Sine Sweeps & Pink Noise

Modern room measuring tools use a swept sine wave, which has several advantages over pink noise. One big advantage of using a sweep is it offers a better signal to noise ratio. After the sweep is miked and recorded the software analyzes the result, applying a tracking filter as it measures the level at each frequency.

So while it’s assessing the volume at 40 Hz, for example, it ignores all other frequencies. This way midrange noise such as from the ventilation system won’t influence what’s measured at 40 Hz. This is a very powerful feature!

Software can also process and analyze a swept sine wave faster than humans can read the level of band-limited pink noise as it plays in real time.

At high frequencies, an SPL meter is stable enough to read quickly. But once you get below 100 Hz the needle dances around, and at very low frequencies it can swing 5 dB or even more over time. So you have to watch the needle carefully for 10 seconds or longer, and mentally note the approximate “average” level displayed. Again, this was what we had to use in years past.

Most audio engineers know that white noise has the same energy per Hz, where pink noise falls off at a rate of 3 dB per octave. So with white noise the band between 100 and 200 Hz has the same amount of energy as the range from 800 to 900 Hz.

Both bands encompass a range of 100 Hz, and so would read the same on a VU meter. But with pink noise, each octave (or fractional octave) contains the same amount of energy. So the octave from 100 to 200 Hz has just as much energy as the octave from 1,000 to 2,000 Hz even though it spans a much narrower range of frequencies.

Most noise generator circuits create white noise, which can then be filtered to reduce higher frequencies to obtain pink noise. Early white noise generators used a reverse-biased transistor as the actual noise source.

But that requires testing many transistors to find one that produces “good” noise, then you have to amplify and filter the relatively small signal.

A more modern design uses a series of digital logic shift registers as shown in Figure 1. This clever design (not mine!) creates a very pure white noise that can be filtered as needed.

However, one big problem remains: The simplest type of high-cut filter, using one resistor and one capacitor, rolls off high frequencies at a rate of 6 dB per octave.

A single filter stage rolls off at 6 dB per octave, two stages roll off at 12 dB, and so forth. This is why loudspeaker crossovers always transition at a multiple of 6 dB per octave.

So a single filter stage gives twice as much reduction as needed for pink noise, and obviously there’s no way to use half a capacitor.

As you can see in Figure 1, four resistors and four capacitors are needed to obtain half the roll off of only one stage! (The fourth stage uses two capacitors because the necessary capacitance isn’t a standard value, so two must be combined to obtain the capacitance needed.)