Pick and place SMT-880 (SMT-660, SMT-550) and SOT-23-3 problem

[koniu12]

I am currently trying to configure a SMT-880 machine and have a problem with camera recognition of SOT-23-3 (BC817) components (20...30% are thrown away).
The software should be the same as SMT-660 or SMT-550. Can someone tell me the settings for this? Thanks very much.

Best Regards
Adam

[48X24X48X]

Please screen shot your settings.

[PartialDischarge]

Try increasing the LEDs lightness to 300 or 350 for the lamps. If that does not solve the problem then you'll have to edit the component library dimensions for it.

[koniu12]

Thanks! The best results I have with 90..100 for the lamps. But I have tried all degrees from 10 to 500. It will be not better.
Where can I edit the component library dimensions for the SOT-23-3 (BC817) ? I found only something under feeder.
What are the additional settings there for like "FW" etc. ?

[Proto]

Although I don't have problems with SOT23-3 and use the FASTCAM and normal algorithm.  I did notice on a sw update that there is a specific camera algorithm for SOT-23 so you might try that as well - I have not.  If you have LW (length and width) checking configured too aggressively where you require a high percentage of the length and width, component picks at an angle may be foreshortened enough to be rejected to the "throw" area.  Be sure they are not bouncing so much on feeder advance that they fall out of the tape pocket or are repositioned as a result of the bouncing way out of a normal alignment.  Fundamentally, you should check what the image recognition camera is seeing and the shape it is overlaying on the part.  Seems you are saying the part does not make it to the board for placement.  Turn on the camera images during operation so you can visually see what the camera sees for each pickup and pay more attention to whatever nozzles are being used (if more than one).  Perhaps it is just one nozzle and the feeder pickup settings associated with it.

Please describe the problem you are having in more detail.

[48X24X48X]

Although I don't have problems with SOT23-3 and use the FASTCAM and normal algorithm.  I did notice on a sw update that there is a specific camera algorithm for SOT-23 so you might try that as well - I have not.  If you have LW (length and width) checking configured too aggressively where you require a high percentage of the length and width, component picks at an angle may be foreshortened enough to be rejected to the "throw" area.  Be sure they are not bouncing so much on feeder advance that they fall out of the tape pocket or are repositioned as a result of the bouncing way out of a normal alignment.  Fundamentally, you should check what the image recognition camera is seeing and the shape it is overlaying on the part.  Seems you are saying the part does not make it to the board for placement.  Turn on the camera images during operation so you can visually see what the camera sees for each pickup and pay more attention to whatever nozzles are being used (if more than one).  Perhaps it is just one nozzle and the feeder pickup settings associated with it.

Please describe the problem you are having in more detail.

The SOT-23 algorithm was already there in the v6.x software but back then I found that it weren't working reliably and instead stick to the normal rectangular which is perfect. You can enable the LxW check but just use something not so strict. On the new v8.x software, use the 2nd order close loop vision modelling only, this is what I was told by HWGC and even jelly beans parts should use that. On the out-of-the-box footprint library, you will notice all the components are assigned with the 2nd order closed loop vision modelling and no longer uses open loop which will be obsolete in the next software release. But more often than not, all this can also be attributed to the bouncing of parts on plastic carrier tape or wrong pick height which both lead to components hanging sideways (or any ways) and a failed vision recognition.

To TS, if you are not keen to give a proper information like a screenshot on your setup, I doubt anyone here is able to help you.

[Proto]

Whenever you set up a part in a specific feeder, you should test pick height by clicking the button PICK-TO-FASTCAM.  You should not move on to setting up the next part until you can do several PICK-TO-FASTCAM cycles with ZERO missing or dropped parts.  Make pick height adjustments for EACH nozzle being used with a given part feeder since there are usually several for an often used part and repeat PICK-TO-FASTCAM for every nozzle.  Send the parts to the THROW area where you should have a bin to collect parts for this task and also for those rejected by the image recognition system

While doing these trial picks, you can test how the vision algorithm "sees" the part.  If the outline the machine draws around the part pin boundary does not follow the path of the pins (evenly), this is a problem but again, it seems the parts never reach the camera so perhaps this is an unnecessary factor but file it as more setup advice.

