RepRap: Blog

Now the probe is reasonably consistent over 200μm or so, I thought I'd have a stab at the 300μm test image with 15μm spacing. Here's how it came out:

The consistency of the dots is much improved, and we can start to see other variations in a useful way. There were other attempts, of course. Some of the test prints did not show white, photographable dots because they were only partially embedding into the Sharpie layer!

I did a series of dots at decreasing height of 1μm by the way, and the Sharpie layer is about 4μm thick. 

So looking at that with a calibrated MkI Eyeball I'd say our current hardware limitations are an accuracy of about 10μm on the XY plane, similar backlash, and within a couple of microns of height over 300μm of planar movement. So let's work within those restrictions for now.

The latest test results on height showed that the levelling bodge was likely stable overnight. I retested it, tested it in the same place, moved Y +100μm, tested that again, and finally did a test of Y +500μm. Results show some linear displacement on X, but I'm not going to fiddle with that for now.

Next steps, put "dipify" code into the PNG plotting script, put Probe One in the machine, and test some resin deposition. Of course, something else will come up...

Got it down to 1 μm in 270μm!

As per previous testing the probe wasn't staying at a consistent height, but the change seemed to be linear. So I tried propping up one side of the slide by sticking layers of that well-known precision shim material, duct tape. Yes, that's right. If it's good enough for bodging Apollo 13, it's good enough for me.

The tape was placed on the edge of the underside of the slide in stacked ~6mm strips. Each strip is approximately 0.35mm thick. The slide was covered with burnished ~10μm kitchen foil, burnished flat, adhered with UV resin that was pressed before curing. I used "Probe Zero" - the robust hypodermic tip. The results:

The Z axis values are reversed because I'm lazy and couldn't be bothered typing all the minus signs. The red and blue lines here show what happens with 3 layers of tape. The lines cross because I did the test with the tape on the -X and +X sides. As expected, the slope is fairly similar but reversed.

The yellow line is with one layer of tape on -X, and the green two layers on -X. As can be seen, one is too little (1:127), two is too much (1:128) because the slope changes somewhere between the two which would put the elevation needed at ~0.5mm. The previous round of testing indicated an error in the order of 75:1. I was unable to split a length of duct tape, but I strongly suspect one and a half layers would have been just about perfect. Total measurement error appears to be about 2μm judging from the delta between readings.

That represents roughly 13μm height gain/loss per millimetre of travel, which is probably enough for resin deposition of small objects. This only tests the X axis, but we know there is a deviation on the Y axis as well which is yet to be tested for.

My conclusion here was that designing a Stage with some kind of levelling in would take out a lot of the height variations and should probably be done before further resin tests to avoid probe damage. This will also facilitate Y testing. So another simple bodge: Stick split washers under the Stage mounting screws:

The maroon line here is the end result of the fiddling with the mounting screws after five tweaking sessions. Range of motion was 2400μm (2.4mm) and the final deviation was 8.7μm. This equates to the height being 1μm out in 277μm of movement on the X axis, so good enough for me to make things about 500μm (0.5mm) across. That's the approximate equivalent of the width of a human hair over an inch in old money.

Now the Y axis... 

Probe One (the acid etched stainless steel one here) was successfully mounted using the mounting jig and put in a nice, protective box until I'm ready to proceed. There's definitely a knack to it, and the jig definitely helps.

I'm going to get some polarized filters and test them on the binocular microscope first, just to see if it makes it easier to see the tiny little resin dots. There was an article on Hackaday about that a day or so ago, though that was about reducing reflections and I kinda want to increase the reflectivity of the resin dots. Not sure how that would work out.

I had been wondering previously whether some variation on a polarizing microscope of the sort used by mineralogists might work. I have a chunk of LCD screen and a heat gun, but outside weather is not cooperating and it definitely sounds like an outside job in the interests of domestic harmony.

I've tried illuminating from various angles, and that wasn't as helpful as I'd hoped. The best result previously was obtained by tilting the slide at an angle as viewing the droplets from the side makes them more visible. The downside here is that you only have a very small section of the slide in focus at any time.

First, I'll have a crack at sorting out the bed levelling using Probe Zero (a disposable hypodermic needle tip) so I don't munt the probe. 

