Kaylor Karl commented:

>We are currently using a 2 layer FR-4 board as a 2-up panel in our assembly
>factory and our board vendor has asked us if we are able to use x-outs.
> Currently, we are not able to use x-outs given our equipment and line rate.
> They say that this problem is causing 6%-8% fallout and want to increase
>our cost.

For a "normal" 2-layer, that would seem to be a very high defect rate.

>Consequently, our vendor has asked us if we can accept boards
>that have two good boards glued together into a panel.  They claim that
>there will be little difference between the panels and want to move forward
>with a trial run.

Your "Frankenstein" boards will have to be glued in such a way as to hold
perhaps +/- 0.002" tolerance to fiducials, etc. during SMT assembly (I'm
assuming here) and that would seem unlikely. If each unit in the array has
its own fiducials, and you program the SMT machines to look independently
at each unit's fiducials, it might work. I'd be worried that the glue might
fail in reflow, etc., or that you'd trap ionics, the glue might have
ionics, yada, yada, yada.

>Does anyone else produce / use glued boards?

Never heard of it...

>Is this fallout unusually high?

IMO, yes, but perhaps there is something unusual in this design.

>What are some problems we should look out for?

When you enter a dark castle, there is always a risk of surprise...
My philosophy with X-outs is that the panel had some problem - perhaps an
etch problem, imaging, etc. and that if an array has one defect, my
confidence in the entire array is reduced. If one unit in the array had
mousebites, then there may be a latent defect in the other unit even if it
tested OK.

We have several 4/6 layer units in two up arrays here for which we accept
no "X-outs". On a couple of designs, we tried using X-outs, but we are
trying avoid that on all new stuff.

In assembly we have to program the machines to find a particular fiducial
on each unit, and to see "no mark" on the X-out unit. In practice, if the
machine sees the fiducial, it populates the unit. We tried having the
vendor mark the fiducial with a black pen, and the result is the fiducial
camera sometimes found the mark (either the vendor blacked the wrong
fiducial, or simply marked an X, which the camera ignores), and placed
components on the board. So then we have a fully populated "bad" board with
a few hundred dollars worth of parts coming down the line.

Perhaps the board vendor has a philosophical disagreement that the unit is
good, so it ought to be sold. With the trend towards smaller board designs,
the use of 2,3, or 4-up arrays is probably growing. Components are
continuing to increase their share of the assembly cost. For our SMT line,
with many handling steps and machine conveyor transfers, arrays of smaller
PC (PCI, EISA, etc.) format boards makes compelling sense. The "good" unit
in an array seems like an unreasonable loss to the board fabricator if the
customer does not buy it; but the assembler will be taking a risk if he
does.

So it's a conflict of interest, and you will probably have to negotiate
your way out of it.


Jerry Cupples
Interphase Corp
Dallas, TX
http://www.iphase.com