Gerald,
Sorry to bother you with this, but there seems to be a bug in the 'render-to-disk' mode (or, in my case, to XMS), which more or less clobbers the calculated image.
I'll have a look. Thanks.
This was introduced sometime between version 20.0 and the current version. The immediate fix would be to remove the float=y from the parameter entry. No, this isn't the real fix. It forces the code to initialize whatever is getting missed when the code determines that floating point is really required to generate the image. I can think of two possible changes that might have caused this. Either the change to the disk video code to use the new memory routines or the change that generates the pixel grid on-the-fly (this second is the more likely cause). Maybe Tim can help narrow this down. Jonathan