Just started imaging and was looking at the Drift graph tab the image below is the first ten 3min subs I know the left graph shows me the pixel difference between the subs and the two grouping offset is due to a dither as they are around each other but separated if that makes sense.
The right hand tab is the field rotation so is showing the rotation difference up the left hand hand side but what is the horizontal scale telling me as I am unsure as to what the units are. Also what should a good graph look like in terms of the values you should see.
the right hand graph shows the rotation in degrees on the vertical axis and the horizontal axis is 'frames ago', so -4 is the frame 4 before the current one and so on.
Rotation is caused by either an alt-az mount or polar alignment error. You are seeing something like 0.03 degrees over 8 frames, or about 0.004 degrees per frame.
That rate is probably nothing to worry about - you can work out the rotation movement at the image corners by
(rotation per frame) * (distance from frame center to corner in pixels) / 57.3
So, for say an IMX571 camera with 6000x4000 frames, the distance from center to corner is about 3600 pixels, so it works out at about 0.25 pixels movement at the corner per frame. Note that the 57.3 is an approximation to 180 / pi - not just some number picked from thin air!
Somewhere I have a note to alert if the drift rate from movement or rotation exceeds 1-2 pixels per frame. One of these days I will get around to it...
Hi Robin
Thanks for that just to clarify though that 0.25 pixel in your example if it was applied to the screen shot would be 0.25 from the first image as I assume the 0.04 degrees is in relation to image ones original position so in reality I am seeing 0.025ish pixel drift frame to frame so as you say nothing to worry about and a rotation of 1-2 degrees would be an issue if that was frame to frame were as I saw 0.04 after 10 x 3 minute frames so half an hour of subs so in reality even at a 30 minute single sub it would be non visible.
yes, the figure shown is relative to the first frame in the stack, so if it is changing in a relatively straight line then you can estimate the frame to frame rotation by dividing the most recent value by the number of frames so far.
I suspect that you would likely get away with a 1 pixel movement per frame in the corners without noticing in the final image (unless you start zooming in to the pixel level). 2 or more pixels per frame would probably start to be readily noticeable.
