---

I’m working on a sound localization project based on three microphones — each connected to an Arduino — and a fourth Arduino connected to a Mac running Processing. The three Arduinos provide the analog-to-digital conversion, and the fourth microcontroller determines which microphone encounters the audio front first. The fourth Arduino also communicates with the Mac Processing, which provides a graphical interface for the sound localizer. I initially tuned the system using a pure 1 kHz tone source, with great results.

However, with real world sounds, performance was much less spectacular. Identifying the source of low 100-200 Hz signals was expected, given the longer wavelength relative to the sensor spacing and the effect of the room acoustics. Humans have similar difficulties localizing a low frequency source — an advantage when hiding a massive subwoofer in a living room, for example.

Unexpected, however, was the performance above 1 kHz. Sometimes it was spot on. Other times, the results were essentially random. My first thought was there was some sort of delay in the output of the microphone processors. Next, I looked at the fourth polling microcontroller. I tried swapping out microcontrollers to no effect.

Finally, I examined the microphones. According to the specifications sheet, the microphones had a deep cardioid sensitivity pattern at 1 kHz. There was no other test frequency noted in the specifications sheet. Delving deeper, I located the datasheet on the condenser microphone element used in the microphone. The sensitivity pattern was identical to that in the specifications sheet of the microphone. Of course, that’s impossible.

The microphone housing and mesh opening necessarily distort the free-space directional response of the microphone element. Clearly, the manufacturer of the microphone simply copied the specifications of the microphone element and reprinted them in their product specifications sheet.

Through testing of my own, I determined that each of the three microphone assemblies had virtually omnidirectional sensitivity at 3 kHz and then something resembling a cardioid response at 6 kHz. So much for the specifications.

I’ve always been skeptical of specifications associated with audio gear, but now I’m doubly so. Whether from carelessness, laziness, or with the intent of intentionally misleading consumers, specifications for electronic components, instruments, and assemblies are not to be believed at face value. You can’t check everything, but if you have that one project that you just can’t get working, consider that the specifications may be simply marketing fodder. This goes for servos, sensors, and the cables that you’re considering for your next robotics project, as well.  SV