[Styno]

Make pick height adjustments for EACH nozzle being used with a given part feeder since there are usually several for an often used part and repeat PICK-TO-FASTCAM for every nozzle.  S

I'm curious. That sounds unnecessary laborious. Does this machine not calibrate each nozzle automatically? Or not even keep a calibration record for each nozzle? An use that calibration to compensate for differences between it and the test-pick nozzle?

At startup our 25 year old Dima calibrates each nozzle length against a known fixed surface. It uses the vacuum sensor to determine when the nozzle touches that surface. After that it picks a certain component with any sensor nozzle with no (re)calibration needed.

[48X24X48X]

Make pick height adjustments for EACH nozzle being used with a given part feeder since there are usually several for an often used part and repeat PICK-TO-FASTCAM for every nozzle.  S

I'm curious. That sounds unnecessary laborious. Does this machine not calibrate each nozzle automatically? Or not even keep a calibration record for each nozzle? An use that calibration to compensate for differences between it and the test-pick nozzle?

At startup our 25 year old Dima calibrates each nozzle length against a known fixed surface. It uses the vacuum sensor to determine when the nozzle touches that surface. After that it picks a certain component with any sensor with no (re)calibration needed.

It does have. It is called feeder base height and usually a re-calibration is done rarely using a standard paper type reels. I only touch the pick-up height if the material of the tape is a lot different from the paper tape and based on my experience it is spot on unless as the above condition. I only did the re-calibration after I killed all 4 heads due to a clumsy crash.

[koniu12]

Thank you very much for your help. The SOT23 was only a test, which I hope to solve soon.
But now I try to explain my way with new project to understand, how the SMT-880 works.
1. I have installed necessary feeder in the machine and adjusted the height, tested the vision etc.
2. Than I have made a new project and imported a TXT/CSV file with fudicals, the feeder numbers I have intalled before, and nozzle type (1...7 = 504 and 8 = 505)
3. I have generated PCB array and the marks.
4. But when I start "Optimize order" then I have 2x problems :
a.) the feeder numbers will be changed (mixed up), and the adjusted settings under FEEDER before are lost (for example OFFSET)
b.) the nozzle types I set will not be accepted (after optimalization I have suddenly 1...4 heads with 504 nozzle and 5...8 = 505), AND I cant change it under "FEEDER/NOZ. PLUG IN"
I would be grateful for any help, thanks

[koniu12]

p.s. The recognition problem can be solved with the footprint set up ? But how can I set up the footprint lib ? Its on chinese (software version x7.0.1)[/img]

[48X24X48X]

Thank you very much for your help. The SOT23 was only a test, which I hope to solve soon.
But now I try to explain my way with new project to understand, how the SMT-880 works.
1. I have installed necessary feeder in the machine and adjusted the height, tested the vision etc.
2. Than I have made a new project and imported a TXT/CSV file with fudicals, the feeder numbers I have intalled before, and nozzle type (1...7 = 504 and 8 = 505)
3. I have generated PCB array and the marks.
4. But when I start "Optimize order" then I have 2x problems :
a.) the feeder numbers will be changed (mixed up), and the adjusted settings under FEEDER before are lost (for example OFFSET)
b.) the nozzle types I set will not be accepted (after optimalization I have suddenly 1...4 heads with 504 nozzle and 5...8 = 505), AND I cant change it under "FEEDER/NOZ. PLUG IN"
I would be grateful for any help, thanks

If you created a new project but want to use the same location of feeders, use the import existing project feature on the optimization dialogue. Once you have done that, when you click on optimize, choose optimize only unassigned feeder. That will keep the imported feeder settings from existing project intact. Noozle type and it's location are decided by the software as they are optimized for each project.

From your image, it looks like your threshold is wrong. Try something like 127. Also, please screenshot your settings if you want people to help you.

I doubt that the footprint library was available/completed on v7.0.1. I think most of us here uses the v8.0.x.

I hope if there's a new question, please post it on the other main HWGC thread so everything is kept centralized and users don't have to look everywhere if anyone come across the same issue.