45 Degree angle mirror for upward facing camera

[thereza]

I just want to stress Juha's point here - needle wobble calibration HAS to be done at the same Z-level as PCB surface

Technically, it needs to be done at pcb height + component height if you want to get super accurate. measuring component offset when base is at the pcb height over the upcamera would be an alternative approach.

[mrandt]

Technically, it needs to be done at pcb height + component height if you want to get super accurate.

As often, you are right Reza

I would suggest the following comrpomise:

1. Small components with few pins (passives, SOT23, etc.) can usually be placed accurately enough without further component measuring using upcam. So these can take standard needle wobble calibration offsets measured at PCB level and placement will be accurate enough, also due to self alignment during reflow.

2. Tall components (elcos, USB connectors, etc.) and those with many pins (mostly ICs in different packages) should be aligned using component measuring with upcam.

If we used "center with multiple movements" approach for up cam, perspective error will not be the issue. However, for very tall components focus might be. So ideally, we would measure component height as suggested in the other thread and "hover" the components at PCB level - so needle tip would be at (0 + PCB thickness + component height).

[Covert]

Isn't needle wobble the same regardless of the height it is taken from ? Looking at the machine I can't see how Z axis height would change needle wobble.

[mrandt]

If you look at the needle only, you are correct - if you rotate it, the wobble will be consistent at different Z-levels.

I think there is yet another effect, which is the moving parts on the Z-axis itself. I would assume that the vertical maker slide will never be perfectly orthogonal to the table surface, so this will affects needle tip offset as well and is dependent on Z-level.

Your biggest problem however will be perspective error and projection size. If the needle is closer to the camera, it will be projected larger on the camera CMOS - to use an arbritrary example, 1mm offset in reality might be 20 pixels offset in camera picture at PCB surface level.

Once you move the needle farther away, the projection becomes smaller. 1mm offset might only be 15 pixels in camera picture.

So it will be important to calibrate the mm to pixel ratio used for measurements at the same level at which you will later measure needle wobble.

This is currently achieved by using PCB surface level (as calibrated during needle height calibration) - which is not 100% accurate as Reeza pointed out before - but good enough for passives. For taller and more complex components, we are discussing enhancements in separate threads.

Not saying it can't be done but it might complicate things. And you might still suffer the Z-axis offset as mentioned above.

[JuKu]

> Technically, it needs to be done at pcb height + component height if you want to get super accurate.

It does that, but currently uses a fixed value of 1mm for component height.

[thereza]

Isn't needle wobble the same regardless of the height it is taken from ? Looking at the machine I can't see how Z axis height would change needle wobble.

good question. i can see a different offset but the wobble should be the same -- thoughts?

[JuKu]

The farther you look at it, the smaller it looks. On the other hand, that's calculable.

[thereza]

see http://liteplacer.com/phpBB/viewtopic.p ... rt=10#p481 for a height dependent measurement of wobble.