Well, I'm assuming that you are rendering sequential stills frame by frame instead of rendering directly through a codec. If not, then some of the noise could be coming from the compression. Likewise the image format for the stills should be as lossless as possible before combining into an animation. I use PNG format but TIFF would work too. In Poser 5 you can also increase the accuracy of frame-to-frame pixel interpretation by lowering the minimum shading rate (to at least 0.5) and increasing the pixel samples to 4 or 5. Don't be too quick to blame Poser's mathematics - it has a higher degree of mathematical precision than any other 3D program I've seen. The problem is with non-CL modules which were originally developed for other applications which had LESS precision than Poser.