RepRap: Blog

I've made a small test flag and printed it with a lug to mount it on the Z Axis with. The "flag" was flipped around so that it was near the contact point - much handier when you're trying to watch everythign through a microscope.

I can certainly see movement that happens over a micron or less. This is done by bouncing the probe 10μm. If it twitches as it gets to the bottom of the movement, then you're contacting the slide. If it doesn't, drop it by a micron and bounce again. It makes for a boring video.

There's a bit of tidying up to do, making things align in the microscope's field of view, making sure there are more straight edges when you're looking for motion etc. But I might be on to something.

Options might include making it into a proper indicator gauge, and being clever with capacitors or inductors to give electrical feedback.

Detecting when the probe contacts the surface is not easy. When it gets there, it just stops. This makes determining bed levelling difficult to do by just probing at it. Even using an electrical probe has problems, as that relies on a perfectly flat contact surface and a perfectly clean one at that, with no oxide build-up anywhere and minimal electrical field effects. This never happens for me.

So I wondered if it might be possible to make a mechanical surface probe. I have a DTI gauge on my lathe which is notionally accurate to 10 microns, after all. I came up with a "Scott-Russell" flexure system with a flag sticking out of it. This is basically a flag on a very unstable toggle joint. I made several sizes: 

The sensitivity to motion less than 10 microns seems to be roughly comparable to that of the DTI, but the flexures require far less force to actuate. By just watching for the flag to start to move, you can fairly easily pick up a 10μm tip motion with the naked eye. 

These examples are printed on a Prusa XL, but it would be interesting to try to print a smaller one on a μRepRap. Doing that with an improperly calibrated μRepRap might not be possible, but if I can keep the delicate parts arranged in one direction then you only have to calibrate reasonably accurately along one axis for a relatively short distance. The theory is that once printed, one of these could be stuck to a probe tip, and then - when observed through a microscope - be used to level the Stage and determine the height of printed structures accurately.

Lots of if's and but's, dubious rigidity of the UV resin,a desire for the thing not to swivel sideways on contact, concern over separating the thing from the Stage etc. Might work, might not. Won't know until we try, will we? 

 

["Do You Have A Flag?" reference is to a comedy sketch by  Eddie Izzard, go look it up.]

The objective of the test was originally to attempt to measure the height of deposited droplets by banging the probe into them and looking for bumpy motion using the 4K microscope. Unfortunately the probe was rather destructive, and I was unable to get anything out of it other than "well, it's 1-2 microns-ish". However, I happened to record the deposition process, and CC subtitled it for YouTube:

I doubt YouTube has done the original resolution justice, but even zoomed in on the original I could only just make out the droplets being deposited. This was all being controlled manually, and thus a wee bit haphazard. I set the CNC step size to 50μm and just slammed it around. Under software control, the probe is decelerated when it gets to within ~10μm and then is lowered gently, so those tend to be a bit more uniform.

As "measuring by braille" didn't work, I stuck the slide under the microscope horizontally and vertically. Here's the images but I couldn't determine much about the droplet height from it:

One of the big problems with imaging things at high magnification is that your depth of focus is about half a gnat's whisker. So even though there were 5 dots, I could only get two of them in focus at the same time. I could probably do a bit better if I made an improved swivelling slide holder, but that's a project for another day.

I've just finished the first draft of the RepRapMicron Microscope Pole documentation https://github.com/VikOlliver/RepRapMicron/wiki/Microscope-Pole . As usual, comments and queries are welcome.

I used the probe-wriggling levelling technique to do the "Hello World" thin with pixels spaced at 30μm. As you can see, the more accurate probe height results in much smaller impacts on the left. The slide appears to not be correctly levelled on the right and the dots smear out a bit as the probe slides under downward pressure.

You can't see half the dots using the USB microscope. This shot was taken with the bench 'scope using strong underlighting. If I level the slide you can expect more fine dots of the size that appear in the 'H'. Those seem to be <10μm across. I'll do a bit more levelling and then use the wiggle technique to estimate the height of a deposited resin dot. That'll be hard data.

I've done a little bit of upgrading of the RepRapMicron poseable microscope clamps. The thumbscrews now have a bit more clearance, and the clamps actually have the diameter printed on them so you know which is which.

 

Files are on github in the PIKA directory and on Printables. Proper writeup in the PIKA assembly instructions to follow.

With the new 4K USB microscope I can see ~5μm motion of the probe. I was mucking around with it (photo below) and had a revelation:

 

If the probe is in firm contact with the slide, and I jog the slide by 20μm in the direction corresponding to horizontally across the screen, then I can see the probe tip jerk. If I raise the probe until it no longer jerks, I am no longer in significant contact with the slide.

I bounced the probe up and down a few times, and (according to my instrumentation anyway) I can get that right to the micron.

Next, seeing what happens with Sharpie when I put that theory to the test in the real world. Probably with a slightly cleaner slide. Hey, I needed to see the surface, right?

I've been looking at a better microscope and found a 4K USB/WiFi one on Amazon AU for NZ$63 https://www.amazon.com.au/Bysameyee-3840x2160P-Microscope-Inspection-Magnification/dp/B09NBY6G9S . It just had to be better than my "Z Axis" one (less than 1MP and ancient) so I got one to try out. There goes a week's beer fund. Here's how my better microscopes stack up excluding aforesaid crappy one. The 4x4 square is 50μm per division and I have manipulated the images to give a pixel-per-pixel comparison at maximum real zoom:

On the left, the trinocular port on my Konus Crystal 45/90 with an 8MP Svbony SV205 image sensor. Quite sensibly, this does not claim a magnification factor (see previous rants on why USB microscopes lie). That can easily make out the 10μm subdivisions, and can resolve down to a micron. It is, however, a great lumbering beast and cost a few thou. 

