Circuit Board Rework

Marc had a tale to tell at lunch yesterday. The company he worked for built a short run of circuit boards to validate a new design. The boards did not work. After some investigation they discovered that eight control lines had been routed incorrectly. Marc, using a microscope, an extra sharp Xacto knife and a very fine soldering iron, and four hours of time was able to fix one of the boards.

So what? you say. Well, let me describe the problem in more detail. Circuit boards, otherwise knows as PCB's (Printed Circuit Boards) or PWA's (Printed Wiring Assemblies) replaced point to point wiring 50 or 60 years ago. Simple circuit boards have a layer of insulating material coated with a thin layer of copper. A design is drawn on the copper, typically using a photographic process. The parts of the copper that are to form the "wires" in the circuit are protected by a masking material. The board is then exposed to acid and the unwanted material is eaten away by the acid. More complex boards may involve multiple alternating layers of insulating material and metal film. Imagine alternating layers of aluminum foil with sheets of paper. That will give you some idea of the thickness of these layers.

The board Marc was working on was six layers. That is, six metal layers alternating with five insulating layers. The whole board was probably no more than 1/16 of an inch thick. The lines ("wires") he had to fix were on the third layer. So he had to cut through two metal layers and two insulating layers to get to the problem. After he fixed the problem, he had to reconstruct the lines from the metal layers he had previously cut away.

The design for this board was derived from a design from an older board. The problem came about because of a flaw in the rules for routing lines in the old design was carried over to the new board. Most lines on a board have very specific endpoints. They go from pin X on one chip to pin Y on another chip.

However, some lines on a circuit board are interchangeable. For instance: the data lines on a memory chip. Typically the eight data lines will be named D0 through D7. Normally you would connect these eight lines to the same lines on the processor. D0 on the memory chip would connect to D0 on the processor, D1 to D1, D2 to D2, and so on. However, this is not a requirement. We don't really care which line on the CPU connects with which data line on the memory chip. As long as the bit comes out on the same line we use to store it, and we don't have any other devices connected to this line, we don't care.

So after the electronic design is done, it is turned over to the layout guy whose job is figure out where to draw the actual lines on those layers of foil in the circuit board. These data lines can be marked as interchangeable, so that if it is easier to route lines using different data bit numbers, it is okay to do so.

The problem was that back when the original design was done, eight control lines had been marked as interchangeable when in fact they were not. Earlier designs had not taken advantage of this error and so they had been routed correctly and the boards had worked. The new design was more complex and the layout guy took advantage of the design rule and swapped these eight lines around. As a result the new board did not work.

Marc did this rework to verify that this was the only problem with the board, and indeed it was. When the rework was complete the board operated properly. They could have just made the change in the layout and made a new batch of boards, but the risk was that the new batch would still not work because of another problem that had not yet been discovered.

This rework was difficult for a couple of reasons. The first is that the change had to be to an internal layer, which meant cutting through the two outer layers and then rebuilding them afterwards. The other problem, and the one that made it really difficult is the size of the lines. After mark had made his "incision" in the board, he pulled it out from under the microscope to look at it, and could not find the spot he had been working on. It was too small to be easily noticed. Going back under the microscope and taking note of the position, he was able to locate the work site. It was smaller than a bread crumb.

To be able to do such fine work, you need a really sharp knife. While an Xacto is plenty sharp for making cuts that are visible to the naked eye, for this kind of work, Marc had to sharpen the Xacto using emery paper. Before sharpening, the blade looked ragged under the microscope. After sharpening, it looked better.