slide was used with no folded foil edge, and it just relied on the resin curing in UV as it overflowed.

Disaster struck at around 100μm when the probe lost contact with the workpiece. Looks like I need thinner layers. PrusaSlicer parameters were:

• 35μm extrusion width
• 3.5μm layer height
• 1 perimeter
• 15μm nozzle diameter
• 1 solid bottom layer
• Random seams 

One improvement was taking screenshots from the side view camera during layer curing. This allowed me to compare the speckled reflection of the UV LED from one layer to the next. No change in image means no new material added.

The probe was cleaned off with IPA and a tissue afterwards. I'll deliberately paint the probe with a Vivid next time, as Probe 9 is robust enough to survive this.

If at first you don't succeed, try again at greater speed. 

I've found the maximum speed I can run at without causing excess vibration: 8mm/min. I noticed the probe accidentally touched the slide when I was moving at 14mm/min doing some tangram shapes, so I did a bit of experimenting.

To find out what the speed limit is, pick your safe Z height -  the height to which you feel happy raising your probe between moves. In my case this is 35μm. Then you put a slide covered in Sharpie marker under the probe at that height, and start issuing manual GCODE commands to move 500μm at ever increasing speeds until your motors slip or the probe vibrates so much that it whacks into the slide.

Here's my test. You can see the horizontal lines moving in a square pattern where the probe starts digging in.

 

This was with Probe 9. I haven't tried this out with other probes, so I don't know if this is a probe vibration issue or a structural vibration issue.

It occurred to me that if I'm drawing with resin rather than placing dots, I'm not so fussed about micron accuracy. That's because my plot is going to need to be 30-40μm wide to have structural strength. So if the probe drifts by 5μm I don't care at that resolution.

Soooo, I can just draw lines rather than do little dots. I'm modifying the "dipify" code to support that straight out of PrusaSlicer. I still need to break the GCODE lines up into short segments so that I know when to go and dip the probe in more resin.

Still, it should speed things up a bit when printing. I can switch back to dots if I need more accuracy or have a more wobbly probe.

Sometimes your experiments do not prove your hypothesis. This is one of those. You have to remind yourself that it didn't "fail" it just returned unexpected results.

After trying to draw solid lines with resin and getting no adhesion from the Dutch Guilding (UNAOIWN Gold foil GF01 AU) I just went medieval on it and put down blobs of resin. Then I dragged out one of those blobs into thin lines with the probe. The foil was then laid down, pressed with a lightly compressed tissue, placed under a sheet of paper, and burnished with a chromed rod. 

Burnishing was used as the foil has a very dimpled texture and there was no other readily apparent way to ensure good contact with the resin on the slide. That should fix it.

After plentiful exposure to UV from underneath, the slide was rubbed first with a brush, then a cotton bud. Much foil randomly adhered to the slide. I rubbed much harder. The slide was then examined under the microscope:

 

As can be seen, adhesion by the foil to the resin is more or less random (the circular structure is part of the slide support). There was very little correlation between any of the resin blobs and the locations of patches of foil, some appearing stuck down with optimism and stubbornness on areas that had not acquired resin.

As well as being dimpled, the material appeared to fracture easily into straight-edged flakes with a distinct grain, in a way that real gold foil does not. 

So, not much of immediate use here. But I do have plenty of guilding for Christmas decorations. Might try using the probe to machine the foil later, so I'll hang on to a few sheets.

Real gold foil may be obtained later, and I suspect that will behave somewhat differently. 

Dutch Guilding is used as a cheaper and more amateur-proof alternative to gold leaf. It's a form of very thin brass sheet, therefore conductive and non-edible. Although thicker and sturdier than real gold, it is still inadvisable to leave the book of it open on your desk when you turn on the air conditioning. Ask me how I know.

The hypothesis was that it would be possible to adhere guilding to the glass slide using deposited UV resin. Previous rough tests were inconclusive, though resin was observed on the resulting fragments of leaf.

I printed an 8x8 hollow square at 50μm spacing using Probe 9, allow a 25mm square of generic Dutch Guilding leaf to lay itself onto the pattern, and pressed it carefully with a loosely-wadded tissue. Then I pressed more firmly, taking care not to push sideways.

Finally, I transferred this to a UV lamp and cured the resin excessively.

The leaf is so thin that you can see the UV LED shining through. After carefully removing the fragile leaf with a camel hair brush, I found ... nothing.

A bit of a null result. I suspect that the resin adheres more strongly to the leaf than to the glass, and when I peel the leaf off, the resin sticks to that. This may be useful at some stage, but it isn't right now.

I did manage to make a test blob approx. 0.5mm diameter stick to the glass. This suggests that the force necessary to peel the resin off the slide for a given area needs to be greater than the shear force concentrated at the edge of the resin blob. The leaf then tears, and the resin remains on the glass. Maybe.

So basically, try bigger blobs (or more dense dots, or both). On the plus side, this should work somewhat better with actual gold leaf, as this is an order of magnitude more fragile than Dutch Guilding leaf.

I put the RepRapMicron through a few more paces and found the Axis Driver I had fitted wasn't quite concentric and didn't have enough tension on the anti-backlash bands. Here's how to fix:

To put more tension on the bands, move the axis so that there is minimal band tension. Lift the groups of bands one group at a time with a screwdriver, and slip the booster in underneath. There are no fasteners, it's held there by band tension and lugs.

