NOTES FROM THE TEST BENCH
By Bruce Hofer
Two months ago when I wrote about parasitic oscillations, I just happened to be the midst of trying to understand and fix an unusually perplexing example in one of my own designs. I mentioned that most parasitic oscillations involve either an unintended resonance or feedback around a loop that exceeds the Nyquist criterion for stability. I suggested the use of an RF spectrum analyzer with a custom made input probe as a tool for chasing these oscillations. Indeed, I practiced what I preached!
This month I would like to extol the virtues of using the tracking generator feature that is present in many RF spectrum analyzers. If you are ever given the choice, ALWAYS include the tracking generator option (if available) when buying or leasing an RF spectrum analyzer. The tracking generator adds the incredibly powerful capability of looking at circuit transfer functions, and revealing resonances. It lets the engineer observe the results of his corrective actions, and evaluate their effectiveness.
In my particular case, it turned out that the circuit board layout had unintentionally introduced a fairly significant amount of stray inductance in the common side of an input attenuator design. This created an unusually high-Q resonance around 50 MHz with the stray capacitances that I already knew existed in my circuit. This resonance was very easy to observe by connecting the tracking generator output to the input of my design, then probing the attenuator circuit with the RF spectrum analyzer and custom probe that I described two months ago. I had been looking for the fundamental cause of the oscillation in the wrong place for months!
Once I had confirmed the existence of the resonance using the tracking generator, I was able to quickly estimate values of stray inductance and capacitance that explained the 50 MHz oscillation. Adding these parasitic elements to my circuit simulator quickly improved the correlation between predicted results and actual measurements. The solution was both simple and highly effective. All I needed to do was add an R-C damping network between the resonant common circuit trace and ground. In my case, it turned out to be easier to connect the added network to a power supply plane instead of ground.
Happy hunting, and enjoy this month’s edition of our newsletter…
Audio Precision co-founder and Chief Analog Engineer