Relation of seeing (arcseconds) to resolution (arcseconds per pixel)

[themos]

I have seen some statements about this around but I wonder if anyone knows of an authoritative exposition of this subject: oversampling/undersampling stars:

How many pixels should stars take up on the sensor?

[blinky]

I suppose ideally they should only take 1 pixel since they are 'point sources' of light. THe optics in telescopes not being ideal along with seeing/atmospherics will blur the star slightly.

[arad85]

Bizarrely, I was looking for an answer to almost the same question and came across this:

Auto Guiding

[brianb]

How many pixels should stars take up on the sensor?

What's the purpose of the imaging?

If you want to do high resolution astrometry then you want the scale as large as possible. For "pretty pictures" you want the stars fairly sharp, a FWHM around 3-4 pixels seems reasonable. For photometry it depends on the brightness of the objects: if you're near the instrumental threshold you need the star images to be pretty sharp in order to be able to identify them; if you've got plenty of light to play with it's better to blur them a bit so as to spread the light around rather than saturating the pixels at the core of the star image.

[Merlin66]

The Nyquist sampling theory ( we use in in spectroscopy!) really says a minimum of two pixel for a resolved image ie Full width Half Max (FWHM) image should sit on at least two pixels.

PM me if you want all the boring bits of the theory!!!!!

[dph1nm]

As far as imaging is concerned:

If you oversample (i.e. too many pixels) then you risk read-out noise per pixel becoming an issue (not to mention guiding!), hence you will lose signal-to-noise.

If you undersample (so that your pixels are bigger than the image of a star), then you get too much sky in each pixel and lose signal-to-noise that way.

Somewhere in the middle is a happy medium ...

NigelM

[roundycat]

If your stars only take up one pixel on the sensor by the time you have processed, re-sampled, smoothed and re-sized for printing or publication you won't have any stars. Nyquist doesn't tell the whole story. To get good resolution of any object you need to image at a resolution three times higher than you think. See Fourier analysis of a square wave as applied to a high contrast target in photography.

Dennis

[themos]

Image Sampling - Article

Craig Stark jumps in.