Maybe this procedure will help?
Or maybe this is due to too large pockets in the tapes.
Try to place many parts with different rotations.
My build
Re: My build
The offset is consistent, tape or loose part placement. Will do the makerslide loosening procedure again. Nothing wrong on the Y axis.
Re: My build
Couldn’t quite make out your pictures when I click on em, try rescaling to reduce size before posting as phpBB does weird things in different browsers with large photos.
As Y is working and X isn’t, I suspect your nozzle / camera aren’t orthogonal to the table or parallel to each other and it’s
worse in X.
It’s fairly easy to rule out, if it isn’t that post back for some more things to check.
So the issue is that you have very tight tolerances to get everything square and parallel, it only has to be slightly out
to show up at the small tolerances we need for accurate placement, and without fine tuning ability (which we don’t
have), near impossible to tune out.
But the good news (essentially the workaround to the problem) is everything is the same at a given Z height. I.e. I think that maybe your component pickup height and placement height, or the height at which you tuned are different. Everything must match exactly, so:
Top of the PCB, rulers when measuring and doing alignment/setup, top of the component, must all be at the same height,
otherwise any offset transfers. So take the following situations as an illustration as to why this is so important.
1. Top of PCB is Z=35. Top of Components for pickup is Z=33.5 and Nozzle / Camera parallel differs By 0.5 degrees (camera height above surface say is 4cm), then that is an error of 0.015mm on pickup. That will be present as a constant offset in X, exactly as you describe.
2. Top ruler is Z=36.6, top of PCB is 35, and you calibrated based on this, now your camera pixel to mm ratio is off, so
everytime you measure a ficducial to check accuracy, if it’s not dead center you’ll have a consistent offset, this will vary according to how far the fiducial is from center.
And obviously, you have to have a really small fiducial to check any of this, and your vision has to be rock solid on that fiducial.
There are other possibilities here, but check the above first....
For example, the way I have mine setup is on laser cut acrylic jigs, so that I can do double sided and single pcbs (with
everything at the same height), and jigs under the tapes for pickup to put the top of those at the same height of the top of the pcb.
Because component height like SOT23, QFP’s, connectors etc. can vary, then I could make sure those are at the same height, but as I use bottom vision, it’s less critical because the component is larger, provided the bottom of the component is always same as the top of the PCB height. I.E. I tune for 0402 table pickup with the smallest nozzle, as it’s the most critical positioning wise, but with larger components I don’t care as much as bottom vision fixes it within limits.
However if you’re using the Liteplacer software, you don’t have bottom vision, so in that case top of all components MUST be at the same level if you want accurate placement.
As Y is working and X isn’t, I suspect your nozzle / camera aren’t orthogonal to the table or parallel to each other and it’s
worse in X.
It’s fairly easy to rule out, if it isn’t that post back for some more things to check.
So the issue is that you have very tight tolerances to get everything square and parallel, it only has to be slightly out
to show up at the small tolerances we need for accurate placement, and without fine tuning ability (which we don’t
have), near impossible to tune out.
But the good news (essentially the workaround to the problem) is everything is the same at a given Z height. I.e. I think that maybe your component pickup height and placement height, or the height at which you tuned are different. Everything must match exactly, so:
Top of the PCB, rulers when measuring and doing alignment/setup, top of the component, must all be at the same height,
otherwise any offset transfers. So take the following situations as an illustration as to why this is so important.
1. Top of PCB is Z=35. Top of Components for pickup is Z=33.5 and Nozzle / Camera parallel differs By 0.5 degrees (camera height above surface say is 4cm), then that is an error of 0.015mm on pickup. That will be present as a constant offset in X, exactly as you describe.
2. Top ruler is Z=36.6, top of PCB is 35, and you calibrated based on this, now your camera pixel to mm ratio is off, so
everytime you measure a ficducial to check accuracy, if it’s not dead center you’ll have a consistent offset, this will vary according to how far the fiducial is from center.
And obviously, you have to have a really small fiducial to check any of this, and your vision has to be rock solid on that fiducial.
There are other possibilities here, but check the above first....
For example, the way I have mine setup is on laser cut acrylic jigs, so that I can do double sided and single pcbs (with
everything at the same height), and jigs under the tapes for pickup to put the top of those at the same height of the top of the pcb.
Because component height like SOT23, QFP’s, connectors etc. can vary, then I could make sure those are at the same height, but as I use bottom vision, it’s less critical because the component is larger, provided the bottom of the component is always same as the top of the PCB height. I.E. I tune for 0402 table pickup with the smallest nozzle, as it’s the most critical positioning wise, but with larger components I don’t care as much as bottom vision fixes it within limits.
However if you’re using the Liteplacer software, you don’t have bottom vision, so in that case top of all components MUST be at the same level if you want accurate placement.
Re: My build
Did all the above and rechecked it with draw box and jogging the machine. Everything is spot on.
When I use the probe down function the nozzle lands exactly on all 4 fiducials, but probing down on a 603 position it allways lands halfway on the rightside pad. 603 pick up is spot on in the middle.
All components have the same shift like this ( Tape, Loose part):
When I use the probe down function the nozzle lands exactly on all 4 fiducials, but probing down on a 603 position it allways lands halfway on the rightside pad. 603 pick up is spot on in the middle.
All components have the same shift like this ( Tape, Loose part):
- 603.GIF (27.92 KiB) Viewed 5826 times
Re: My build
Do you have a download link?Jet wrote:
1. If you wish to stay with Liteplacer, I have a compiled binary you can have which allows you to change the amount of
Backlash Compensation, it’s a drop in replacement for your .exe
Re: My build
Found the problem!
When re-adjusting the camera I discovered the lenshousing is mounted incorrectly on the pcb and the lens focus ring is very loose. This is what it looks likes when adjusted to have the cross in the lens center:
When tap against the lensring de cross shifts in the picture....
When re-adjusting the camera I discovered the lenshousing is mounted incorrectly on the pcb and the lens focus ring is very loose. This is what it looks likes when adjusted to have the cross in the lens center:
- 1down.GIF (267.01 KiB) Viewed 5820 times
Re: My build
Took the camera board out, the housing is mounted correct on the pcb but there is a lot of sideway play between the lens and housing, so the angle changes constantly. The up looking camera has no play.
Re: My build
Please email me your delivery address, I'll replace the camera. I'll respond to other items on this thread when I get better (I've been somewhat sick since start of December, and I haven't been able to catch up on everything that I should).