Sound Card Swept Frequency Response

The traditional approach to frequency response measurement is to drive the speaker with a pure sine wave and slowly sweep its frequency over the range of interest, while monitoring the output of the microphone.

The Daqarta Generator can provide the swept sine wave using the Frequency Sweep modulation option, and you can observe the microphone response in Spectrum mode. But on its own, that would simply show a slowly moving peak whose height changed with the response. A better approach is to use Spectrum Averaging in Peak mode, which will cause the peak height to be retained at each frequency as the sweep moves along. A complete frequency response will thus be obtained on one screen.

But there are some problems with an ordinary swept response. The swept response is made up of many individual spectra, each obtained by applying an FFT to 1024 samples of the signal. Unless the FFT sees an exact integer number of cycles of an input frequency, it will show the spectrum of any fractional cycle as leakage from the main peak into adjacent sidelobe or "skirt" regions. This reduces the peak value, with the biggest reduction coming when the signal has an extra half-cycle, and smaller reductions for smaller or larger (closer to the next full cycle) fractions.

Since during a sweep the instantaneous frequency is constantly changing, each FFT sees a constantly-changing fractional cycle as well, with the result that the peak bounces around between the true value and something less, giving a ragged-looking response. Additional sweeps may result in "filling in" some of the dips, but in general this is unsatisfactory.

The customary way to deal with this leakage problem is to use a window function with the FFT. Most window functions are designed to give a narrow peak with low sidelobes, but those windows still give a sizeable peak error for the "wrong" input frequencies. Here you should use the Flat-Top window, which has a wider peak and higher sidelobes, but gives a very good frequency response when used for this purpose.

See also Frequency Response Measurement