Sound Card White Noise Response

White noise, by definition, has equal average energy at all frequencies. Thus, if you measure its spectrum you might expect to see a perfectly flat line. And you will, in fact, but only after a long Spectrum Average to smooth out the moment-to-moment random variations.

White noise has a higher average energy than Impulse or Step signals, roughly 32 mV per spectral line (- 30 dB) for a +/-1 volt noise signal. The equivalent instantaneous waveform voltage is about 0.5 volts RMS. If you use this to obtain an absolute calibration, instead of just a response shape, then you can reduce the External Gain on the Input line by a factor of 32. If you are starting from a default of 1.00, just change it to 0.03125. Daqarta will compensate by reporting a 32x higher response.

White noise is sometimes used to avoid acoustic standing waves when measuring room responses, but Pink noise response is usually preferred. The reason is that sound systems are designed to handle higher amounts of power at low frequencies; since White noise puts out equal power at all frequencies, the overall power must be kept within the limits of the high-frequency drivers.

Note that you should never use a Window function to view a noise response; use them only for continuous waveforms.

Low-Frequency Response via Decimation

You can use Decimation with the White noise method to measure response at very low frequencies. The Generator output produces continuous White noise as described above, while the input is effectively sampled at a lower sample rate that allows high resolution at low frequencies.

For example, with Decimate Factor set to 100, the effective input sample rate is reduced from the default of 48000 down to 480 Hz. The Daqarta Spectrum response is 512 spectral lines (from 1024 raw samples) spread over a frequency range equal to half the sample rate (the Nyquist frequency). The default spectrum thus has a resolution of 46.875 Hz, from 0 to 24000 Hz. This is not adequate to measure the frequency response of a sound card's input and output AC coupling. But with 100x decimation the resolution is 0.46875 Hz, from 0 to 240 Hz, which is more than adequate for most sound cards with typical response down to 20 Hz or so.

You can set the Decimate Factor even higher for finer frequency resolution, but note that this increases the time for the Spectrum Average to complete. At 100x, each frame takes 100 * 1024 samples, or about 2.133 seconds. Thus a typical 32-frame average takes well over a minute.

Even if you don't mind the wait, remember that as the effective sample rate goes down, so does the highest frequency shown in the spectrum. You typically want to see some of the flat part of the response above the low-frequency cutoff, to be sure you are getting the whole low-end response. If you set the effective rate too low, you might miss too much of that upper end.

You can use Decimation with a DC measurement system to measure very low-frequency devices or systems, limited only by how long you want to wait for the average to complete.

See also Frequency Response Measurement