neoflash,
The Kolmogorov-Smirnov test is one way to do it. See the following: references:
. http://www.itl.nist.gov/div898/handbook/eda/section3/eda35g.htm **broken link removed**
**broken link removed**
.
Te third reference provides a correction to the critical values for a discrete, as opposed to continuous distribution.
Regards,
Kral
neoflash,
Yes. As the first reference indicates, the math is raltively simple. Mnay DSPs have no hardware divide instruction, but if N is always the same, then the divide can be replaced by a multiplication by the fixed value 1/N. The cdf could be a look-up table.
Regards,
Kral