Greg,

Thanks for looking at this. After I posted, I came up with similar parameters, basically using as few samples as possible and putting them evenly throughout the gradient. I think division seems to work out a little better than subtraction in this case.

Gabe Cardona