Middle is the $63 Amazon special. It's quite nice and actually has proper focusing: When you turn focus, it actually focuses the image on the sensor plane rather than shuffling the entire imaging assembly back and forth to hit a fixed focus.

It also has a detachable transparent endcap, so one can get it closer in. 

On the right, the original "decent" Digitech microscope I bought from Jaycar. A reasonable device when I bought it for $300, but now outclassed by the 4K technology. Came with a very nice stand for bench use though.

So why isn't maximum magnification all the time the best thing? Well, two reasons:

 1. With great magnification comes lousy depth of focus, and μRepRap is making 3D structures, not flat microscope sample slides.

Note that I have experimental implants in my eyes with infinite focal depth on the macro scale. This means that when using a binocular microscope I get a radically better depth of simultaneous focus than most people, but I can't take pictures of exactly what I see, and few others can see what I do. It also means my Svbony imager focus almost always needs tweaking before I take a photo.

2. Doubling the resolution quarters the amount of available light. Adding more light only works up to a certain point, then you have a small sun on your workbench and your samples melt. 

Next step is to try the 4K out on imaging the probe as it closes in on the glass slide, and see if I can use that image to get the probe within a few microns of it. The old <1MP microscope was obviously not capable of doing that. If it works out, I'll get another. 

 The GRBL stepper setup is as follows:

CNCjs 1.10.3 [Grbl]
Connected to /dev/ttyACM0 with a baud rate of 115200
Grbl 1.1g ['$' for help]
client> $$
[MSG:'$H'|'$X' to unlock]
$0=10 (Step pulse time, microseconds)
$1=255 (Step idle delay, milliseconds)
$2=4 (Step pulse invert, mask)
$3=3 (Step direction invert, mask)
$4=0 (Invert step enable pin, boolean)
$5=0 (Invert limit pins, boolean)
$6=0 (Invert probe pin, boolean)
$10=17 (Status report options, mask)
$11=0.100 (Junction deviation, millimeters)
$12=0.200 (Arc tolerance, millimeters)
$13=0 (Report in inches, boolean)
$20=0 (Soft limits enable, boolean)
$21=1 (Hard limits enable, boolean)
$22=1 (Homing cycle enable, boolean)
$23=3 (Homing direction invert, mask)
$24=5000.000 (Homing locate feed rate, mm/min)
$25=8000.000 (Homing search seek rate, mm/min)
$26=10 (Homing switch debounce delay, milliseconds)
$27=1000.000 (Homing switch pull-off distance, millimeters)
$30=1000 (Maximum spindle speed, RPM)
$31=0 (Minimum spindle speed, RPM)
$32=0 (Laser-mode enable, boolean)
$100=31.000 (X-axis travel resolution, step/mm)
$101=33.642 (Y-axis travel resolution, step/mm)
$102=32.000 (Z-axis travel resolution, step/mm)
$110=15000.000 (X-axis maximum rate, mm/min)
$111=15000.000 (Y-axis maximum rate, mm/min)
$112=15000.000 (Z-axis maximum rate, mm/min)
$120=200.000 (X-axis acceleration, mm/sec^2)
$121=200.000 (Y-axis acceleration, mm/sec^2)
$122=900.000 (Z-axis acceleration, mm/sec^2)
$130=6500.000 (X-axis maximum travel, millimeters)
$131=6500.000 (Y-axis maximum travel, millimeters)
$132=6500.000 (Z-axis maximum travel, millimeters)
ok

 

Paul asked if NEMA14 steppers could be used in a μRepRap. Well, I guess. 80:1 mechanical advantage will do a lot if your motor hasn't got much torque. So I've split the PIKA version of the Axis Driver away from the MAUS version, tidied it a bit, and made the screw holes into slots that should take NEMA14 or NEMA17:

 

The motor shaft and boss are (I think) the same size on both, and the slots should self-centre the motor. As this is gloriously untested there are no STLs, but you can download and build from here

https://github.com/VikOlliver/RepRapMicron/blob/main/pika/pika_axis_driver.scad

Let me know how it goes. 

The guide to automated etching of RepRapMicron Probe Tips and their integration/assembly is now on the Github wiki: 

https://github.com/VikOlliver/RepRapMicron/wiki/Probe-Tip-Fabrication-And-Assembly

Comments and suggestions welcome.

 

I have now posted the results of FPath Experiment 012 which is part of the FPath project. 

Experiment 012 builds on the Graphical Stigmergy control techniques discussed in earlier experiments and embeds commands into the virtual path. These commands can automatically trigger actuators when the device under control traverses that section of the path.

Also, which may be of some interest, is a brief discussion of a new Subsumption Architecture which has been implemented in the controlling software. It has proven to be quite versatile in operation.

I just thought that some of you might be interested in how all this is done. The video explains all: https://youtu.be/7SanmrL5zJI

Labels: FPath, Graphical Stigmergy, Image Recognition, Sub-millimeter Control, Subsumption Control Architecture

If you want a preview, here's a link to the Probe making documentation:

https://github.com/VikOlliver/RepRapMicron/wiki/Probe-Tip-Fabrication-And-Assembly

I've only covered the etching part so far, but I'll write proper docs for putting the etched tip on the Probe Arm later.

As an aside, I tried using the automated probe dipping with a variety of over- and under-etch roughened wires (by time and acid strength) and they generally end up with pretty much the same shape on the last 100μm:

 

As far as I can tell at this point, the upper profile is preferred.