Then loosen the two screws next to the bands a bit so that the bar supporting the bands can be moved around. While driving the axis, adjust the position of the bar for minimum unwanted wriggle of the drive screw. I do this by watching the squared protrusion that contacts the limit switch. When it is in a good position, carefully tighten the two screws again.

I tighten the screws in gentle increments alternately until they're tight. This will all minimise unwanted axis deviation. It's a shame this is still necessary, but at least it's much simpler to do than the previous Axis Driver design!

While looking for reference images I came across the technical paper that started it all for me: The technical report on the original Meccano RepRap proof-of-concept. For those interested, I've put the document up on ResearchGate https://www.researchgate.net/publication/398027280_Construction_of_Rapid_Prototyping_Testbeds_Using_Meccano

I already had a company called Diamond Age back then. Marvel on how things have gone in the last couple of decades.

 

It's been a while coming, but at last I have managed to do multiple layers. The big question is: How many? Well, about 10 before things turned to custard on this attempt. But the custard was very informative. Take a look:

There are three major objects, and some insignificant little lumps that are me checking probe height. In the foreground, slightly fuzzy (sorry, depth of focus is not a well-known trait of microscopes), is where I started printing a 200μm long "test lollipop (a 100μm circle with a similar size line coming out the side). It became apparent pretty darn quick that Probe 9 was depositing bigger blobs of "Top Coat" nail resin than it had in dip testing, so I aborted that one.

In the middle we have a 400μm long (i.e. 200μm diameter) lollipop made with layers 4μm high and assuming a 30μm wide dot size. I got a few layers in before I noticed irregularities around the circumference which you can just make out. This was where the probe was not touching the surface below and so no resin was transferred.

A few notes:

First, it is really, really hard to see the layers or indeed what is going on. So it is not obvious when the RepRapMicron is not working. Consequently things can go hilariously wrong and you can't tell until you take the slide off, tilt it sideways, and shove it under the big microscope.

Secondly, it appears that any defect in the layer tends to accumulate resin. This means that if all your print seams line up, a really big defect will happen (I'm looking at you at the back). So, lesson learned, stagger your print seams. When the probe goes to dip and recharge with resin, that can create a seam too - that's where the biggest blob came from. It may be that making sure the probe runs out of resin will lessen this problem.

Kinda related to that, the first resin drop is always a bit big. It may be necessary to make some form of print tower ("The Tower of Ooze" as we called it way back) to get rid of that.

Finally, the probe is not just brushing the upper surface. It is stabbing in to semi-set resin. On that scale the resin doesn't just run downhill because surface tension is having a massive effect on the resin - far more than gravity. So your resin will stick to the top surface, the side of a surface, or any damn where it comes into contact. This may be useful for bridging later.

Back to our picture.

Bringing up the rear is a 500μm diameter single-walled cylinder with some kind of blob sticking out the top. That was done at 3.5μm per layer with an assumed dot size of 40μm ,and a dot spacing of 15μm. It was depositing 60 dots between dipping the tip. Every time it was dipping, the dip code briefly turns the UV LED on to gel the already deposited resin.

That one got to about 10 layers. You can see the top surface (ignoring the darn blob from the seam problem) has a square edge, not a rounded one. This indicates that layering is happening rather than I'm just making a thick line - there is no surface tension curving the upper surface.

So by that reckoning that cylinder is roughly 35-40μm high, and the deposited line is roughly 40μm wide. Which looks about right.

For a first stab at a multi-layered object from an STL file I'm pretty pleased with that result. A slightly thinner or more steeply angled probe tip, staggered layers, a more accurate measure of how many dots per dip, some kind of method of knocking off the first dot from the probe, and I think we can get better resolution and significantly more than 10 layers out of this. Maybe overhangs too.

Just a quick update. Fitted the Axis Driver that I just made for Siddharth's photos. Moved straight and homed first time! Note that I had previously moved the black bracket it is attached to up one hole when fitting the previous test axis. I haven't fitted the booster that increases the band tension, but that can be slipped in later if needed.

 Next steps:

• Make a new slide.
• Test layers with Probe 9.
• Finish Everything Open presentation script. 

Carrying on from a couple of posts back, here are the remaining stages of assembly for the RepRapMicron V0.05 Axis Driver.

The things not obvious from the photos are:

• The red Flexure Coupling on the motor is in a separate OpenSCAD file.
• You can use different height NEMA17 Plates to match the length of your motor's shaft.
• Adjust the position of the pair of locknuts so that the base of the Drive Screw goes into the hexagonal socket on the Coupling. 
• Very lightly oil the drilled-out M3 nut bearings, the M3 x 50mm Drive Screw, and its washers. 

Pipe up if I've missed anything.

 

Labels: driver, Pantograph, reprapmicron

Not my own work, but these enterprising people have adapted the mouthparts of a dead mosquito into a fine extruder nozzle, getting down to 20μm. An interesting approach. While not one we might use immediately (though do go ahead if it tickles your fancy), it does demonstrate that we can nick bits from nature if that's what it takes. I'm quite interested in the delicate silica structures of diatoms myself.

Screenshot from https://techxplore.com/news/2025-11-repurpose-mosquito-proboscis-3d-nozzle.html below.

 

Siddharth pointed out that the assembly of the Parallelogram Axis Driver isn't shown anywhere. Pending proper documentation I've taken a bunch of photos for the tricky bits.

I updated the OpenSCAD files an hour ago.

Probe 9 has been checked under a microscope after the last test, and it stood up well to being cleaned with isopropyl alcohol and the corner of a tissue. 

Hospital visit did not go as expected and I'm minus a couple of strips. Going to take a few weeks to heal up.