Printed Circuit Board Assembly & PCB Design SMT Electronics Assembly Manufacturing Forum

Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.


Fiducial Recognition

Views: 7648

#62417

Fiducial Recognition | 10 August, 2010

I'm having problems with Quad QSP-2 pick n place machines recognizing fiducials. We are cleaning the fiducials, but that's tedious and time consuming ( I don't know anyone who routinely cleans them). We have solder mask clearance around the fiducials. Any suggestions so we can stop cleaning fiducials?

reply »


JMW

#62420

Fiducial Recognition | 11 August, 2010

If the fiducials are bad. There are 2 lights on the downward camera. The one that you can see is mainly used for bright and shiny fids....the other light is for dark/dirty fids. Turn this light on and set the other for 1 (not 0 because of a software bug). Then go in and adjust your gray value numbers down till you can see the fiducial clearly and most everything around it is still dark.

reply »

#62463

Fiducial Recognition | 18 August, 2010

Why not use some PADs inside the board for X-Y index location?

reply »


JMW

#62465

Fiducial Recognition | 18 August, 2010

Pads typically have paste on them. Unless you have some parts that arnt placed and you can wipe off the paste or tape off the holes in the stencil. Not a good option.

reply »

#62472

Fiducial Recognition | 18 August, 2010

You can try using a PTH or a tooling pin hole. It's much more consistent than fiducials.

reply »

#62499

Fiducial Recognition | 21 August, 2010

I do not have a QSP-2, but build boards where others have done the layout. Some have not included fiducials, and I am left finding other things. Some boards have complete tenting over the vias, so not all PTH pads will work. Drill holes are not as accurate as the copper either. I have had to resort to pin holes on a couple of boards though when there is nothing else available. I cannot easily change the lighting on my machine, but I would try the first option suggested to see if it works for you.

reply »

#62591

Fiducial Recognition | 1 September, 2010

Drill holes or tooling holes have higher tolerances than Fiducial marks during PCB fabrication. I would certainly recommend Fiducial recognition as a more reliable approach for better accuracy. You could re-calibrate the Vision system file used for recognizing this Fiducial. I have seen similar problem on other machines, and this method helped to resolve. What is the Fiducial Dia? Should be between 1mm to 2mm to get good results.

reply »

#62601

Fiducial Recognition | 1 September, 2010

***Correction*** I stated, "1-2mm is pretty small (that's roughly 4-8mil). You might consider making your fids larger, perhaps 7-10mm (30-40mil)." My conversion was incorrect. I was brash in thinking I could do it by hand. Sanchu 70 is right, 1mm = 40mil. I still don't think in terms of metric yet (I'm too American):)

You will get more room for error based on your camera resolution. 1-2mm does not have very much surface area. Sachu 70 is correct that tooling pin holes have a higher tolerance than fids, but for lack of a workable fid, they may work. Of course fiducials are always better, but if your experience with HASL finish is anything like mine, they can be a headache at times. Retraining your system to recognize this particular fid will work, until you run the next batch and you have a fid that it can't recognize.

By the way, what I meant about tooling pin holes being more "consistent" than fids was, they are more precise as far as their appearance goes. They don't deviate as much in how they look under the camera as HASL finish fids do. I wasn't speaking about their physical location. Sorry that I did not make that more clear.

reply »

#62610

Fiducial Recognition | 1 September, 2010

Reese,

you meant to say 78mils right?

reply »

#62611

Fiducial Recognition | 1 September, 2010

I use 30mil fids too.

reply »

#62612

Fiducial Recognition | 2 September, 2010

Hi Resse, there seems to be some mistake in translating aa to mil. I thought, 1mm should have been about 40 mil. Pl. confirm.

The larger the fid dia, lesser is the accuracy during recognition.

reply »

#62620

Fiducial Recognition | 2 September, 2010

Yes, you are correct. I converted it (tryed to at least) by hand and did not carry by decimal place far enough in converting mm to mil. I double-checked my conversion and confirm 1mm = 40mil. Thanks for correcting me.

reply »

#62621

Fiducial Recognition | 2 September, 2010

You stated, "The larger the fid dia, lesser is the accuracy during recognition."

I don't necessarily disagree as far as the positioning of the fiducial in relation to the camera. In fact, I believe this to be the case and with a perfect fid, this is what you would want; however, the problem he seems to be having is with oxidation on the fiducials. I'm thinking in terms of pattern matching (grey scale coorelation). The larger the area, the more pixels you have for the inspection which will give you less error deviation from your original template. With fewer pixels, more room for deviation. So this may help in terms of oxidation which will effect the inspection of a smaller fid more than a larger one.

reply »

#62637

Fiducial Recognition | 3 September, 2010

In case of pure pattern recognition, i agree with your view. However, the basic functionality of the fiducial is determine any positional offset of the PCB and provide necessary data for correction during the intended operation such as Paste printing / Glue dispensing / Pick&Place. Hence the need for smaller distinct fiducials approx dia 1-2mm to provide greater accuracy in obtaining the offset data by the machine. Also, larger fiducial boundaries may not fit into the recognition window of the Camera.

reply »

#62640

Fiducial Recognition | 3 September, 2010

Agreed, but his fiducials aren't working so he may have to resort to something a little bigger that returns a more consistent pattern. You are correct, this will provide less accuracy than his fiducials but will save him the expense of cleaning his fids manually. I have had to resort to such tactics in the past and haven't had any major issues. The better option is to use a PTH (60-80 mil) over a tooling pin hole, if one is available. If the PTH is too small, he will have the same issues with the recognition. Not due to the size, but the surrounding plating which will most likely be effected by oxidation as well (a larger hole will give a higher ratio in the inspection area of non-metal/metal, which should raise his score). However, ultimately he needs to resolve his issue with the vendor and correct the problems he is having with his fiducials, which sounds like oxidation. But tricks like this can get him by for now.

reply »

#62644

Fiducial Recognition | 6 September, 2010

Agree .....

reply »

Lead Free Wave Solder - 1 Click SMT

PCB Rework / Repair Services