The Case of the Invisible Signal

By Daniel Goff and Ian Stewart

We were testing a new circuit board design for a hydrophone amplifier. The board fits into a PCI slot in a computer next to a data acquisition board. A custom program works in conjunction with both boards to perform a self test on the hydrophone amplifier board. The self test involves injecting a signal of various frequencies and amplitudes from the data acquisition board into the amplifier board to test an Automatic Gain Control (AGC) loop circuit on the board.

The computer program showed that the self test failed so we removed the amplifier board from the computer and set it up for troubleshooting on the bench. The board takes a 0 to 5 V amplitude signal at a set frequency from the A/D card and attenuates it by a factor of 500 resulting in a 0 to 10mV signal for use by the AGC circuit since a very small amplitude signal is all that is needed to test the main AGC loop. The AGC circuit takes this attenuated signal first into a Voltage Controlled Amplifier (VCA) with a gain of -40 dB to + 40 db (0.01 to 100).

Using a laboratory signal generator, we injected a signal at the required frequency and amplitude and traced the signal through the AGC circuit using an oscilloscope. We initially deduced that the VCA in the test circuit had failed because the voltage signal displayed on the oscilloscope looked like random Gaussian noise at the AGC output.

The digital scope we were using had an FFT mode so one of us suggested that we try using that just to see if we could detect any trace of our signal. Using the oscilloscope in the spectrum mode with a logarithmic scale, the highly attenuated signal could now be clearly seen and was in fact being amplified by the VCA. The real problem could then be traced to a missing trace on a terminal of a relay. It turned out that the relay manufacturer’s component diagram was not very well drawn. Some of the lines on the manufacturer’s diagram were drawn so close together that it looked like two pins of the Dual In-line Package relay were in common when in fact they were not. When the printed circuit board was designed we had a 50:50 chance of using the correct pin and of course we got it wrong. The trace on our circuit board was connected to the DIP relay terminal that had no internal connection. The relay manufacturer was contacted and we learned that they were re-doing that diagram for the next version of their data sheet.

This example illustrates the importance of considering all details before plunging headlong down some incorrect path. There is usually more to a problem than initially meets the eye. In this case we had to look at both the time and frequency domains to find the solution to our problem.
Daniel Goff is a design engineer at NAVFAC ESC. He works on instrumentation, programming, system and circuit board design. Dan holds a BSEE from Portland State University.

Ian Stewart, P.E., is a design engineer at NAVFAC ESC. He works on instrumentation, programming, system and circuit design. Ian holds a BSEL from California Polytechnic State University and an MSEE from California State University, Northridge.