Yes, disassemble number five. After smacking it around while investigating the probe height problem in the previous test the tip looks like this:

Well, damn. This is why we do not use micron probes for etching. I used the Probe Assembly Jig to get the probe out by the way, and it makes a decent Probe Disassembly Jig. May as well practice the technique on a damaged one.

This has implications for the previous test, in that it was unlikely I was ploughing the point per se into the marking dye. I have no idea at what stage I stuffed the tip, but I strongly suspect the initial touchdown into the resin-covered touch plate.

I'll make a new probe assembly using one of the acid-etched tips and try that out for resin deposition. Probably should image it after after the first touch and see what it looks like before I try depositing resin. It won't be anything like as accurate, but we'll see what data we can get out of it - particularly with regard to robustness!

The upgrades to the Y axis are done, by the way. I've added an 11mm booster under the tension bands to put more pressure on for anti-backlash. I added an extra band on each side too, though the little hook that they hang on to looks slightly over-stuffed with band.

If that works out, I'll put it on Github. Strange though that the other two axes don't show the problem. The Y axis isn't even the one with the highest loading.

I attempted a resin deposition run, unsuccessfully. This was due to a number of factors, most of which are visible in the following micrograph. This was taken with my phone through a microscope that cost me 30 quid from Liddell's, and focus is particularly blurry on the left, but it's the best magnification I can get:

 

This is an attempt at doing the "300μm" plot at a 10μm pitch using the acid/salt Nichrome probe. Why etch rather than deposit resin? Well, the height slope on the Y axis meant that the resin-dipped probe did not contact the surface, so I put down some Sharpie to see what was going on. 

Setup Issues at 1000V/mm 

During the Z-Touch phase I noticed that by dropping the probe speed to 2mm/min I could touch the foil reliably several times in the same location, but not if the tip was in resin. I think this is because the voltage will jump the air gap between the probe and the touch plate. Inaccurate rule of thumb suggests air breaks down and becomes conductive at roughly 1000V/mm so we'll go with that. The probe is at 5V, so at somewhere in the order of 5μm we'll start to pass current through it. The deceleration of the probe will stop it in less than 1μm, so by going sufficiently slowly the probe doesn't actually contact the foil much if at all - in air. On average.

The Y axis driver was noticeably  (or rather, audibly) unhappy. There was insufficient tension on the anti-backlash bands, and the Drive Screw was periodically binding on the printed flexure assembly. I'll relieve the binding and increase the height of the bar with the backlash bands on.

Resin Issues

The main problem was making contact some distance from the touch plate. The next problem was trying to tell what a resin dot looked like amongst the scratches and debris on the slide. Illuminating clear resin for imaging is tricky. Coloured resin tends to have pigment in it, with a particle size of ~5μm. So, no resin results today.

Accuracy Analysis

Proceeding with the probe and a Sharpie, the actual impact points are pretty well distributed along the X axis. I'd call that 10μm or near enough, and I'd hazard a guess to say I could position at 5μm or better without too much trouble. Except for the backlash. The tail on the micron character should be under its left arm. It misses by about 5μm on the X axis. That's the only dot that is subject to backlash. So it appears I have that much backlash in the system.

Levelling Analysis

As the probe strikes the surface at a 60 degree angle to vertical, it will move across it at roughly 1.5x the distance of the over-penetration. From the image, after 100μm of movement on the X axis, the probe skids roughly 5μm sideways. So there is effectively a 1 in 30 slope on the slide. The dirty hack would be to shim up the left side of the 25mm wide glass slide by about 0.8mm, and the proper thing to do (eventually) would be to provide for bed levelling.

So, a few things to fix up, then we'll try again. I'll post a picture of what the probe tip looks like after all this when I take the machine apart. 

While assembling those nice, pointy probe tips a certain amount of ... carnage was involved. I shall write "101 Ways To Trash Your Probe" at some stage. It became very worthwhile creating a jig, as it takes a while to re-make a tip (mostly in getting all the equipment out on the bench, mind). So I created this jig to hold a Probe Tip Arm in place while you glue the probe in it etc.:

The jig features a mark showing where the tip needs to be, and extensions to protect it as you work. Hopefully I have allowed enough clearance so you don't smash the tip on the side of the protective bits. Oh the irony.

The probe wire needs to be bent a bit so that it sits flat in the probe, and I'm just holding it in place with a bit of masking tape while I glue the probe wire into the Probe Tip Arm. I'm not sure if it's worth trying to make a tool to angle the probe tip. Let me know how you go.

