Inspecting Printed Circuit Boards and Assemblies

How do inspection discipline and sampling plans decide whether a shipment ships or gets torn apart?

Before going into PCB design, my employer was in the telecom business. I started out putting PCBs into antistatic bags, then into individual boxes with appropriate labels. A group of eight distinct boards was placed in a larger box to form a die group. The big box labeling reflected the part-dash number and revision for each board. This was called “final prep” and was the last step prior to shipping.

A few selected to represent the whole shipment. Someone would audit the shipping boxes, looking for incorrect information on the labels and any other obvious defects. If they found a problem with the ones sampled, the entire lot would be re-examined by final prep. A missing antistatic awareness label could jeopardize the whole shipment. It could happen, especially toward the end of the quarter when we were sprinting to make our quotas.

The QC inspector and QA auditor were the last line of defense before submitting the die group to AT&T or one of the Baby Bells that came out of the antitrust settlement of 1984 that broke the phone company into several regional companies. After a few months, I transitioned to assembling card shelves and heat shields for telecom equipment racks.

Inspectors would check the card guides for burrs. They looked at everything from correct labeling to bent pins and even fingerprints. Each assembler had their own rubber stamp with their unique number. Mine was “ASSY68,” surrounded by a square in black ink. Traceability matters.

It wasn’t long before I moved to the Rack Wire and Test group, which assembled entire equipment racks. You’ve probably seen IT racks where the wiring looks like a plate of spaghetti and others that look so clean and orderly. I strove for cleanliness.

Eventually, I returned to a seated job assembling fuse/alarm panels installed at the top of the racks. The fuse panels had a 46-point wiring harness (23 wires) that wrapped around the inside of the two RU (rack-unit) enclosures. The wires were for signals, while busbars strapped all the fuses to power and ground.

I had my own way of dressing the wires into an S-turn so that the service loops to the fuses didn’t overlap. When they made me the lead assembler, I wanted all nine (plus temps) of us to take the time to form the 18-gauge wires just like mine so that it was easier to spot a mis-wire. Aside from the ASSY-number, there was no way to tell who built the fuse panel.

One day, my manager asked me if we were on “tightened inspection” – where the sample size was much higher than normal inspection. I stated that if the lot submitted to the AT&T source inspector passes, we would be eligible for reduced inspection. He shook his head and said, “You’re the only one” in the TransMUX department that is not on tightened inspection.

A sampling plan is derived from MIL-STD-105E. The document has several tables that specify how many units out of a lot or batch must be inspected based on the AQL (acceptable quality level) and the total number of units in the lot. A track record of rejects leads to greater scrutiny. A control chart would be the basis for monitoring a specific failure point if it becomes an issue.

One day, a job opening came up in the Radio division. Point-to-multipoint radios were like a precursor to cellular networks that we know today. One of the use cases was offshore oil rigs where the land-based managers stayed in touch with the array of drilling platforms at sea. This job was an in-process inspection. I sat at the end of a row of fine-fingered ladies who stuffed the boards and put them through the drag solder machine before soaking them in freon to clean off the flux.

To reject or not to reject? If I couldn’t find anything wrong with a PCBA, I pressed my triangular QA23 stamp into a red ink pad and applied it to the board. If placement or workmanship was off, I stuck little red arrows to the board, with a red tag to describe the defect. One thing the ladies didn’t want was a red tag on their work.

There was an issue with letting them fix a noncompliant solder joint without first rejecting the board. I found that out the hard way when I asked for a touch-up on a borderline solder joint. Ruth, my manager, said that “if it’s good, accept it and if it’s not, then reject it. It’s on you to decide what is acceptable.”

To be fair, the company needs that information so that it knows where to focus its process improvement efforts. The better course of action was to get in touch with Henry, the QA inspector, for his opinion. He would audit the lot, choosing a random selection of units for visual and electrical qualification. The source inspector would do the same with their chosen samples.

Analog circuits can be an unusual case. I found the meniscus of a mica cap jammed down to the surface of the board unacceptable. There was no room for a top-side solder fillet. Although that was a clear violation of the workmanship standards, these were analog boards, and any length of wire acts as an antenna. As a practical matter, we accepted the defect on the radio boards, where we wouldn’t have it on the digital boards.

One day, Ruth invited us four inspectors to move from in-process inspection over to receiving inspection. The only requirement was that we enroll in two concurrent Quality Assurance classes at a local junior college. My hand shot up immediately. One of the things that lifted my hand was a reluctance to put a red arrow on a PCBA with assemblers sitting right there. Rejecting something from a vendor seemed much less personal.

Studying the works of Dr. Edwards Deming (the godfather of statistical process control) and Dr. Joseph Juran (the architect of quality and author of the Pareto Principle) was indeed a life-shaping experience. We learned the theory about the standard deviation of a Poisson distribution along with practical aspects of quality assurance.

Receiving inspection: the first line of defense. Running the Receiving Inspection lab meant checking bare boards before they went into the stock room. Of course, all the components also had to be compared to the source control drawings. Looking for absolute perfection wasn’t the thing. What mattered is that the PCBs and their components “met the requirements.” Our field guide in these matters is IPC-TM-650, “Test Methods Manual.”

When a new part number came into the RI lab, it underwent first-article inspection. That meant that every dimension or feature was thoroughly examined. For example, if soft gold is specified on the fab drawing, we use a nondestructive method to verify its hardness by applying a specially cut diamond tip to an indenter. A fixture holds the board or a cross-section, and the indentor drops down, making a tiny indent in the metal. A bigger indent registers a lower number on the Knoop scale of hardness.

Pure gold is the softest and rates between 50 and 90; graded as ASTM B488 Code A. The next level of hardness is medium gold, which tests between 91 and 126 on the Knoop scale; classified as Code B. Hard gold is rated between 130 and 200, which would fall under ASTM Code C. For reference, diamonds are rated at the high end of the Knoop scale at 7000.

The receiving inspection lab also included a machine for checking the copper thickness in the via barrel. Another one measured the thickness of gold on the edge fingers. An optical comparator could measure tiny objects projected onto a backlit glass with graduation marks. The shadow of the part was projected at 10X actual size to facilitate the math. We also had a binocular microscope to find cracks in the PTH barrels.

A set of Mitutoyo pins was calibrated to check board warpage and hole sizes. We used plain old Scotch tape to see if any ink, mask or copper could be peeled from the boards in accordance with IPC-TM-650.

These days, we add automated optical inspection (AOI), x-ray and other technologies to look deeper into the PCBs. A reliability lab performs destructive testing, whereas all the other inspections are nondestructive. Weeding out the defects is one thing. Improving the process so that there are fewer defects is the ultimate goal of the inspection team. This is the inspector’s value-add that goes beyond a rubber stamp.

John Burkhert, Jr. is a principle PCB designer in retirement. For the past several years, he has been sharing what he has learned for the sake of helping fresh and ambitious PCB designers. The knowledge is passed along through stories and lessons learned from three decades of design, including the most basic one-layer board up to the high-reliability rigid-flex HDI designs for aerospace and military applications. His well-earned free time is spent on a bike, or with a mic doing a karaoke jam.