Specifying position and orientation of a thin part on a PCB

[bean777]

http://files.engineering.com/getfil...n_and_orientation_of_a_thin_part_on_a_PCB.pdf

I had a PCBA with two Fiducials and an image sensor (Fiducials are exposed copper pads on the PCB typically about 1mm in diameter. They are used as Datums or reference points for locating other copper traces or pads on a PCB.) The Sensor was wire bonded to the PCB (see Figure 1).

Figure 1 (See attachment or follow this link:

http://files.engineering.com/getfil..._and_orientation_of_a_thin_part_on_a_PCB.pdf)

For the purpose of system level of Tolerance Analysis, we needed to specify the position and orientation tolerances of the sensor to the two Fiducials. Just to be clear, we weren’t try to control the leveling of the sensor to the PCB surface. So the drawing looked like Figure 1 below. Basically the PCBA was treated as a 2D part. To some of us there was no ambiguity in term of how the part should be constrained and inspected.

Figure 2 (See attachment or follow this link:

http://files.engineering.com/getfil..._and_orientation_of_a_thin_part_on_a_PCB.pdf)

 
Then a coworker insisted that we must define all three Datums to completely constrain the part for inspection. So the new drawing looked like the one in Figure 3. When our supplier got the new drawing (as in Figure 2), they complained that the new drawing required them to measure the position and orientation to Datum A as well, which was not only difficult to do because the height of the sensor was only 1mm but also not part of the design intent. I’m writing to this forum to solicit some expert opinions. How would you specify the position and orientation of the sensor in a drawing?

Figure 3 (See attachment or follow this link:

http://files.engineering.com/getfil..._and_orientation_of_a_thin_part_on_a_PCB.pdf)

Thanks in advance.

 

[MikeHalloran]

If the coworker who insisted on 3D tolerancing was the cognizant optical engineer, then (s)he was just doing the job. The location of the image sensor surface in 3D space is presumably important to the operation of the assembly.

If the coworker was just the usual random idiot, well, you wouldn't be here, right?

[A] is not a real good datum for measuring the edge angles, and probably shouldn't be referenced in those callouts.

[A] may be read to refer to a faying surface on a gage, which is also the collective high points of the PCB. Since they are sometimes curled, it might be better to use a third fiducial on the PCB surface as [A].

If you are _not_ trying to control the planar orientation of the sensor, e.g. variations in solder or epoxy bond thickness, then the true position callout on the sensor center is unnecessarily restrictive. The reference to [A] should be removed from that, or true position wrt [A] might be given a looser tolerance, like .5 mm.


You probably need to call an informal meeting with your vendor and your optics people, and do some whiteboard sketches until you can all agree on what is acceptable and manufacturable. Do take photos of the whiteboard. A video of the whiteboard and the meeting would be better.



Mike Halloran
Pembroke Pines, FL, USA

 

[3DDave]

Your supplier is incorrect. The tolerance zone is the same whether datum A is referenced or not. Adding A simply requires that the part be oriented to A before locating B and orienting to C. Nothing in the drawing locates the part normal to datum A.

The obviousness for the need for using datum A is this - do they lay the part down on a table before they fix a simulated datum reference frame to the fiducials? That table is the primary datum reference simulator and should not be assumed.

The fiducials are not capable of eliminating the last three degree of freedom, two rotations and one location.

The fiducials are two points which can be viewed from any direction. If they are the only references, one could have the board tilted in any direction necessary to meet the location requirements. Adding datum A controls the orientation (rotation about X, rotation about Y, unused Z movement,) fiducial A removes two location degrees of freedom (X movement, Y movement), and then fiducial B removes the last rotation (Rotation about Z).

One assumption that remains is the slope of the chip face relative to the board. Presuming this is a BGA or similar component, soldering will cause the part to lay flat on the board within some small variation. This is a system level constraint that should be part of the tolerance analysis even though it is not explicit on the drawing.