PCBs that work – and then they don’t

I recently received a frantic call from one of my old clients. About three years earlier I had designed a detergent-soap-mixing control circuit for them. The circuit monitored the alkalinity of a soap and water mixture and by controlling the water flowing into the mixing tank, it could maintain a fixed soap concentration. 

The company sold a few thousand mixing systems per year. They typically bought about 500 printed circuit boards at a time and soldered the surface-mounted components onto the boards in their own shop.  For about three years they had no problems.  But on the day I got the call, my client said they were experiencing a rash of strange problems.


About 75% of the boards would not pass a functional test. The symptoms were all over the map and were never consistent.  To keep things going, they were forced to cherry pick from the 50 or so completed units, installing only those units that passed a functional test into their machines.  They put the rest of the nonworking units aside.  I requested that they send me a few of the defective boards as well as a few blank unsoldered boards.

My client assured me that they had not made any component changes.  The trouble all started when they began using printed circuit boards from a batch of 500 they purchased.  But, since the boards all came from the same manufacture, the cause of the problems was in doubt. They were perplexed.

On the following day, I received three non-working circuit boards and three blank boards.  I fired up the boards.  The symptoms were indeed strange.  I picked one defective board and started the process of tracking down the source of the fault.  It was time consuming, but I finally found the problem: A spot of extra copper, which was shorting out two traces on the board. The spot was a star-shaped flake of copper under the green solder mask.  When I removed the short, the board worked perfectly.  Under magnification, I looked at every trace and every plated through hole on the other boards. 

Sure enough, in random locations on both sides of the circuit board, I found the same star- shaped solder flake. On one of the boards, the copper flake fell in a blank area where no traces or components resided.  Clearly, that board would have worked fine, since the solder flake would not have caused any problems. The flake was clearly some dust or dirt that somehow got into the circuit board fabrication process and resulted in a bit of extra copper deposited on the boards in a completely random location.

Ultimately from the stack of 500 blank boards my client was able to cherry pick about 50 where the solder flake fell in a safe area.  Knowing what to look for, they also were able to fix about 50 of the soldered boards by removing the copper flake.  The rest of the blank boards were shipped back to the manufacture. Although only about 400 boards were returned to the PC board fabricator, they quickly provided 500 new boards without charge.  In about a week, my client s production line was up and running again.

D avid A. Johnson, P.E. is a consulting electronics engineer with over 30 years practical experience in the areas of analog & digital design, applications, fiber optics and RFID and the founder of discovercircuits.com, where he originally wrote about this case in his monthly Circuit Diagnosis column. His article has also been published on EETimes. 

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.