Electronic Assembly Fault Detection

300

Just a guess... gut feel.

Del

Many will say my guess is low, but inspection alone will never get 100% especially if there are faults like short circuits/open circuits due to poor quality pcbs.

Better to put the quality control in to the actual build rather than inspection after it.

If you combine inspection and test you will get that figure upto nearer 498.

(Some 'faults' will not prevent correct operation of the circuit)

Del

You assume that all of the faults CAN be detected visually, which is not necessarily the case, especially for open or short circuit situations.

Your detection will approach 100% as the time spent detecting approaches ∞.

Assuming, as was stated earlier, that all faults can be detected visually, which they can't.

100 components on a board is not a lot to check - think of it as 10 columns of 10 components each. If the inspector has a "good" board to compare to, rather than just a drawing, I'd say no more than 5 seconds per component would be required for the first board, and no more than 1 second per component for the last board, since the inspector will become way too familiar with the board over the time spent inspection ALL 1000 of them. So, on average, let's say 250 seconds per board. About 5 minutes. Times 1000 boards = about two weeks.

Less if your inspector goes insane.

The visual inspection would catch most missing components. But I believe that some type of diagnostic test circuit placing board in working environment would be more efficient. This would reduce the time one would take searching the whole board. Would also catch faults not visible by the eye.

You have made the statement that of the 100,000 components there are 500 faults which is .5%. But if each one of those faults is on a different board then of the 1000 board 50% of them are bad. Seem a little high to me. I would be addressing the problem of the missing component, wrong polarity, wrong component and bad solder joints.

Batch=1K Boards (3 shifts)

Board=100K Components

Faults=500=0.5%

A well trained inspector and FM guidelines he should be able to catch them all. Assuming that the faults were spread among the batch of boards and only involved one 8 hour shift, 1 operator should find 33% and since we're talking about visual inspection only (attributes), I'd say about 2.0-3.0 minutes /board.

could try using this system, Landrex Optima 7300 automated optical inspection system

You'd be very lucky to get above 60% to 70% of faults by visual inspection alone. If you've got anything like that number of faults you should invest in some kind of pre-screen or in circuit test as well as some kind of functional test. How much will it really cost you when one of those boards fails in the field?

Oh: five minutes per board on first pass through; half an hour per board for the next 100 or so, which don't work; half a day for the next few 10s of boards, and, give up with a "bone pile" of about 50 to 100 boards.

Experience has shown that human visual inspection detect only about 80% defects.

So, while we want and are capable of detecting 33%/shift of the total 500, in actuality the inspector will have detected only about 80% if and only if they have been well trained in the possible failure modes outlined in your Query...