We clean (using a pencil eraser) the fiducials on all our PCBs before processing them through a QUAD QSP-2. We're discussing pcb's with our fabricator to see if they have recommendations. Some of our Engineers believe it is the QUAD that's the problem.
It seems that we're the only company that cleans fiducials. Does anyone clean fiducials? Do you have any suggestions?
We only clean ones that are rejected (typically 1 out of 250 or more). Very well could be an issue with the boards supplier if your having a inconsistency in plating/hasl process, under a scope do the fiducials look the same between boards? More likely that your lighting or fiducial acceptance score on the machine could use some tweaking, try adjusting it to produce a better contrast since that all the vision algorithm works on.
I clean all fids with a fiberglass brush pencil so that I know there will be no rejects (it doesn't take too long). The problem is the machine lighting. My Essemtec machine uses a light ring (red LEDs) around the cam. Here's one of many documents that may help explain the differences: http://www.graftek.com/pdf/Marketing/MachineVisionLighting.pdf
(google "machine vision lighting filetype:pdf")
I think, each engineer had this situation. Try to change something - pattern of fiducial, finish cover on your PCBs (HASL to flash gold) or solder new LEDs in to your mounter.
On my Mirae's mounters we have two rings of LEDs - red LEDs and blue LEDs - and few times per year I have problems with fiducial recognition, and only if PCB made in my country. If I order PCB in Europe or in Asia, I assemble it without any problems.
When fiducials on pcba's have poor plating and the vision cant find them repeatably there are a couple of things you can try. 1. Use the WAVI light and turn off (setting of 1) the main bright LED ring on the downward camera. After doing this in teach fiducials...you need to go into extended paramaters and adjust your grey value scale down very low to get a good contrast between the fiducial and board. 2. You can also lower your acceptance tolerance. You can only do this if the fiducials are isolated from anything else the vision can find.
I have only had a few different boards over many years that the machine had a major problem with finding the fiducials. Usually the ones that had the problems had the fiducials surrounded with no color difference between the board and the fiducial.
We've had our share or troubles with this over the years. Flex arrays are the worst. If you machine allows search box sizing, make sure the x/y coordinates of the fids are perfect (CAD #'s only as always!) then make the search box as small as possible. As a temp. fix, you might have to hit the bads ones with an iron from time to time. You'll probably see if you look close that they are not flat so you're machine is reflecting a cresent moon or some other shape.
We have a Quad QSV-1 and have had some trouble with it reading 0.040" fiducials. Our work is very low volume so we just manually spot the ones it can't read. We do not clean the fiducials. Instead we increased the fiducial size to 0.060" and it has been working far better. I know it is not an overnight solution and you may not have the freedom to modify the artwork if you are a CEM. Perhaps you at least have the freedom to change the panel fiducials.
ENIAC, 0.040" equates to 1mm, not 0.1mm. 0.040" is the IPC minimum recommended fiducial size and when we started putting fiducials on our boards I assumed that all equipment manufacturers should be able to automatically spot that size. Since then I've learned to be more conservative and use 0.060" or 1.5mm.
DeanM, thanks for explanation, I will know about this "feature" of some mounters, cause I worked with Samsung, Assembleon and now work with Mirae and I never had a problem with fiducials recogntion with 1 mm diameter.
I have some suggestions. We have been fighting bad fids for years. Some things we have tried that have helped is:
1) Look at your relief area around your fid. Does it butt up against the solder mask. We allow 20 mil around the fid as a relief area. This helps with the contrast.
2) Is your fid thickness within spec? Check QC and see what the specs are. The board house might be able to put a thinner HASL finish on the fids, which will give you more consistent results under the camera.
3) How do you handle the pcbs when they come in? Are they left wrapped up, or do you unpackage them and put them on the shelf? This is OK for a little while, but the HASL finish will oxidize, which will effect your inspection score (crummy image).