It handily protects the probe while the glue dries and as you shuffle things around the workbench. I'm backing off now as our electricity is turning on and off like disco lights. The jig is on Github already as https://github.com/VikOlliver/RepRapMicron/blob/main/maus/maus_probe.scad

>Fzzzit!!!< CLONK WhrrrrimmmM!!!! 

Power Resumed

Sorry about that, folks. Electricity is back. The jig now has a slot in it to allow the Probe Beam to be fitted onto the Probe Arm without removing said arm, and you can nut the two together in the jig - admittedly a fiddly process unless you have a small nut driver. So anything that reduces tricksy handling of the probe manually is a good thing! There's also a little notch underneath so your finger can push the screw up into the Probe Arm. I've updated the Github.

 

I got some actual 316L stainless steel wire and some Nichrome 80, after the "Art of Bulshito" non-stainless non-steel wire incident. I decided to try etching them in either 1% nitric acid or 5% salt solution using the rig and instructions here on Printables. Using 3xAAA batteries for power, immersing the wires 8mm into the liquid, and with a stainless steel spoon/cathode 40mm from the liquid's surface. Blimey, wasn't expecting this:

 

Using nitric acid seems to produce a very robust probe tip, with an almost foamed shaft. There was indeed no cloudy precipitate during etching, but the tips of the wires did not fall off. What you see in the top photos is just the eroded end of the wire.

The salt was as cloudy as ever. The 0.12mm Nichrome dropped off at the meniscus. The 0.3mm Nichrome parted about half way down the immersed section.

The tips were gently rinsed with water and sprayed with isopropyl alcohol to remove any remaining liquid.

It appears that nitric acid is best for creating a robust tip. Nichrome creates extremely fine tips, but using wire as fine as 0.12mm diameter produces a very short tip that appears uneven and particularly delicate.

I think the most impressive tip of this batch was the very fine, spire-like tip produced by the 0.3mm Nichrome in salt. The tip is exceptionally small (too fine for the camera), the length has a gentle taper, and the sides look like they would hold a good dose of resin by capillary action. Here's an image zoomed in on that fine tip, with a piece of 0.3mm wire (almost in focus - sorry, it's hard) for scale.

That tip though is really fragile, and finer than I need - I had to make a jig to protect it during assembly. I wanted something half way between nitric and salt in dimensions. Well, lets put about 5ml of 1% nitric and 35ml of 5% salt solution together and... interesting. That stops the precipitate forming, and I can see the point sharpening nicely - hydrochloric acid might do the same and is easier to get, which I'll try later.

Anyway, that combination allows me to get a good tapered point, and tune its thickness to my requirement of the day. Now to join it to the μRepRap and see if I can deposit an even smaller resin Jolly Wrencher with it. 

Doesn't look as fragile as the long thin one (foreshadowing).

Yep, stung again by an Amazon supplier. Advertised as 0.3mm Stainless Steel Wire. Absolutely not, as you can see. Some plated copper muck:

 

So three spools of that can go in the bin. My first clue was that it started copper-plating the spoon, er, cathode in my electrolysis setup:

FYI I was using 1% nitric acid rather than my usual salt water. I intended to see if that would stop the production of cloudy iron hydroxide so I could tell when the end fell off the etched probe more easily. Turns out I need to try a different batch of wire, huh?

The objective was to find out how consistent the μRepRap was when using an aluminium foil touchplate. Simultaneously an overall difference in height was observed on the Z axis when the X axis was displaced.

A glass slide with UV resin-bonded aluminium kitchen foil burnished in place was placed on the stage and connected to the Z-Touch ground probe. 0.5mm approx of one edge of the foil had been bend right over prior to burnishing, presenting two layers of foil. There is a limitation here in that irregularities in the foil will cause cumulative errors in the height of the second layer of foil.

cnc-js was then used to position the probe over the foil, and Z-Touch data noted at 300μm intervals along the X axis on single and double layers of foil. The separation in the Y axis was 300μm, however it had to be manually adjusted in cnc-js by approximately 50μm due to inconsistencies in the preparation of the foil edge in the XY plane.

Repeated probing of the same point was not possible. The probe point cannot be instantaneously stopped, and deceleration after time of contact results in holes being poked in the foil.

