Thanks to a suggestion by @iannesbitt, we also changed the spectrogram color pallet to be more color-blind user friendly. It also helps those of us that can see the full spectrum.
The old spectrogram used a rainbow color scale, which now known to be ineffective at conveying numeric information properly, and virtually impossible to read for most colorblind users. @kwinkunks shared this image of a colorized LiDAR elevation dataset on Twitter not long ago, which is a good demonstration of how rainbow colormaps distort our visual perception of data:
Both colorblind and non-colorblind viewers will notice that the two right-hand color scales are more difficult to interpret in the blue and green bands, while those on the left convey stunning detail of abandoned river meanders in those same areas. This is because, in essence, the brightness of the image does not vary co-linearly with hue, and human eyes are much more attuned to perceive dark/light differences than differences in hue.
In light of this, we changed spectrogram color scale to something perceptually uniform and linear. You can read about perceptual uniformity/linearity here.
I ran the sound app last night and it seems to me that the color scale needs to be adjusted somehow so that the image is more centered in the color range. My graphics came out to be almost entirely orange with relatively little variation in the color pallet.
I think a lot of people suffer from man-made noises that tend to take the form of a continuous βtoneβ probably corresponding to he speed of some rotating equipment. This produces a high-pitched whine in the resulting sound track. I have imported the sound track into Audacity and notched out the tones (two of them, in my case) and the sound track is really much nicer to listen to. It would be really great if that notching function was built into the app (user would specify the normal-speed frequency of the notches).
Failing that, a simple approach (although not as nice) would be to have an optional low-pass filter. Distant quakes have all their energy pretty much below 4 Hz so an optional, steep, low-pass filter at 10 or 20 Hz would do the job.
Hi Ken, both of these are good suggestions. I fixed the color scale issue in the code yesterday and am running some tests to make sure it always snaps to the noise floor of the instrument. Your suggestion of the creation of stop bands to get rid of mechanical tones is also a good one. I have tested and had success using a bandstop filter to remove these tones. This may take more time to implement since it involves changes to the frontend, so stay tuned.