Message written at 13 Oct 1997 21:37:28 +0100
--- Copy of mail to wicked-mentality@usa.net ---
In-reply-to: <20185.199710131440@zeno.maths.exeter.ac.uk>
	(james@maths.ex.ac.uk)
References:  <20185.199710131440@zeno.maths.exeter.ac.uk>

Do you mean the nlfilt opcode?  It is explains in

@InProceedings{JPF62,
  author = 	 {Richard Dobson and John ffitch},
  title = 	 {Experiments with Non-linear Filters; Discovering
		  Excitable Regions},
  booktitle = 	 {On the Edge},
  year =	 {1996},
  organization = {ICMA},
  publisher =	 {ICMA and HKUST},
  month =	 {August},
  pages =	 {405--408}
}

The actual formula is
\begin{equation}
          Y_{n} =a Y_{n-1} + b Y_{n-2} + \delta Y^2_{n-L} + X_{n} - C
\label{eq3}
\end{equation}
for various a, b, delta, C and L.

eg a=b=0, L=20, \delta = 0.8, 0.9, 0.7, C = 0.4, 0.5, 0.6 gives
reasonable values, as do
a = 0.4; b = 0.2; \delta = 0.7; C = 0.11; L = 20, ... 200
or
a = 0.35; b = -0.3; \delta = 0.95; C = 0,2, ... 0.4; L = 200

Of course if you meant something else, sorry to blast you with useless
information.
==John ff