Updated gravity transform, cleaned up graph shader

shaders/graph/1.frag

@@ -7,10 +7,50 @@ uniform ivec2 screen; /* screen dimensions */
 #request uniform "audio_sz" audio_sz
 uniform int audio_sz;
 
-#request setavgframes 4
+/* How many frames to queue and run through the average function */
+#request setavgframes 6
+
+/* Whether to window frames ran through the average function (new & old frames
+   are weighted less). This massively helps smoothing out spikes in the animation */
 #request setavgwindow true
-// #request interpolate true
-#request setgravitystep 0.06
+
+/* Gravity step, higher values means faster drops. The step is applied in a rate
+   independant method like so:
+   
+   val -= (gravitystep) * (seconds per update) */
+#request setgravitystep 5.2
+
+/* When we transform our audio, we need to go through the following steps: 
+   
+   transform -> "window"
+       First, apply a window function to taper off the ends of the spectrum, helping
+       avoid artifacts in the FFT output.
+   
+   transform -> "fft"
+       Apply the Fast Fourier Transform algorithm to separate raw audio data (waves)
+       into their respective spectrums.
+   
+   transform -> "fft"
+       As part of the FFT process, we return spectrum magnitude on a log(n) scale,
+       as this is how the (decibel) dB scale functions.
+       
+   transform -> "gravity"
+       To help make our data more pleasing to look at, we apply our data received over
+       time to a buffer, taking the max of either the existing value in the buffer or
+       the data from the input. We then reduce the data by the 'gravity step', and
+       return the storage buffer.
+       
+       This makes frequent and abrupt changes in frequency less distracting, and keeps
+       short frequency responses on the screen longer.
+       
+   transform -> "avg"
+       As a final step, we take the average of several data frames (specified by
+       'setavgframes') and return the result to further help smooth the resulting
+       animation. In order to mitigate abrupt changes to the average, the values
+       at each end of the average buffer can be weighted less with a window function
+       (the same window function used at the start of this step!). It can be disabled
+       with 'setavgwindow'.
+*/
 
 #request uniform "audio_l" audio_l
 #request transform audio_l "window"