The eyes have it: Failing amplifier boards no match for engineer's keen vision

In the mid-1970s, I joined a company that designed and manufactured scientific instruments for studying plants. LI-COR provided instruments such as the world’s only portable leaf-area meter, used to measure the area of a leaf without detaching it from the stem of a plant. My first assignment was to develop a new light-measuring instrument—a quantum radiometer photometer. The core was to be a chopper-stabilized amplifier used in a transimpedance configuration. Back then, there were no monolithic ICs to perform this task. The design involved using MOSFET and JFET transistors as the chopping switches along with some monolithic amplifiers surrounded by resistors and capacitors. The design was challenging. An early problem was that the offset of one of the amplifiers drifted for the first 100 hours of operation, causing the instrument to lack long-term stability. The cause was an ion-contamination problem at the IC manufacturer. After solving this problem, the final amplifier achieved about 50 pA of input current, offset drift of less than 50 nV/°C, and noise of approximately 30 nV/

Later, a production technician asked whether I would help him troubleshoot a bunch of the amplifier boards that had failed to pass their performance test. He had been unable to determine their cause of failure and had accumulated about 40 boards in a box.

The first amplifier had a peculiar behavior. When the gain was set with a feedback resistor greater than 10 kΩ, the amplifier output would saturate to its supply rail. With a feedback resistor of less than 10 kΩ, the amplifier would behave within specification. Failures on the production line are always more challenging to troubleshoot, because the circuit has never worked normally. This one ended up being a microscopic solder short between the positive and negative inputs of the amplifier.

However, most of the failures were boards that seemed to fully work but had excess noise. In examining signals on all the nodes throughout the circuit, everything looked normal, but the meter readout was unstable. Although its mean reading was correct, it erratically jumped around. I examined the circuit operation, then looked at the defective board alongside one that was working properly. Then I saw it: a resistor installed backwards.

To minimize board space, the assemblers mounted the resistors, most being axial-leaded 1/8W resistors, on the board standing up. The first amplifier stage had two resistors to set its gain. The resistor going from the negative input of the amplifier to ground had been assembled with the lead going to the amplifier input facing up in the air. In this configuration, the lead became an antenna that allowed the injection of interference directly into the input of the first amplifier stage—the source of the excess noise.

We inspected the rest of the amplifier boards; most had the resistor in backward. We added a note to the board’s test procedure to carefully check the position of this resistor before beginning the electrical tests.