This is Fire. It looks pretty good.

It makes sense that 26 has more difference between red and blue. The transmission cutoff is farther toward the NIR end of the spectrum, so less red is included. So the red channel gets less light. But the red channel has a higher proportion of NIR:red, and the red channel is supposed to be visible red light. So the increased difference between the red and blue channels does not add any information to NDVI.

Ok, that's a good explanation. The fire does look pretty good. I really like the the images aesthetically too, the red sky/blue plants is eerie.

The blue channel (NIR) is a little bright in the 19 photo (and the others as well). There are lots of pixels with 255 for blue, so the dynamic range of NDVI is being clipped. It might be worth trying something like R 310, G 500, B 660 next time.

Here is is again in full sun. with B at 660 there is still substantial clipping. Should I keep dropping the white balance for B? :

That one is really contrasty. I guess some experiments are in order to see how various white balance settings work and how the other photo settings help. I set the "Exposure" a third of the way toward "Dark," and increased the "Contrast" a little. So there are lots of variables to be fine tuned. I guess I would try a lower blue setting next. But that's kind of a nice IR photo.

I wish we could switch the settings around while the camera was in webcam mode, it would make this process a lot easier. that said, I'm complaining about features that are very rare in cameras like this.

SYSCFG.TXT

Great, copied this into http://publiclab.org/wiki/infragram-point-shoot