It defintely is "possible" since I did this a few years ago in P5 as part of an experiment, but not in an elegant way. I can think of a few new techniques that may I may try again in the near future as I can see some benefit in commercial terms, but not likely as a useable public product. There are a lot of issues that would probably make it impractical, such as a likely inability to load or resave the nodes in convential pz3 files. However for arccos/sin/tan and other linear functions it occurs to me that one thing you may want to try is using a colorramp node as a lookup table. The value input is fed into a closed spline curve and the output is the blended result of the colour values. So far as I can tell it is possible with most colour nodes to set the value outside the normal 0-1.0 RGB values, so long as you connect them to external (scalar) channels. Arctan (IIRC) goes to inifinite values at +-90 degrees so it may be an idea to output the logarithm from the lookup and feed that into a power node. I think it should be possible to get a couple of significant digits accuracy across most of the input range. Bill