Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.

SMTnet
»
Electronics Forum
»
Fiducial recognition using laser

Fiducial recognition using laser

Subscribe to Thread

Many years ago I remember the Philips CSM series SMT machine... - Apr 01, 2004 by Mike

Are you sure it was measuring a fid? Is it possible it was ... - Apr 01, 2004 by

Yup. It was definitely checking the fiducials as well as the... - Apr 02, 2004 by

Why not use tooling pins (locating pins)? Sounds like the s... - Apr 02, 2004 by Kuru

Ken, You'd think so but the tooling pins only register to... - Apr 02, 2004 by Mike

I have used tooling holes and fiucial registration since the... - Apr 03, 2004 by

Ken, Thanks for the feedback. Unfortunately (or fortunate... - Apr 12, 2004 by Mike

Mike

#27864

Fiducial recognition using laser | 1 April, 2004

Many years ago I remember the Philips CSM series SMT machines used a small tubular laser system for identifying fiducials on the PCB. Can anyone suggest suppliers for these devices.

Thanks.

reply »

Ken

#27870

Fiducial recognition using laser | 1 April, 2004

Are you sure it was measuring a fid? Is it possible it was measuring a block-skip mark?

reply »

cedams

#27882

Fiducial recognition using laser | 2 April, 2004

Yup. It was definitely checking the fiducials as well as the block skip marks and it used to work pretty well.

What it did was started at the center of where the fiducial should be and scanned left. When it found a change in contrast it scanned right. When it fell off the fiducial on the right edge it took the average to get the fiducial center in X. It did the same for Y to get the overall center.

The disadvantage is that it really only works on round or square fiducials and if the fiducial is badly etched you will get offset errors. However it's cheap and for my needs more than adequate.

I have found a company http://www.keyence.com that do fiberoptic reflective sensors that should do what I need at a very reasonable price <$500 and since I only need the fiberoptic head on my dispense head I can keep the weight down, have the sensor closer to the dispenser, etc, etc.

Regards, Mike.

reply »

Kuru

#27894

Fiducial recognition using laser | 2 April, 2004

Why not use tooling pins (locating pins)? Sounds like the same accuracy...

reply »

Mike

#27896

Fiducial recognition using laser | 2 April, 2004

Ken,

You'd think so but the tooling pins only register to the drilling process for the PCB. The fiducial registers to the copper pattern and for both glueing and placing SMT components the pads are more important than the holes. That's why the machine manufacturers give you fiducial recognition on SMT machines and tooling hole registration on axial and radial machines.

The sensor I found has a beam of 0.1mm (0.004 inch) so I should be able to get within 2 mils of the center of the fiducial. Good enough for my needs and less than 500 bucks. I know a camera system is going to cost me a lot more than that to give me a little more accuracy that I don't need...

Mike.

reply »

Stefan Witte

#27904

Fiducial recognition using laser | 3 April, 2004

We had a similar system on the Siemens machines, quite some time ago. However, it was not a laser light but a LED and a fiber cable. Actually the center fiber or receiver was surrounded by a bundle of small fibers for the LED light source. You may also look at Omron for a supplier. The system had to detect four edges of each fiducial to find the center, which took about 3-4 seconds per fiducial. Because the sensing distance is so small (about 1 millimeter), the sensor was lowered upon recognition and was retracted to avoid crash with feeders, components and conveyors. Because this system had a very short life time on the market, it would be quite difficult at least for the Siemens machine to locate the appropriate software, which supports this type of recognition. I do assume this will also apply to your Philips machine

reply »

Ken

#27905

Fiducial recognition using laser | 3 April, 2004

I have used tooling holes and fiucial registration since the late '80's. In fact if your maximum "vision" tolerance is +/- 0.05mm, your in the same ball-park as locating pins.

It is true that fiducial registration is more accurate. However, I would point out that all artwork (fids, traces, land patterns AND holes) are referenced to the same origin. This is regardless of a "drill process". The drill process has the highest degree of inacuracy. Typically this is due to drill wear and drill stack height.

Good luck in your search.

reply »

Mike

#28001

Fiducial recognition using laser | 12 April, 2004

Ken,

Thanks for the feedback. Unfortunately (or fortunately) I am in Mexico and we make our own PCBs. That is another conversation entirely and another of my fix-it projects but currently we drill the PCBs and align the artwork to the holes visually. Most of the time it's pretty good and most of our boards are single-sided so it's good enough but I can't trust the registration of the artwork to the tooling holes. When I worked in Solectron and we bought all our boards we could pretty much guarantee the registration within a known tolerance but here it is just too variable to trust.

The sensor I found has a target distance of 5mm and my process has almost nothing on the glue side to cause collisions so I think I will go with that. The algorithm for detecting the fiducial is very simple IF the starting point is withing the fiducial. If the board is so far off that you have to hunt for the mark then it gets a little more complicated but it's still achievable.

Thanks, Mike.

reply »