RepRap: Blog

Now I have a binocular microscope. Ho-Ho-Ho. Santa was exceptionally kind to me this year (thank you very much, Suz). It's a Konus Crystal-45 with a 2x objective, so it has 90x zoom magnification - even more if I jam my old Barlow into the camera port and peek through that in monovision!

I've not got the camera attachment working properly yet. Working on it. I suspect 3D printing will happen. However, I have MacGyvered the phone onto it, and it's not half bad.

Anyhoo I decided to try a backlash test on the X & Y axes, and a Z height drag test. The hypothesis was that I would be able to see line displacement along the axis as it changed direction, and from that calculate backlash.

The method was to make clear contact with the slide using an expendable 24ga hypodermic tip, uplift, and move to (0,0,100μm). There the probe was lowered to Z=0, moved along the +X axis by 0.5mm, down -Y for 0.05mm, back on -X for 0.5mm, down on -Y for another 0.05mm and repeat until I get fed up with it.

The procedure was repeated at 90 degrees to exercise the Y axis. Following that the probe was raised 1μm at a time and stepped along -X and -Y, unfortunately not on a regular spacing (sorry). Results look like this:

As you see, colossal improvement in image resolution. Shame about the probe movement. Not at all what I was expecting, which is why we experiment, right?

Analysing that, we see a big divot at the probe start point in the top left. The downward velocity of the probe may have been a bit high. It also does not line up on the X axis with subsequent lines. As this was done manually, it may be operato,r error (Note: Write actual GCODE with slower Z descent, and some repeatability).

Also obvious is that motion on the Y axis is notably more consistent than the X axis. There is a suggestion that motion on the X and Y axis is not proportional.

There is a significant curve on the zigzags where there should be sharp corners. This persist as the Z lift manoeuvre progresses along the staircase.

My overall conclusion here is that the probe probably dug in too deep, and is being dragged around as it scratches the marker off the slide. X/Y distortion differences may be due to the geometry of the probe, which is aligned along the Y axis. As a side note, the removed material can be seen accumulating at the end of the longer movement segments.Nonetheless, the major lines are meant to be 50μm apart ("human hair"), and so positional accuracy under these condition still appears to be less than 25μm.

it would appear that processes that have high drag on the probe are not conducive to accurately measuring backlash or probe positioning capability, but that with more consistent test procedures the effect could be better analysed.

The GRBL is now being driven in 1mm = 1μm scaling, and as a separate experiment I drew the 400μm logo again. The regular distortion on the arc segments of the logo is no longer present indicating that those were indeed artefacts caused by attempting to use GRBL for very small movements. Other distortions exist, but may be due to the effects described in the first experiment.

# posted by vik-olliver @ 6:03 AM