Mono M13

[Ags]

This is my latest attempt at lucky imaging (or lucky tracking) with my C6 and AZ-GTi. This is 48 one-minute sequences of 400 millisecond frames. Compared to previous efforts I have added a 6.3 reducer and changed from 2x2 binning to no binning.  Each sequence is processed in AS!3, the the TIFs from each sequence is stacked in DSS, and finished in Gimp. I retained 80% of the frames in each sequence, not only was the night calm and windless, but also the reduced focal length made a big difference I expect.

I think the principle of this approach to imaging works, but I need to resolve the problems I have with darks and flats - the flats don't flatten and the darks leave traces of amp glow. Also processing is too laborious - I would like to shoot longer sequences than one minute so need to get a wedge and shoot in equatorial mode.

Edited May 22, 2020 by Ags

[Ags]

On the subject of binning, does it ever make sense to bin? Why not capture at full resolution and then do more sophisticated stacking of pixels. You can downsample at the end. Binning seems to force small stacks of 4 subs (assuming 2x2 binning) which can't be great mathematically...? I had to bin at F10 because the signal was so attenuated... but in general am I right in thinking that binning is best avoided if at all possible?

[Martin Meredith]

That's a really nice result in terms of tight stars and good resolution, so I think your approach does work.

My feeling about binning is that it is nearly always better to do it in software. The only benefit from doing it on the sensor is to save a little noise -- how little depending on the sensor. Although it is almost never mentioned, on-sensor binning also introduces aliasing noise, because the energy from those high spatial frequencies that are present in the unbinned image has to go somewhere, and ends up being aliased at lower spatial frequencies, thereby contributing spurious photon counts i.e. noise, although I've seen no studies of whether it makes a noticeable contribution or not. One advantage of downsampling in software is that one can apply a spatial lowpass filter to remove aliasing. Another advantage is that one can downsample by non-integer factors, which permits precise matching of the resulting reduced resolution to seeing, giving the best possible noise reduction while not oversampling relative to seeing.

I'd be interested to hear other opinions!

Martin

Edited May 22, 2020 by Martin Meredith

[Ags]

Thanks. Initially I was making the mistake of thinking I would get a brighter image by binning, it took a while to realize the image would stay the same brightness but with a bit less noise. A major reason to bin is the size of my hard drive - the above image is 80GB of SER files! I only have 150GB free space.

In another question, does flooding an SSD with data every week reduce its lifespan?

Edited May 22, 2020 by Ags

[Ags]

I found the problem with my darks. I was applying darks and also removing horizontal banding in AS!3. Turns out the AS!3 horizontal banding feature results in a very slight gradient.

[Ags]

Here is a new version reprocessed from scratch without using the AS!3 banding removal feature.

[don4l]

On 22/05/2020 at 13:02, Ags said:

On the subject of binning, does it ever make sense to bin? Why not capture at full resolution and then do more sophisticated stacking of pixels. You can downsample at the end. Binning seems to force small stacks of 4 subs (assuming 2x2 binning) which can't be great mathematically...? I had to bin at F10 because the signal was so attenuated... but in general am I right in thinking that binning is best avoided if at all possible?

I've spent the last three months imaging at 2000mm F/L instead of my usual 500mm, and I have (slowly) been wrestling with this thorny subject.

My last image (Whirlpool Galaxy) was taken with the L binned 2 x 2, and the RG and B at 4 x 4.  I have to say that selecting the "4 x 4" binning option really, really went against the grain.  However, it was the only way that I could gather useful data.

The problem for me was that all the images that were unbinned turned out "blurry", and had to be scaled down before viewing anyway.  They just looked rubbish at full resolution.   Increasing the exposure time was just producing smoother looking blurriness!   I think that this is why people say that there isn't much point imaging at less than 1.6" resolution.

Like you, I thought that there was nothing to lose by software binning if necessary.  This didn't seem to work very well - and I don't really understand why.  If I collect my R G and B at 4 x 4 (which is still 2.4" per pixel), then I get a useable image in a reasonable amount of time.  If I try the same thing binned 1 x 1, then I never get anything useable.

Perhaps I should point out that my camera is CCD.  Maybe the answers are different for CMOS. 

BTW,  Nice image, and really nice to see different techniques in action.