REaching critical sampling - need some clarification please!

[CoolhandJo (Paul)]

I finally decided to Learn/calculate what my equipment is best position for critical sampling.

So far I have established the following for the 6.4um pixel size of my DSI III Pro (and Canon 350d which has the same pixel size).

1. I am imaging wavelengths of say 680 nm avge with my ED80 at f6
2. The focul length is 480mm
3. Using the formula of aperature x (pixel um/wavelength)
4. I divide the result by a factor of 2.5 for poor seeing conditions and using a meade mount!

The most appropriate FL is then given as 595mm. This is pretty close to the FL of 480mm. I can assume then that this is now going to give me critical sampling.

I also understand that sampling at 2 arc sec per pixel is considered to be the best.

So by using the formula (205 x pixel um) /Focul length, I get 2.73 arc sec per pixel. Pretty close again!

All this makes sense until I run the formulas based on my LX200. This is where I can't resolve the numbers.

When I do they same calcs for my Lx200 at f0.63 and FL 1575mm I get critical sampling at 1860mm (Still pretty close).

HOWEVER, at that critical sampling FL its only sampling at 0.83 arc sec per Pixel!

So am I right in assuming that even though you can match the FL for critical sampling to your Chip, this doesnt mean that you will achieve 2 arc sec per pixel? And if that is correct then oversampling/undersampling can still take place? And if that too is correct then my chips based on 6.4um are only really ever going to critically sample at FL of around 500mm

[rat156]

The 2" per pixel is used as a guide only.

The theory goes...
a) Not many sites have seeing that is better than 2".
b) Nyquist theory of sampling means that you should sample every point with at least 2 pixels.
c) That means that the highest sampling rate you should consider is 1"/pixel. Probably more as your seeing will only be 2" on the better nights.
d) Match the focal length of your scope to this sampling rate. In reality most people have a scope and a CCD camera, what you get as a sampling rate is determined by the equipment you own. Of course some cameras allow you to bin the pixels, so your 6.8um become 13.6um.

In your case I would say that the 2.73"/pixel is undersampled (point sources, i.e. stars will end up as squares), and the 0.83 is oversampled. I usually run oversampled (0.87"/pixel) as I find there is more room to use deconvolution software to fix the seeing. Most deconvolution methods require the picture to be oversampled. The trade off is sensitivity and image scale. If I bin 2x2 I get a much better sampling rate in theory, but the image scale goes down. I like the bigger images.

Cheers
Stuart

[AlexN]

I wouldnt worry too much about critical sampling to tell you the truth.. as long as your sampling is better than the average seeing conditions at your location you should be right... Under sampled data is the enemy. Avoid that if possible, over sampling can be a good thing.. especially if your seeing supports it.

I too image at 480mm F/L - With my new camera I achieve 2.23 arc sec per pixel.. I would essentially like to see it down near 1 arcsec per pixel, though I fear that with such a short focal length, this is probably not going to happen...

I would say anything between 0.8 and 3.5 arcsec per pixel is going to be pretty nice at such a short focal length. My previous camera, the Orion SS pro with 7.8um pixels gave me 3.35 arcsec per pixel, and the results I got from that setup were fairly sharp, and and relatively good... This leads me to believe that at the shorter end of the spectrum, the sampling makes less difference than at the longer end...

with my ST10 + 8" RC I got 0.87" per pixel, and the results were very sharp when the seeing was good enough to support it...

[CoolhandJo (Paul)]

Thanks guys. Clears it up - The 2 arc sec per pixel is a rough guide and 0.8 up to 3 is OK. And its better to be oversampled than under etc.

Good. Thanks.

[rat156]

There's an excellent article by Craig Stark over at Cloudynights.

http://www.cloudynights.com/item.php?item_id=2042

Cheers
Stuart

[CoolhandJo (Paul)]

Thanks Stuart. Just checked it out. Yes it helps a lot. Especially when you consider FL and SNR vs Arc min optimisation.

[Bassnut (Fred)]

Download the free CCD calculator from here , and you wont have to calculate anymore, its simple and usefull .

If you undersample,(or even if you dont), doubling the image size in PS before you start processing I found gives less pixely stars and better results generally in the end (especially with the sharpen and deconvolute filters). Its sort of cheating, and guru math says its rubbish, but just through trial and error, it works. Although processing is of course slower.

[CoolhandJo (Paul)]

Fred, Thanks for the link. Yes, I read on that the other day, that if you resample up and then process, then resample down it gives an overall better "percieved" quality.

[AlexN]

Thats something I've not tired before...

Fred, have you tried doing that a different way, but drizzling the data 2x to give you the larger image, then scaling it down after processing? I would imagine it would have much the same affect on the end result... I suppose it makes sense, an image can look moderate to average at full size, cut it down to post it on the forums and it can look amazing...