Updated "Calculate Amplitude for Frequencies": white is flat?

Discussions
1

There’s a few changes in the updated Calculate Amplitude for Frequencies node.

The values scaling is different: is it now between 0 and 1? If it is, good! But the frequency bias has also changed: before, it was “pink is flat” (equal energy per octave), now it’s “white is flat” (equal energy per frequency interval).

Could it be an option rather than one or the other?

2

The values scaling is different: is it now between 0 and 1?

Yes, assuming the input audio samples are between -1 and 1, the output amplitudes are between 0 and 1.

But the frequency bias has also changed: before, it was “pink is flat” (equal energy per octave), now it’s “white is flat” (equal energy per frequency interval).

When Frequency Bin Averaging is None, isn’t “white is flat” always the correct behavior? My understanding of white noise vs. pink noise is that white noise should (ideally / on average) have an equal amplitude for every frequency bin, and pink noise should have decreasing amplitudes as the frequency increases.

If you do want to go back to the old behavior, you can still use the old version of the node. In Vuo 1.2.3 and earlier, the Node Library lists vuo.audio.analyze.fft. In Vuo 1.2.4 and later, it lists vuo.audio.analyze.fft2, but vuo.audio.analyze.fft is still installed. If you create a composition in Vuo 1.2.3 containing vuo.audio.analyze.fft (the deprecated node) and open the composition in Vuo 1.2.4 or later, it will still contain the vuo.audio.analyze.fft node with the old behavior.

Or you can paste the old node directly from Vuo 1.2.3 to Vuo 1.2.4. Or just paste the following text onto the canvas:

digraph G
{
CalculateAmplitudeForFrequencies [type="vuo.audio.analyze.fft" version="1.0.0" label="Calculate Amplitude for Frequencies|<refresh>refresh\l|<samples>samples\l|<frequencyBins>frequencyBins\l|<frequencyBinAveraging>frequencyBinAveraging\l|<amplitudes>amplitudes\r" pos="459,67" _frequencyBins="\"255\"" _frequencyBinAveraging="\"quadratic\""];
} &nbsp;
3

Thanks for the info. Depending on what type of audio data you want to show, both white and pink bias are useful.

From Waves’ PAZ Analyser user manual:

Pink noise is defined as noise that has constant energy inside any frequency band with a constant Q (i.e. frequency/bandwidth=a constant value). Pink noise is an important audio signal since it represents the average spectrum of a typical audio signal, and because the ears analyze the sound in approximately constant-Q bands. Because PAZ has almost constant-Q bands, a pink noise input will show up as a flat spectrum when no weighting is applied.

And from Metric Halo’s SpectraFoo user manual

EQUAL POWER PER OCTAVE AND EQUAL POWER PER HZ ACCUMULATION MODES:
SpectraFoo’s underlying FFT generates more data than can be displayed on screen at once. The way that this data is treated when the system computes the trace is called the accumulation mode. SpectraFoo provides two accumulation modes that you can choose using the Spectragraph Controls window:

  1. When the Accumulation Mode is set to White Noise is Flat, the Spectragraph shows the maximum power of all of the spectral lines in the FFT that fit into each point in the displayed trace. This has the effect displaying white noise (or noise that has an equal amount of power at each frequency) as a flat line. This mode is traditionally used when the frequency scale of the graph is linear.
  2. When the Accumulation Mode is set to Pink Noise is Flat, the Spectragraph shows the sum of the power of all of the spectral lines in the FFT that fit into each point in the displayed trace. This has the effect of displaying pink noise (or noise that has an equal amount of power in each octave) as a flat line. This mode is traditionally used when the frequency scale of the graph is logarithmic and corresponds to the type of display available on a traditional RTA.

So I think both “accumulation modes” are useful and that’s why an option would be interesting. But, if it’s not possible, I’ll happily live with the new bias. The new scaling is the big progress anyway and fits better with other data processed by Vuo.

Thanks!

4

Helpful descriptions, thanks. The part about “when the frequency scale of the graph is logarithmic” leads me to believe that the “accumulation modes” are similar to the Calculate Amplitude for Frequencies node’s Frequency Bin Averaging options.