fiducial mark

I published some info under http://www.geocities.com/stefwitt/Vision.html

reply »

Is this problem with local fiducials or global? Is the problem on the same equipment? Is this a cad file o centroid? How are you measure the coordinates?

reply »

How does a fiducial work? I am going to venture a guess as how it ralates to the parts you are trying to place...

A PCB fiducial is a landmark. The fiducial tells the machine where the board is located in the machine. A single fid can only determine x and y position. It takes two or more to determine x,y and theta position.

How does the machine know where to place the parts after it scans the fids? The fiducial has a coordinate like any other part on the board. That coordinate is referenced to the same coordinate system as the parts. IF the absolute position of the fid is know then when the actual fiducial(s) are scaned by the machine, the difference in position is the offset position of the PCB and corresponding parts. This of course assumes your machine is calibrated.

It (fiducial)is merely a landmark to tell the machine where exactly the board is located and how to offset the mount data to exactly mount the parts.

Local part fiducials work the same way, but are used to judge the local position of the pcb for the a component. These can be useful to correct for board shrink / stretch.

In order for fids to work, you must have two things: 1. calibrated equipment. 2. cad data which references the fids and parts.

Many machines don't even need fids. Some Chip shooters and even older machines used locating or "tooling" pins...these often were before the days of vision systems.

Hope this helps

reply »

To clean up other replys I will add that some most all machines have global correction. Most placements are an off sets from pcb 0,0 usually the lower left corner of the board. When the machine locates the fiduitals and these fids are not exactly where the data indicates they are it will make adjustments called global correction. This is to compensate for pcb srinkage or slightly shifted patterns that sometimes occure on panalized pcb's. Some machines allow for global correction to be turned off. Shifting of parts will be in one direction if all fids are off from data provided in the same direction. If first fid is correct and second fid referance is off then parts near the second referance point will be shifted further than parts near the first referance point. We run into problems when pcb's are rev'ed and pcb changes to fid locations are made and nobody informs us. To fix try this if all parts are shifted equally change fid data the amount and oppisite direction that the parts need to move. Say that all the parts are placeing 1mm too far to the left. Then change x cordnate of all fids to x-1mm. If parts are shifted more towards one end of the board then work with the fid on that end of the pcb. As far as search aria is concerned only use as much as nessary to consinstantly catch the fid, being mindfull of other nearby objects that may be confused with the target. If your machine locates the pcb consinstantly in the same place this should not be much larger than 2x fid size. On some machines that have fixed rails such as MyData table type, search aria is suppose to be less than the pcb's set into the rail. If search aria is too great it will detect a pcb not completely set into the rail.

reply »

Hi Chris, About marks detect: If you have vision system you can recognize the fids and not center them manual. If only camera you have to put the center of the camera in the center of the mark.

reply »

Chris- I use the SMEMA Standard as a reference. Goto http://www.smema.org and click the Council Standards link on the Home page. the SMEMA standard you want is 3.1

or

http://www.smema.org/smema3.1.pdf

Hope this Helps.

Enjoy, Cal www.CTDI.com

reply »

How well should the fiducial be centered in the teach mode? Consider that you see the fiducial with probably 40 times magnification on the monitor. If you see the cross-hair a bit off center, then this is most likely just by a few hundreds of a millimeter on the machine. This in itself will not contribute to a large offset of placement or mis-print. However, to answer your question, the machine will assume the center, where you teach it. Accepting the fiducial mark, does only mean that the vision system recognizes the template compared to the one you taught. The vision system feeds the information of the fiducial position within the field of view to the machine controller. The machine controller compares the fiducial position with the axis position and calculates the board position. If your machine displays a box or cross-hair during fiducial recognition, which is off center by far more then you taught the fiducial, then this particular fiducial may not be suitable for recognition. I recommend to determine if your machines operate correctly with known good data (test board, aluminum plate or glass plate). From there, you narrow down the failure in: Un-known fiducial position (no CAD data), pre-tinned fiducials with dome shape and make-your-own fiducial using through holes or via.

reply »