On Sat, 27 Dec 2008 20:04:33 -0800 (PST), Greg Lovern wrote:

Thanks, apparently the error only happens at unrealistically high
rates (such as 500% in this case..), and therefore doesn't matter.

I was testing with randomly generated input values, not limiting the
test to realistic values. I changed the test to keep the rate down,
and at those rates the error doesn't seem to occur.


Greg

I am surprised that there is that large an error (1%) , but I don't have time
to dig into the reasons. I'm glad it works well enough for you.
--ron