My building Experiance.

[Covert]

Well today I hopped to load 4 pannels. I got 1 done. It was smooth at the start with all the white paper tape 0603 parts going in without any issue. The Black plastic tape started to cuase some issues and some strange crash/lockup on the TinyG. Turns out the lookup was a Slack Compisation. Without access to the source code I would still be running around in circles. Took me a long time to find the problem but once I worked it out and made a few simple changes to the source it started to load the next lot of tapes. Then trouble again with my different tape sizes and rmod not supporting the. A few more changes in source and some of my larger tapes are loading.

Tomorrow I hope to get one smooth run in. Already seeing a few changes I want to do.

//Notes for myself of changes I made.
Slack Compensation A axis.
New tape types
Check disconnected before placing more parts.

[ravng]

The FET for the vacuum pump is stuck close. It's not fully closed so the pump just idles away then "rev's" up when the fet gets powered. I will replace it soon but for now it still works as needed.

I have the same issue with both the vacuum pump and the solenoid.
When I measure the voltage I get 8 Volts when supposed to be off, and 24 Volts when on.
Did you figure this out?

Not an issue with the pump, but when the solenoid is activated, 8 volts seems to be enough to keep it engaged, so it never cuts off the vacum.

[JuKu]

The FET for the vacuum pump is stuck close. It's not fully closed so the pump just idles away then "rev's" up when the fet gets powered. I will replace it soon but for now it still works as needed.

I have the same issue with both the vacuum pump and the solenoid.
When I measure the voltage I get 8 Volts when supposed to be off, and 24 Volts when on.
Did you figure this out?

Not an issue with the pump, but when the solenoid is activated, 8 volts seems to be enough to keep it engaged, so it never cuts off the vacum.

Your FETs are broken (I assume that you have checked the connections and that the TinyG does drive the gate from 0 to 3.3V). It could be static, or they could be faulty start with. You might get replacements locally faster. Anyway, please get in touch with me for replacements or refunds.

[Covert]

Your FETs are broken (I assume that you have checked the connections and that the TinyG does drive the gate from 0 to 3.3V). It could be static, or they could be faulty start with. You might get replacements locally faster. Anyway, please get in touch with me for replacements or refunds.

I addressed the issue shortly after. It was a dead FET. I replaced it with a P55NF06L. I have a tube of these from unbuilt RAMPs boards. Been working ever since.

Today while running a small job I had trouble with parts staying on the end of the pic and place pickup. The soldier paste was sticky enough most the time and I got the job done but later I diagnose the problem to the Solenoid FET being dead. Same type of death, stuck partly on.


I addressed the noisy air pump. Printed a clamp and nested the pump into it with foam. I really reduces the background noise from the machine.

[Covert]

My machine is getting used every second day now. I'm having to baby sit it less and less. I solved an issue with calibration using a metal square from the hardware store.

I edited the source code to take a calibrated needed into the down camera -> needle calibration. I have written a custom down camera / needle calibration. It goes like this.

I have a location on the machine called "BlueTac". I have a little blob of blue-tac to detect the location of the needle.

-------------------
Rub my finger over the BlueTac to remove last needle mark.
Calibration needle on down camera.
Move to BlueTac.
Needle to PCB
Needle up.
Move down camera over blue-tac location the current down camera offset.
Select a 8x zoom down camera filter.
I click on the down camera image to align the crosshair into the center of the mark made in the blue-tac.
I click "Done"
-------------------
I find this simple and effect to get the needle-camera very closely calibrated.

I printed a new up camera mount simply because my last one had play.


I address the down camera light issue. I found with black tape I could not get a clear and reliable image to detect holes. After playing with a few LED's and even a 3 WATT cree I soon found out the 12V output on the TinyG was not up to multiple LED lights. I had a 12V-3A reg around and run that of 24 Volts from the PSU.

I ended up using a "24 LED 5050 SMD 12V White Car Interior Light Panel". It's what I had. I 3D printed a mount that bring the light out towards the front of the machine. 100mm away from the camera and 60mm of the table pointing towards where the camera looks on the table.I find this gets the light underneath the edge of the black tape and helps light up the white double sided tape I use on the table.

Next I had to deal with the different reflection with the different black tapes I have. I currently use 9 different black tapes. I made a spreed sheet and entered the different threshold ranges I had to use to get a circle. the lowest was 50-110 and the highest was 90-190. Decided to add 4 new filters so I have a total of 5 black tape filters (more than is needed I think but I want a reliable machine). On the spread sheet I decided on these 5 threshold levels. 85,90,100,130,145. I added the BP-2,BP-1,BP+1,BP+2 filter profiles to the source code and set them up in the software. I then tested the machine with these new filters assigned to different tapes and the machine dry ran placed all 9 tapes with over 200 parts. I think this will ran well for me.

I added a new tape calibration routine. I have a new field in the tapes for TapeLength. I then can look for a start hole, a hole mid way along the tape and one at the end. This is quick and more accurate in getting the angle of the tape, something I found was a issue with long tapes, towards the end of the tape the machine would seem to get lost because the original measured angle was different to the real angle. Ideally I would like the machine to re-calibrate the tape angle as it finds new holes while running, I think this would be a even better option. Using the last hole used to find a part as a new calculation for the tape angle.

I got a big red stop switch. I hopped to use it on the reset line. The switch is normally closed. I'll need to wire up a transistor to invert the signal for it to work. Another day..

I'm still cant get clear tapes to work well. I'll try it out with the new light soon...

[WayOutWest]

Rub my finger over the BlueTac to remove last needle mark.

Neat! I've been thinking of using a cheap trackpad with a "crosshairs" printed on it to do something similar.

I ended up using a "24 LED 5050 SMD 12V White Car Interior Light Panel". It's what I had. I 3D printed a mount that bring the light out towards the front of the machine. 100mm away from the camera and 60mm of the table pointing towards where the camera looks on the table.I find this gets the light underneath the edge of the black tape and helps light up the white double sided tape I use on the table.

Sounds interesting, but I can't quite visualize it. Could you post a photo?

Next I had to deal with the different reflection with the different black tapes I have.

I've been thinking a lot about reflection issues, because my PCB provider's soldermask seems to be unusually reflective.

Once the camera is calibrated, it ought to be possible to:

1. Snap a photo
2. Move the head a tiny amount
3. Snap a second photo
4. Subtract (step3 photo shifted a tiny amount) from (step1 photo)

The idea here is that because the reflective glare is coming from the camera head it doesn't move by the same amount as the rest of the visual field when the head moves. So it should be possible to remove the glare with relatively stupid vision processing by exploiting this fact. But doing so many extra moves and snapshots may reduce throughput.

[Covert]

Here are some pictures of the Automotive light panel. One with the ring light on and the other with the ring light turned off.