The nozzles are far better than the needles currently used. And they can be automatically changed, of course; that's the point:). There is almost no wobble*. Also, there are new pickup tube and better bearings. These should fix the issues some of you have had with a not so smooth rotation. And finally, the new tube and nozzle adapter is significantly heavier than the old parts. Therefore, I added another spring and associated hardware to allow tuning of the placement force both to heavier and lighter direction. I ended up using neither, but if you want to tune it one way or another, you can now do both.
*: In my proto, I had no measurable wobble first. During the development I crashed the machine to the holder (hard) a couple of times. I now have about 0.1mm wobble and some nasty sounds on my X axis. I haven't yet figured out what got bent and if I will bother to rebuild it.
Action video:
Pardon for the wrong length screws, I just used what I had.
Development status:
Hardware: The prototype holder works fine, and the production model will be better. Changes will be: Right length screws, 1.5mm stainless steel instead of 2mm (easier Z setup) and some minor dimensional trims. There will be no alignment marks that you see on the video; setup using nozzle mouths is easier and more accurate.
Software: The setup page is done and the change routines work. Yet to be done is to add nozzle info at tapes setup and allow placement order sorting by nozzles.
Estimated availability by the end of September, hopefully sooner.
While doing the software for this, I made some experiments. during development, I have crashed the machine hard, several times. Hopefully the setup process and the software is now such, that you don't have to! I think I'm able to start shipping early next week. There might be a short period where shipping the kits, finalizing the instructions and releasing the software are out of sync. My apologies beforehand if you find yourself with a kit but no software or with the kit, software but inadequate instructions. All pieces are coming, but they will not all come at once, there is only myself working on this.
The parts are here. I haven't figured out the packing yet. I have a second release candidate for the software. It looks good, but it is not fully tested yet.
Performance:
Summary: I had a period where I didn't have any measurable wobble. After a few crashes, I now have some. This does not really matter, as the repeatability of the nozzles is good; so good, that I made the software able to store the calibration and use stored calibration values.
Long version: The software can calibrate and store the values for each nozzle. I calibrated all nozzles in turn and stored the values. I then repeated the five times. I put all numbers in Excel and looked maximum and minimum values for X and Y in several directions. The results are (all numbers in mm's):
Wobble for individual nozzle (can be calibrated out): 0.09 to 0.16
Differences between nozzles (can be calibrated out): 0.18 to 0.20
Repeatability, how much difference there is for a nozzle calibration values when it is loaded, unloaded and loaded again (*): 0.03 to 0.046, smallest nozzle 0.055 to 0.08.
*: The last line is worth a few notes. First, this is the most important measure, as this cannot be calibrated out unless the software is set so, that the nozzle is re-measured after each change.
Second, looking at the individual numbers, I think that the measurement process has a significant influence here. You would think that any wobble creates a circle or an ellipse, but not so; the deviations look more like noise. The numbers are small, about one pixel on my setup. I have improved the needle/nozzle calibration measurement routine already, but there might be room for more. If you think that the repeatability is good enough, the nozzle calibration needs only be done once. Some extra measurement time doesn't matter then (besides, the process is not only ore accurate now, it is also faster). The nozzle is currently only measured once. Maybe doing several measurements and taking either the median or average gives us the real performance number. The median throws away noisy measurements. If the sequence is 2 2 2 4 1 2 2 2 2, 2 is the true answer and the median is correct. On the other hand, if the sequence is 1 2 1 2 1 2 1, the median would be 2, while the true answer is 1.5, the average. Clearly, this needs some more work and experiments, but it is pretty good already; maybe good enough for first release.
And last, I need to take a deeper look at the smallest nozzle. This is where the accuracy matters most, but the calibration errors are largest. Again, as of today I'm unsure if this is true variation or a measurement artifact. The nozzle tip is only a few pixels wide and there is some noise on the video, so maybe the computer has issues finding the center reliably? Again, it will be interesting to see if repeating each measurement and taking median or average changes things. It certainly looks to be in the same position each time, when just looking at it.
But for now, it is very usable and I must not get distracted to getting the final 0.02mm performance out of it right now - there is better time for that next month, when the kits are out.
