bad component localization and calibration

[instrukcion14]

hey liteplacer's guys

I have bought liteplace about 4 moths ago.
In my case there are still one unresolved issue. I have similar problem like user "Covert". You can see his post here: http://liteplacer.com/phpBB/viewtopic.p ... travel+per
When I try to go to measured position of components there are still some offset from real position.
Covert resolved it by setting up "travel per rev."
At this moment I have a lot of boards ready for placing but i can't do that with low accuracy machine.

Can i ask you for your values in theese boxex? Juha didn't mention it in guide and I tried everything how to increase accuracy.
Or any ideas how to resolve similar problems?

Also I have one more problem with fiducial measurment. on figure i attempt to explain problem. Black dots are fiducials on board and red dots are where machine expect fiducials. It looks that expect bigger board. Or i don't know what can by wrong with that.

I'm open to your suggestions a advises.

many thanks

[JuKu]

The "travel per rev." should not matter much. The nominal value is 40mm (20 teeth, 2mm/teeth) and should be really close. I have some super accurate rulers http://www.schaedlerprecision.com/index.htm and got something like 39.96mm per rev. I guess I was measuring the belt manufacturing accuracy. To set, start with 40mm. Move the machine a known amount, say 100mm. If you find that your machine does not move exactly 100mm, but say, 99.8mm, the value to put in box is 99.8/100 * 40 = 39.92mm. But this should not matter much, everything happens in machine coordinates anyway, except the part from tape hole distance.

Here, it looks like the issues are elsewhere. First, does the machine move smoothly and repeatably? You should be able to take the machine to a mark (say, PCB zero or a bookmark), move it around and get back with high precision. If not, you need to fix this first. Look into loose screws, tight or loose V wheel adjustments, too tight or loose belts, loose pulleys etc. As a side not, if you tune Y belt tightness, loosen the pulleys first, then tighten the belts, then move the machine manually back and forth a couple of times (to even the tension on both sides of the carriage and finally, keeping the machine square, tighten the pulleys. To keep machine square when doing this, one way is to manually hold the bar against the end of movement (with a block between the carriage and the end post to avoid the switches).

If the basic movements are ok, the issue could be in calibration, data interpretation or position transform (calculation from CAD coordinates to machine coordinates). When you say " Black dots are fiducials on board and red dots are where machine expect fiducials", what do you mean by "machine expect"? Did you get the red dots by going to fiducial nominal locations or measured locations? If "measured", and you don't see anything else being off, this might be in the data or the interpretation of it. It should be able to find the fiducials, measure the locations in machine coordinates and find the transform between CAD and machine coordinates. That transform should also handle scaling issues. Like, in your image it looks like the board is smaller than the machine expects. But even if that is the case (a cross manufacturing error), the measurement should handle it.

Does this happen with all boards or just this one?

[instrukcion14]

hey dudes

update from my todays calibration

there is no offset as i posted on figure. that's perfect
and about the bad positioning on mesured location: i disable squarness correction and results are incredible. don't know why but im totally satisfied.

Juha can i ask you for your comment on disabling squarness and its influence on accuracy?
I know that correction works good but in my case give bad results. how program determine position of component with enabled correction?

[JuKu]

> how program determine position of component with enabled correction?

The squareness correction is deep in the system, and affects all movements. All the movements use machine XY coordinates. All the squareness connection does is applying a small adjustment to X in terms of Y so that the machine XY better corresponds to a square machine in real world. In practice, this should have little or no effect, as everything happens in machine coordinates anyway, and mostly everything (like part positions, board location etc) is measured, without assuming that things are really aligned to the machine.

With or without squareness correction, the components are found by the holes or using the location directly, and the correction does not really matter. Your components are not precisely aligned to the machine anyway. If you use locations directly, the result is that from the machine's point of view, the tape is a bit more or less skewed. When the components are found by the holes, the skewed placement of the tape is also automatically compensated when the holes are measured. When a part is picked up, the pick up process does not take the tape misplacement into account anymore, but assumes that it is straight enough that a part can be picked up from (say) 2mm down, 3.5mm left from the hole. (Resulting errors from this, with or without squareness correction, are in the small 0.0x mm range, well below the machine resolution).

I don't really have an explanation to your experience, other than maybe the squareness factor wasn't somehow properly taken into account. (I went to the code, and found that I have work to do in the internationalization and sensible checking of values). Maybe you used comma or missed a zero or two when putting in the value?