LEARNED IT · 11/26
One star is not an alignment
Aligning on the target alone tells the mount where it is pointing, not how far it is from level. The tracking rates that come out of a one-point model are wrong by a steady amount, and the planet drifts at a steady speed with the motors running exactly as told. Two points, well apart, or it is not an alignment.
On 3 November Saturn drifted about a pixel a second across the ZWO for an hour, with the tracking on and the motors doing what they were told to within 0.6%. Nothing was broken. The rates were wrong.
Why one point cannot do it
An alt-az mount on a tripod is never quite level. The model that turns a sky position into motor positions has a tilt in it, two small angles, and a single alignment point cannot fix them. It can only fix where the mount thinks it is pointing right now. From that one point the software predicts how fast the sky moves in each axis, and the tilt it could not measure turns into a steady error in those speeds. The target is centred when you align and walks away from then on. See [[measure-dont-derive]]: the rates were derived from a model that had not been measured.
The same night a second thing hid behind the first: near its highest a planet climbs slowly, and a floor in the software treated under 1 arcsec a second as zero and stopped the altitude axis. Two pixels a second on this camera. The floor is now 0.2.
What to do
- Align on the target, then on a star 40 to 60 degrees away. The second point gives the model its tilt, and the drift goes with it.
- A one-point model should say so. rig-ui will track on one point, and it should warn that the rates are provisional until a second point is in.
- Keep the planet stacker anyway. It finds the planet in every frame before it does anything else, which is the right thing even when the tracking is good.
This is the same mount habit as [[a-sync-cannot-reach-the-encoder-frame]]: a single correction fixes the position, not the geometry.