Hi
Have just completed the final stage of designing a high res system for imaging galaxies and am posting this summary in case anyone else is trying the same thing – would be interesting to compare notes, particularly on cameras.
An earlier study determined that there is vanishingly small resolution gain from a bigger scope once you get above about 10 inches aperture in my expected best seeing of 2 arc seconds (there will be more photons with a big aperture though, so bigger is still better). The study also showed that around 1 arc second per pixel was the optimum scale for best resolution plus maximum sensitivity in the chosen seeing. On that basis, I investigated three system options based on 10 inch scopes. Mike Sidonio recently posted details of his fast 12” scope and Sony 694 camera combo and that is also included for interest, since it was designed for the same purpose.
Available cameras divide into two groups – 9 micron and about 5 micron pixels. These match well with 10”/f8 and 10”/f4 scopes respectively, with the following combinations all having a little less than 1 arcsec/ pixel scale:
- 10 inch f8 with Kodak 11002
- 10 inch f4 with RCC1 and Sony 694
- 10 inch f4 with Paracorr and Kodak 8300
- Mike Sidonio’s 12” f3.8 with Sony 694 (included for interest)
The analysis included a fairly complete model of the responsivities of these systems, along with shot, read and thermal noise. The operating temperature was chosen to be -10C for the Sony and -20C for the Kodak chips.
The sensitivity results are presented in terms of the single exposure time required to obtain a broadband SNR = 5 with NGC247 as the extended target (nominal surface brightness of roughly mag 23 per square arcsec ). Other source and SNR levels could be chosen, but the results would be generally similar - except for much dimmer sources. Sky brightness has been ignored, as has noise from dark/flat processing, since this will depend on technique. Camera data came from the Starlight Express website and I estimated the broadband QEs (450-700nm) from their graphs.
The results are summarised in the attached spreadsheet capture and are as follows:
- The most sensitive system is Mike’s 12” f3.8 with the 694 chip, which takes 3.2 minutes to get to an SNR of 5 on NGC247 – it also has the smallest field of view.
- This is followed closely by the 10 inch f4 with the 694 chip – it can get to an SNR of 5 in 3.5 minutes, but also has a relatively small field of view.
- The 10 inch f4 8300 system needs 5.6 minutes to the same SNR, but the field of view is significantly larger.
- The 10 inch f8 11002 system has the largest field of view, but it needs 11.7 minutes to get to an SNR of 5, due mainly to the relatively high read noise and low QE of the CCD.
- The dynamic ranges of the systems are similar and all have significant overload margins.
- The 8300 system is easily the lowest in cost, the 694 system is $1200 more expensive and the 11002 system is about double the cost of the other two.
Any one of these systems could do the job of imaging galaxies – they all have about the same angular resolution and dynamic range, although Mike’s will provide better resolution in exceptionally good seeing.
Comparing the 10 inch systems that I am interested in, the 694 f4 system wins out over the 8300 f4 on the basis that a smallish field of view is OK to capture most galaxies and the extra cost of $1200 is a small enough price to pay for the extra productivity due to the very much shorter imaging times. The extra sensitivity will also bring fainter stuff within reach before the noise floor takes over. The 11002 f8 system is much less sensitive than the alternatives in this high resolution application, but it could still be considered where a large field of view was desired or if the alignment/correction/focus advantages of working at f8 were considered to be as important as imaging efficiency.
After going through this exercise, I think that a system modelling approach provides a sound basis for decision making, since it is possible to see what parameters matter for a given application and to investigate "what if" combinations of components. I have had enough of buying gear on an educated hunch, only to find after trying it out that it is not really ideal for what I wanted to do. I will be ordering most of the bits of an f4 system next week, so will find out just how good the system model is in capturing my requirements – provided I can get the real hardware to perform at design levels (in particular am hopeful that an RCC1 does not introduce anywhere near as much SA as an MPCC).
Thanks for reading.
Regards ray