unusual placement offset

[Timbosaurus]

Hi,

I'm finding that the placement accuracy on the run that I am doing to be a bit off. The main difference from my last run is the PC software version, but I'm not ruling out user error at all...

What I am finding is that certain components are about 0.6mm offset. (the offset seems inconsistent from component to component, but consistent from batch to batch)

- When picking the 0603 from the tape (using normal speed placement), the image detects the center of the hole perfectly. (also, when using the alignment test on the tape setup, it centers the image on the part perfectly too).
- it does pick the components up perfectly centered on the nozzle, and I think I've tuned the camera to needle offset well.
- but then it goes and places the component on the PCB offset by about 0.6mm

So, to give real values for one particular component
- the X location on the CAD data is 42.5mm
- after re-measuring the fiducials, the X-machine location in the CAD data becomes 177.632 (which is what I expect)
- because my X nozzle nominal offset is 73.8mm, I would then expect this particular component to be placed at X = 250.839 (which is the X-machine value in CAD data of 177.632 + X nozzle offest of73.8), but when I actually run the job, the liteplacer moves to X = 251.432 before placing. So there is some other offset that I don't understand, and is not working for me.

For this reason, I dont think it is nozzle/slack/squareness compensation (which are all disabled).

FYI, I have tested picking up a component, manually jogged the machine to the expected X = 250.839 , and placing the part... and it is placed perfectly centered on the pad. So it is this small difference between the calculated placement value and the actual placement command value that I do not understand, and that is causing the innacuracy.

Regards,
Tim.

P.S I dont think it's related, but at this time I am using the "bottom" feature when loading the CAD data, adding the PCB width as a Job offset, and also using the panelise feature to do a 2x2 array of boards in one job

[JuKu]

I'd like to look deeper into this. If you can share the data (pick and place file and PDF of the copper), I'd try to replicate. Even the log of the measurement, and pickup and place of the component would help.

[Timbosaurus]

Hi Juha,

I have looked through my notes, and I see something I wrote above is misleading... In trying to troubleshoot, I reloaded the CAD data, and the numbers that I wrote above are incorrect (they are after reloading the CAD data, not from when the calcuations didn't add up)

So now that my head is not clear on what I did and in what order... so how can I expect you to try and understand!

Let me try simulate here again.

[Timbosaurus]

Now that I've reloaded and restarted, I cant get it to do it again, sorry Juha.

Apologies for wasting your time...

I have notice, however, that every time the pump comes on and off(manual control or program control), the XY move back and forth a little (0.5mm). Is this normal?

Also, I find it very easy to crash the program with an unhandled exception when loading data into the job data. The easiest way for me to reproduce is to load tape data with (eg:) a missing nozzle value. Do you see this, or is it just my PC?

[JuKu]

> that every time the pump comes on and off(manual control or program control), the XY move back and forth a little (0.5mm). Is this normal?

Yes. The g-code command is synchronized to movement, meaning that it is executed together with next move. In other words, I need to do this in order to really turn the pump on.

> Also, I find it very easy to crash the program with an unhandled exception when loading data into the job data. The easiest way for me to reproduce is to load tape data with (eg:) a missing nozzle value. Do you see this?

I'm not in the office to check, but I believe you. I'll fix this.