registax difficulties?

I've had a couple more attempts at pointing
my camera at the night sky.

This time (at 4x zoom) I used a 2 second exposure,
which means stars are pretty much points.

I took 5 minutes worth of 2 second exposures,
which comes out at around 118 frames (there's
some overhead in the camera recording data, so
I don't have the "obvious" 150 frames).

On stacking these frames (which are VERY dim,
since I used ISO 80, f4 2 second) the results
are disappointing.

With all those frames I was hoping that noise
would be reduced, and gray scale depth
increased.

I have found that if I pre-multiply
my data up a bit before Registax (I tried multiplying by 7)
I get much better results.

Clearly they're brighter, but there's also
far more colour depth, and hence less banding
when I edit the curves in cinepaint.

It's (I guess) something to do with
scaling, fixed point arithmetic, ranges
and quantisation, but I don't know
enough about registax to be certain
of the details.

Can anyone explain?

BugBear

bugbear wrote:
I have found that if I pre-multiply
my data up a bit before Registax (I tried multiplying by 7)
I get much better results.

Clearly they're brighter, but there's also
far more colour depth, and hence less banding
when I edit the curves in cinepaint.

Here are the samples;

If I simply stack in registax, I then have to multiply
the samples by 70 to get this (reduced to 8bit for web use)

http://i48.photobucket.com/albums/f2...h/sword_70.png

If I pre-multiply by 7, stack, and then multiply by 10 (for
a net multiplication by 70) I get this superior result:

http://i48.photobucket.com/albums/f2...sword_7_10.png

BugBear