Pixels and why does it matter?

[kirkster501]

^^^

Sorry for numpty sounding question, I am sure it's quite straightforward when you are in the know...

I have been playing with CCD calc and I see different scope/sensor combinations give different arc/sec per pixel and this is another parameter over and above the absolute field of view offered by that combination? So this is down to resolution of the sensor right? So what are the optimal measurements? Also, I read some folks say that the pixels with some cameras - 490EX in one thread - are too close?

I confess I have got a little confused...

I want a CCD that will work nicely on my ED80 and my RC8 - both great scopes. I intend the RC8 for individual small objects and the ED80 for wider field stuff. The 314L could do the trick - I am aware f the small sensor size. Then again, I might get a 460EX and get a manual filter wheel for a bit. After my £££ balance recovers I might get a refractor a little faster than the ED80 as well.

Appreciate some thoughts please about the pixels and noise points? Again, I am happy with the absolute field of view revealed by CCDcalc but just want to understand this pixels business a little beter please?

[IanL]

Three different but related issues you must trade off:

- Pixel scale (arcseconds per pixel): Put simply you need a pixel scale that is half the smallest detail you wish to resolve (Nyquist Theorem). Bear in mind that your resolving power will be limited by the aperture of the scope and also by seeing conditions which vary with time and location. So too small a pixel scale and you are not getting any more detail, too large and you are not getting as much as you could.

- Pixel area: Bigger pixels give you better SNR. You can bin pixels to improve SNR but it isn't quite the same as having a big pixel to start with.

- Full well capacity: Small pixels hold fewer electrons, which mean they will saturate with shorter exposures. This limits the exposure time and potentially the dynamic range on targets (so you might have to do longer and shorter exposures and HDR combine them in software, but you need a chip with decent anti-blooming capacity to do longer exposures without the saturated pixels leaking in to other pixels).

You probably want to strike the best balance you can between the first two issues and then see what your options are for the last issue.

[MartinB]

The smaller the pixel size the higher the resolution in theory. Unfortunately, atmospheric movements cause what amounts to a bit of camera shake which reduces the resolution of long exposure deep sky images so there is a limit beyond which smaller pixels don't improve resolution. Smaller pixels gather less light over a given time than larger ones and are hence less sensitive so you want a pixel size which is big enough not to be compromised by atmospheric movements (seeing) yet small enough to get the highest resolution possible. Given that the quality of the seeing varies from night to night and hour by hour it is inevitably a compromise. There is no golden sampling rate (number of arc secs per pixel) but 2-3 arc seconds will be reasonable.

Having a small refractor and a long focal length scope can work very well because you can "bin" the exposures when working at long focal length, binning groups of 4 pixels together (2x2) doubles your sampling rate and increases the sensitivity of the chip. Obviously this would lead to "under sampling" at short focal lengths i.e. not achieving the optimum resolution.

Hope that makes sense!

[swag72]

... Full well capacity: Small pixels hold fewer electrons, which mean they will saturate with shorter exposures. This limits the exposure time ....

...Smaller pixels gather less light over a given time than larger ones and are hence less sensitive so you want a pixel size which is big enough not to be compromised by atmospheric movements (seeing) yet small enough to get the highest resolution possible.

Confused !!! Isn't Ian saying that small pixels will be more sensitive as they will saturate quicker and Martin is saying that smaller pixels will saturate less and so are less sensitive? ............ Or am I reading it wrong?

[kirkster501]

Thanks. So the ED80 comes in at between 2-3" per pixel on the 314 and the 460 is not dissimilar. That's good then ?

But the RC8 comes in around 1" per pixel...? So is that good or bad? Or neither ?

[MartinB]

Sara, think of tin cans left out to catch the rain. Little tins will tend to fill up faster than the big ones (because not only are they wider but also shallower). The big ones will collect more water over a given time though. The capacity of a pixel is known as it's "full well depth" but is not related to it's sensitivity. Quantum efficiency is a measure of how efficiently pixels convert photons to signal but a small pixel with high QE may still be less sensitive than a larger pixel with a lower QE (because the large pixel is scooping up more photons even though it might be leaking some of them).

In reality the images a chip will deliver don't always match the theoretical data. For instance, my old SX H9 had a much lower full well than my QSI 532 but you wouldn't have known it when it came to stretching the image. Then you can start loosing sleep over read noise (which I think is important!), dark current and so on.

I think the best starting point for choosing a chip is deciding what field of view you want (in reality what you can afford!), and roughly matching pixel size to focal length to give a good sampling rate.

[MartinB]

Thanks. So the ED80 comes in at between 2-3" per pixel on the 314 and the 460 is not dissimilar. That's good then ?

But the RC8 comes in around 1" per pixel...? So is that good or bad? Or neither ?

With the RC8 you would normally be better off binning he 460 unless you have a night of very good seeing. This will double the arc secs per pixel. Binning the 314 will also work well except, because the 314 has fewer pixels to play with, you will be getting a pretty small finished image.

[swag72]

Thanks for that explanation Martin

[ollypenrice]

Tip top explanations from Martin.

In reality I think you can deviate by a long way from that which thoery suggests. Say your pixels are too big. You lose resolution but you gain signal. Not the end of the world! When all this talk of optimum sampling rate comes in we must remember that it's based on the idea of having an infinite number of perfect subs. When will this happen?

Olly