Sound Card Pink Noise Response

Pink noise has equal energy per octave, which means that the spectrum falls at higher frequencies where there are more spectral lines per octave.

Pink noise happens to be fairly close to the spectrum of typical musical signals. There is less energy in the 1 to 4 kHz range (2 octaves) where human hearing is most sensitive, and much more energy at low frequencies where our hearing is poorer (about 5 octaves). Sound systems are designed with this in mind, and are typically capable of much more power at low frequencies than high.

A Pink noise test signal can thus deliver a lot of power to big low frequency drivers without exceeding the limits of small high frequency drivers. In comparison, White noise (with equal energy to all frequencies) would need to be used at low levels to avoid damage to the high frequency units.

The spectrum of a Pink noise signal falls at -3.01 dB per octave. Thus, to see a normal frequency response plot you need to apply a +3.01 dB per octave Tilt using the Spectrum Curves option in Y-log Spectrum mode. If you view the raw Pink output signal it will appear as a flat line at about -26 dB re: 1V Pk, or about 4 dB higher than a comparable White signal with the same +/-1 volt range.

But in actual use the difference could be even greater, if you adjusted both to have the same effective output at higher frequencies. For example, the Pink signal has about 12 dB less power at 700 Hz than at 40 Hz, or about -38 dB, whereas the White source has the same -30 dB at all frequencies. So you could set the Pink output 8 dB higher than you could set the White source, if power at 700 Hz was the limiting factor for the overall speaker system.

If you do a long spectral average on the direct Pink output from the Daqarta Generator and view it with the +3 dB/Octave Tilt, you may notice that it is not a prefectly flat line; there is about +/- 0.85 dB of ripple, due to the nature of the Pink noise generator. If you want to eliminate that, you can create a Mirror Curve file that perfectly compensates. You can then use that .CRV file instead of Tilt.

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

See also Frequency Response Measurement