Drizzle vs smaller pixels

[ramv]

Hi folks,

At what point does a 2x drizzle on a camera with say 10 micron pixels start to match the resolution of a camera with smaller pixels: (7 microns or even 5 microns)?

Does your answer change depending on the "quality" of the dithering, i.e. something ranging from not dithered to perfectly random dithering. Under perfectly random dithering, how many more subs need to be captured by the camera with larger pixels?

How much noise does dithering add? Assuming equal amount of time devoted to shooting the target, will it be compensated by the lower noise levels of the camera with larger pixels? Which camera eventually has the better SNR, the camera with the smaller pixels or the camera with lower read noise, i.e. camera with larger pixels once they are dithered to the same size.

If dithering 2x can actually improve resolution by 100% or more without increasing noise by the same factor, it appears that we are getting something for free here.

Is the old saying of matching FL to pixel size irrelevant when dithering is involved? Why not undersample all the time?

I've searched long for an answer to this question but havent found any references anywhere. Mathematical or even empirical data would be much appreciated.

Cheers,
--Ram

[MortenBalling]

This is purely empirical,but:

Drizzle is great! However it was invented for undersampled images. I personally like to postprocess images oversampled/upsampled. That makes me jugde my settings more critical, like deringing etc. Therefore I try to work at a resolution at least twice (on both axis), the resolution I aim for with the final image. Before Drizzle in PI, I upsampled my calibrated lights to 200% (using a lanczos algorithnm that introduce a slight ringing) before staraligning. To be honest I find it very difficult to tell the results of those two methods apart, even though Drizzle is a far more complicated process. Nasa don't use simple cowboy tricks like mine 

Undersampling doesn't rely on pixelsize alone, but on a combination of pixelsize and FL. It's the final image pixel scale that counts. Using large pixels with a short refractor will quickly get you in the undersampled territory, but with a wide field.

Until recently I've used an older Orion Starshoot Pro, which have pretty large pixels (7.8 microns). The chip is cooled, and have a reasonable SNR, and with an ED80 refractor reduced to F/5.6 the image was undersampled @3.62"/pixel. Perfectly suited for Drizzle.

Then I bought an Atik 428M and even though it has much smaller pixels (4.5 microns), it produce much better results. First of all, the microlenses seem to do their job pretty good these days, so SNR is better with the same amount of exposure time, as with the Orion camera:



There are a lot of different factors at work here, like mono/osc etc, but my gut feeling is that pixel size isn't the only one parameter worth considering.

I live in a city center, so seeing is horrible here. I also have a C8 sct and even reduced @0.66x it still gives me a resolution of 0.69"/pixel. Combined with the awful seeing, that makes an oversampled image on most nights. Then Drizzle doesn't really make much of a difference.

Cs

Morten

Btw. do you know this one?

http://www.12dstring.me.uk/fov.htm