Quote:
Originally Posted by Ryderscope
There is one aspect to the discussions on FWHMs that are achieved/measured during imaging sessions that I would like to throw out there for a discussion. That is the relationship between:
- the image scale of the system in use; and
- what can be expected at a particular site; and
- what can be, or is being, measured during a particular imaging session.
My conclusion is that it is inconclusive to cite a particular achievement in seeing as measured by FWHM unless one also cites the pixel scale of the optical system being used to capture the data. For the same reason, it is inconclusive to form a view that a particular site is only capable of a particular seeing as measured by the FWHM unless one takes into account the specifications of the optical system being used for the measurement. My reasoning is this...
Based on the Nyquist sampling theorem, in a ideal world we would design our imaging system to be 1/3 of the target FWHM we would like to be able to image. For example, if we expected a particular site to be able to achieve a 1.5" seeing then we would design our system to have an image scale of .5". Does that mean that we can only ever achieve an FWHM with this system of 1.5" or greater? No - of course not. On a good night we can do better than that but based on the Nyquist theorem, once we get below the 1/3 sampling the accuracy of the FWHM figure we are measuring become less accurate. Yes, we can definitely see the improvement in the star quality in the sub if the seeing improves so all good.
In my case I have two systems:
- a TSA120/QSI683 for an image scale of 1.64"; and
- a 200mm Canon lens/ASI6200 for an image scale of 3.9".
According to the Sub Frame Selector tool in Pixinsight, I have measured FWHM figures for the TSA120/QSI down to as low as 2.15". For the Canon 200mm lens/ASI Camera combination I have measured FWHM figures down to as low as 6.85".
Clearly, the seeing at my location is better than that reported by the 200mm lens/ASI camera combination. Is it better than that reported by the TSA120/QSI camera? I have no way of knowing unless I set up a system in the order of a .5" image scale and start imaging with that. Another way of stating this is that it would incorrect for me to assume that the best seeing that I can get is 2.15" as I am below the 1/3 sampling rate and somewhat under sampled which will affect the accuracy of the measurement anyway.
So again, my conclusion is that "seeing" conditions as reported using FWHM figures should always cite the image scale for the optical system in use.
Unless of course, one is using a DIMM (Differential Image Motion Monitor), to measure seeing and that should be stated too  .
I would appreciate the input of others into this discussion.
Clear skies,
Rodney |
I think you are right Rodney. The measurement of seeing as a function of the FWHM is impacted by the image scale/ pixel resolution and thus the sampling. It only stands to reason that if your stars are spread over too few pixels, the measurement of the spread of the illumination is going to be compromised.
I was able to show this well with some work I did whilst imaging the Antennae Galaxies. I have an image scale of 0.5 arc secs per pixel. If you accept that a 3x multiplier is sufficient sampling to get a good estimate of the FWHM/ seeing, my system should give decent estimates when the seeing is 1.5 arc secs or greater.
I measured the FWHM using the PixInsight Subframe Selector algorithm for a range of images with FWHM values from 1.7 arc secs to 4.1 arc secs. I took these images and then used the Photoshop function to Filter>Pixelate>Mosaic to increase the size of the pixels and therefore effectively down sample. This is a good method to simulate the effect of using a camera with a larger image scale. If you double the pixel size, the software takes 4 pixels and averages the result to create the equivalent of 1 larger pixel, exactly as a camera with larger pixels would do.
I then remeasured the FWHM values. in all cases the estimates of the FWHM values went up as I down sampled, which proved that inadequate sampling gives inaccurate FWHM and therefore seeing estimates. See the graph and here is a DropBox link to it if you can't see the attachment:
https://www.dropbox.com/s/rkcvrf7rfl...eeing.JPG?dl=0
What I found interesting was the measurement when the seeing was poor at 4.1 arc secs. If a Nyquist multiplier of 3 is sufficient sampling to make a good estimate of the FWHM/ seeing then sampling of 4/3 = 1.33 arc secs per pixel ought to be good enough. Therefore the estimates for seeing at 1.4 arc secs when the sampling was 0.47 arc secs and 0.94 arc secs ought to give comparable numbers.....and they do!! The graph is quite flat between these two figures, but once the image scale gets up to 1.41 arc secs per pixel and higher the FWHM figures go up.
Cheers, Niall