LitePlacer.com forums

Hi Adam,

WayOutWest wrote:Wow, Karl, that nozzle changer looks AWESOME.

I can only second that

AFAIK, Karl is using Juki Nozzles and custom machined holder from China:

Nozzles:
http://www.ebay.com/itm/SMT-JUKI2050-20 ... 1783968281

Holder:
http://www.ebay.com/itm/Customized-SMT- ... 1826225528

The holder has a steel ball lock (similar to air compressor connectors) - thus the changer needs to have a angled surface to release the nozzle I suppose...

Options for new nozzles are also being discussed here:
http://liteplacer.com/phpBB/viewtopic.php?f=11&t=192

Some inspiration also came from OpenPNP google group discussion.

WayOutWest wrote:How do you do z-min (i.e. hitting the table or PCB) detection? The original liteplacer depends on a "stiff" pickup rod to trigger the limit switch, but the Juki nozzles seem to have a built-in spring that would absorb the pressure. Do they have some kind of sensor switch in there?

The Juki nozzles have a pin inside that moves up if nozzle touches something.

Karl described his solution before - he is using a electric contact inside the nozzle holder in a true hacking style

Check it out here:
http://liteplacer.com/phpBB/viewtopic.p ... t=10#p1292

Nozzle tip needs to be grounded. A slip ring (or electrically conductive ball bearing) could be used instead of steel wool to make reliable contact to the rotating contact pin - but then it should work fine.

Another option would be to put a longer plastic or metal pin inside the nozzle holder and us a piece of transparent tubing somewhere upstream. We could then use optical sensor to detect that pin's appearance / movement along Z-axis when nozzle spring is compressed.

A third option was introduced by jarekk - he uses 3D printed discs and a capacitive sensor:
http://liteplacer.com/phpBB/viewtopic.p ... =192#p1324

Personally, I will probably try the Juki nozzles and coupler with mechanical / optical detection. At the same time, I will change my LitePlacer's head design to use a hollow shaft stepper and get rid of tube + pulley + belt.

Regards
Malte

For some reason i stopped getting notifications about updates in these forums so i was under the impression nobody was talking, anyway i guess ill start looking at this manually to see what's going on.

I'll look into the bugs reported at github and see what i can do, for now i'm currently trying to finalize my feeder design and once that's done i will dive back into the code as it being stable should be the only thing left for me to do to have a fully working pick-and-place with feeders and automatic nozzle change

Ill make a video showing how to do use all the new automatic calibrations, to me these routines improved the accuracy of my machine tremendously.

I still want to figure out a good way of doing the camera-to-nozzle measurements automatically, so far the only thing i've been able to come up with is using a small touchscreen which projects a circle detectable with the camera and then move the nozzle over the touch screen until it hits the dot. This seems awfully complicated though, there has to be a better way...

I put some blue ink on the nozzle tip -> go down to table and hit a piece of paper -> move camera to the location and search for center of blue circle.

Hi Karl, thanks again for sharing your ideas... I ordered the nozzle-holder you found, can't wait for it to arrive.

Knas wrote: I still want to figure out a good way of doing the camera-to-nozzle measurements automatically, so far the only thing i've been able to come up with is using a small touchscreen which projects a circle detectable with the camera and then move the nozzle over the touch screen until it hits the dot. This seems awfully complicated though, there has to be a better way...

I mounted a second upward-facing "calibration" camera (they're cheap!) for this purpose.
I have a piece of glass over the top of both upward-facing cameras (the bga chips are resting on it in the photos), with a marker above the center of the calibration-camera's field of view. It's very easy to use the calibration-camera to position the needle-tip (shiny circle) over the marker (shiny dot), and then use the downward-camera to position its own field-of-view over the same dot. The dot is a permanent part of the installation which is why I can't use a single camera for both calibration and alignment (but... cameras are cheap!).

The marker needs to be on the top surface of the glass and the needle needs to be (almost) touching the glass during the first step -- otherwise any difference between the angle of the calibration camera and the downward camera will show up in the calibration results. I originally tried having the downcam "look for" the upcam's aperture-ring, but that was too sensitive to mounting alignment. Using a marker that the needle physically touches eliminates this, but requires a second camera since you don't want the marker in the way when rotating components. Since the second camera is used only for this purpose and nothing else it doesn't need to be centered with respect to the illumination ring.

The glass is about the same thickness as a PCB, so the focal depth is correct from both above and below. There are a lot of accuracy benefits to being able to (gently) press a component against the glass. I was using acrylic before, but it seems to develop microscopic scratches, even the supposedly "un-scratchable" stuff.

Spikee wrote:I put some blue ink on the nozzle tip -> go down to table and hit a piece of paper -> move camera to the location and search for center of blue circle.

I was doing this before (using blue putty actually), and it worked reasonably well, but a human still has to remove the old marking before each calibration.

Well is a second upcam really necessary?

How often do you calibrate camera-to-nozzle ? Why not exchange the clean glas with the dot glas for camera-to-nozzle calibration and then replace a clean glas.

The absolute position of the dot is IMO not important as long as it does not move between the two measurements.

I would make a black circle dot which could be measured like the optical homing.

Although cameras are cheap, don't forget the USB count and the extra processing load and the extra software to switch cameras.

mawa wrote:Well is a second upcam really necessary?

How often do you calibrate camera-to-nozzle ? Why not exchange the clean glas with the dot glas for camera-to-nozzle calibration and then replace a clean glas.

I think it depends, the camera-to-nozzle measurement is an approximation of median needle wobble and camera position, since you remeasure needle wobble every time you turn on the machine it makes a difference. Also because the needle wobble measurement isn't perfect, it sometimes doesn't do the best job and you have to do it again. In my thinking, the needle-to-camera position should optimally be remeasured every time needle wobble is remeasured.

mawa wrote:How often do you calibrate camera-to-nozzle ?

My goal is to have automatic nozzle-changing working. But you need to recalibrate the nozzle-to-camera every time you swap out the nozzle, so it can't be truly automatic unless the calibration is automatic too. Plus I like having all the calibrations be fully-automatic so I'm never tempted to get lazy and skip doing it when I should.

mawa wrote:The absolute position of the dot is IMO not important as long as it does not move between the two measurements.

Well, the distortion gets pretty bad as you get farther from the center of the camera's field of view, which means multiple cycles of moving the needle before it "hits" the mark.

However today I switched from using a dot to using a pair of QR barcodes printed on translucent film; the "target" is the point halfway between the barcodes' centers. It turns out that recognizing QR codes is much much much much much much much much much much much much faster and more robust than pretty much any other vision task. So the target is still (almost) at the center of the camera's field of view, but doesn't get in the way of whatever the camera is looking at. So:

mawa wrote:Well is a second upcam really necessary?

With the QRcode-based targeting, maybe not.

First screenshot is the calibration camera. Second screenshot is using the downward camera, which is much higher quality+resolution. The red crosshairs are the mouse, not the target. Upward-pointing camera images are mirrored horizontally.