1) shrinking the window size doesn't help (also it turns out I was wrong the actual number of samples needed is higher than I thought)
2) filtering doesn't do SQUAT for these purposes
3) this technique seems to be brittle when the fundamental ISN'T one of the frequencies we're looking for (it starts picking out the strongest frequency it IS looking for which can be quite far away depending on the signal presented)
it seems like I might still have some good success doing a bit of simple analysis. Picrel is the flavor of thing I'm thinking of doing. Not exactly it but nice to see I might be on the right track here