The resulting data were crudely spreadsheeted.

Analysis shows that the difference in probe heights between the one layer and two layer probe points on the X axis was in the range of 10-11.7μm, suggesting that this technique is likely capable of locating initial probe Z height to within 2μm. 

Closer examination showed the Z values decreased as X displacement increased at what appears to be a linear rate given the sample size. Maximum vertical deviation was 28.5μm over an X distance of 2.1mm

Concluding, Z-Touch capability is suitable for at least initial probe height determination, and possibly may be sufficiently accurate for subtractive operations. If the probe tip is to be kept within 1μm of the surface, a working area of +/-75μm could be covered without compensating in software.

Note that the cause of this error is unknown. It may be that the stage is slightly tilted, or that tilting the stage will fix the issue regardless. However the Stage can reliably relocate within 50μm when homed, less than the distance for 1μm Z deviation. If the error really is relatively linear it will be possible to consistently compensate within 1μm accuracy - at least on the X axis.

Well I had a hypodermic needle probe in, and there was some resin, so I tried it as a deposition probe. Not very good results, wasn't really expecting much. Here's a micrograph:

 

Initial contact point is the top left of the square. White lines are probe tracks in Sharpie, blobs on the corners are cured resin. The sides are roughly 100μm a side (sorry, done manually) and so the blob of deposited resin looks to be roughly 30-40μm diameter. Not good enough. Will have to make a proper micron probe, which last time got me better than 10μm dots.

Aside on microscope image sensors: This cheap one (sub-Mpixel) I'm using here is heinously sensitive to IR light and grubby as all heck. If I try illuminating the sample with a halogen bulb it instantly whites out. Had to take this using the LED desk lamp. Better 23mm dia USB image sensor is on my Xmas list if Santa is listening... 

Every time I see those black and white micro scale images posted by the MEMS crowd I get a bit of calibration envy. The images all seem to have those really useful scale bars which provide an indication of the small size of things they are displaying.

Well, no more. I have added calibration tools to the software which controls the experiments in the FPath project and, henceforth, my images will now also contain a scale bar.

I just thought that some of you might be interested in how it is done. The video explains all:  https://youtu.be/nirIiC6hnc0  

The image below shows the scale bar in place while measuring the length of the claw at the end of a bee leg.(click on the image to enlarge, watch the video for context)

 

 

Labels: Feynman Path, FPath, millimeter scale calibration, Windows Media Foundation

I've been building up to a resin print test campaign, starting end of next week. So have been working hard on getting V0.05 into a usable state. Hardware is running quite happily down to 10μm per pixel, even in engrave mode. Here's the hackaday.io "Jolly Wrencher" logo etched into Sharpie marker. That's a 50x50 PNG done at 10μm per pixel, making the whole thing 0.5mm wide. Zoom in and you can see a bit of a gap between pixels, so I might even make 8μm.

 

 

This is a big little deal. Achieving 10μm is what I said I'd be happy with as a positional accuracy goal for this prototype. Of course, I need to get the resin going still. But the reason I picked that number is that it was the feature size of the original Intel 4004 CPU, and it seems to me to be some kind of tipping point.

Below is the new probe holder, allowing three degrees of freedom. The new probe holder requires you undo one of those vertical screws at the top to remove it, and requires a bit of fiddly setup to get the length right. But it is much more easily positioned over the UV source than the V0.03 design, and as V0.05 has such a massively improved range of motion on the Z axis you don't need to manually adjust it so often - just wind the Z axis up out of the way.

Below is the view I get on my "microscope" console with the print about half way through. The other console is taken up running cnc-js. Using mpv with custom pan/zoom macro keystrokes on the keypad instead of vlc has made the view much more useful.

Oh, note that the 10μm image was done using a 0.5mm hypodermic needle, not even a fine tip probe. That's the big, fuzzy grey thing centre to top right.

List of To Do's is growing a bit. I want to run resin tests using a hypodermic, after seeing how it did in the above test. I want to test Z Touch on a doubled sheet of aluminium foil to give me actual numbers on the thickness of the foil (which I shoulda done and never did). Finally, the bugbear of having to do version control and docs is getting kinda important.

Just as well I'm kept up at 3am by a helluva storm raging outside, ay? 

Update: Build files now on Github https://github.com/VikOlliver/RepRapMicron