[This could be of some interest as it shows how a ‘flat frame’ was produced from the ‘light frames’ to correct the vignetting that would be expected when using a refractor with a full frame camera.]
There are, in principle, two problems. The majority of refractors are now equipped with 2 inch focusers and these will almost certainly badly vignette (darken the corners) the image if a full frame camera is used − which is why I have always used cameras with APS-C sensors to couple with refractors except when I have used an astrograph specifically designed for use with a full frame sensor. The second problem is that of field curvature as, without correction, outer parts of the image will tend to be out of focus towards the extremes of a full frame sensor. Field flatteners can correct for this but their position prior to the camera may well make the vignetting worse.
I thought that it might be instructive to see how these problems manifested themselves so set up an imaging exercise on a night of poor transparency. The target was the ‘Leo Triplet’ of galaxies seen towards the south in April. I was also interested in the tracking performance of my mount so I could ascertain the length of exposures that could be used when unguided. [For reasons outlined in the last section of the article, I do not want to autoguide when using a DSLR or mirrorless camera.]
A Sony α7S, 12 megapixel, full frame camera was mounted onto a 127mm, f/7, refractor requiring (as usual for refractors) the use of a 2 inch barrel extender to gain focus. The α7S has amazing low light performance so one can observe stars in live view and use ‘focus peaking’ to bring them to focus. Whilst it was still quite light, I sent the mount to Regulus so I could synchronise the mount’s pointing in order to increase its ‘goto’ performance in that region of the sky. Very pleasingly, Regulus appeared in the live view screen quite close to the centre of the frame. Having synchronised the mount I selected M65 and an immediate problem arose: the mount did a meridian flip, so drove right around the sky and I could not really expect the M65 to be centred in the frame. As it got darker, the stars in the frame became visible but, unsurprisingly, not the galaxies. These did, however, show up when short 5 second exposures were made and were towards the left of the frame. I should, perhaps, have centralised them in the frame but, later, I was pleased that I had not done so.
The α7S has an internal ‘gain change’ to reduce the sensor’s ‘read noise’ when ISO’s greater than 2,000 are used and so ISO 3,200 is quite a good choice for astroimaging. As I had yet to learn about the tracking performance of the mount I set a short exposure time of 10 seconds saving both Jpeg and raw files. ‘Long Exposure Noise Reduction’ was turned off as it would have halved the time taken imaging the sky and also added some noise into each light frame. The Sony (and Nikon) sensors have such low read noise that dark frames are not really needed – and it would be a problem using an additional set of dark frames as the sensor temperature would vary during the hour long imaging period. A total of 319 frames were taken using an intervalometer to control the camera giving a total exposure of ~52 minutes. Scanning quickly through the Jpg files, two frames were rejected as satellites (one very bright) had passed through – one reason for taking Jpegs even if raw files were to be used for stacking and alignment.
The frames were processed in Sequator (see article relating to Deep Sky Stacker and Sequator in the Digest) and the output exported as a 16-bit Tiff file for processing in Adobe Photoshop. [The excellent, low cost, alternative Affinity Photo could just as easily have been used.]
The stars were easily visible within the frame along with hints of the Leo Triplet of galaxies and, whilst not expected, a faint galaxy, NGC 3953 over to the right of the frame. It is shown on the superb (and almost certainly the best) star atlas, Interstellarum.
Aligning and stacking – first producing a ‘flat frame’ from the ‘light frames’
The 317 raw files produced by the α7S were loaded into Sequator for the first iteration of the alignment and stacking procedure. Sequator produced a 16-bit Tiff file which was loaded into Adobe Photoshop and brightened using the ‘Brightness & Contrast’ tool until the sky background was prominent as shown below. The positions of the galaxies were just visible.
The vignetting that I had expected was present though not as prominent as I had expected. The fact that light pollution was present made it possible to produce a ‘flat frame’ which could then be used in a second iteration of the aligning and stacking process to enable it to be corrected.
To this image was applied the ‘Dust & Scratches’ filter with a radius of 40 pixels. The stars are removed but some evidence of the galaxies remained. These areas were cloned out from adjacent areas and an application of the ‘Gaussian Blur’ filter made with a similar radius. A very smooth image thus resulted which showed the vignetting towards the corners of the frame.
This process has actually made an excellent flat frame.
A second pass of the alignment and stacking process – but now including the flat frame
The alignment and stacking process was repeated but, this time, the flat frame was added into the process. When the result was brightened using the ‘Brightness & Contrast’ tool, it was obvious that the vignetting had been very largely eliminated. Excellent.
This image was duplicated and, on the top layer, the same process was used to remove the stars and galaxies to produce a smooth image of the sky background. The two layers were then flattened using the ‘Difference’ blending mode. The result was an image of the stars and galaxies with the light pollution removed.
Stretching the image
The image was stretched using four applications of the curves tool: the lower part of the curve was lifted up so that the fainter parts of the image were brightened relative to the brighter parts.
As discussed below, hot pixels had left some faint streaks within the image. These could easily be removed by lifting the black level (the left hand slider in the ‘Levels’ tool). But this would also affect the galaxies, so these were first selected and the selection inverted so that the levels command would not be applied to the galaxies. With the selection inverted again so that only the galaxies were selected, their contrast was enhanced using the ‘Smart Sharpen’ filter with a large radius and small amount (this helped bring out the dust lane in NGC 3268) and their noise was reduced with a very light application of the ‘Gaussian Blur’ filter to give the final image.
NGC 3628 is the upper galaxy of the triplet with M66 and M65 (on the right) below. The fact that I had not centred M65 had, pleasingly, allowed the faint galaxy NGC 3593 to be present in the frame.
The Mach-1 Tracking Performance.
From the fact that Regulus had appeared close to the centre of the frame, I knew that the mount’s alignment on the North Celestial Pole was quite accurate. This had been set some months before using the QHY Polemaster. (See article in the Digest.) I do not want the polar alignment to be perfect as I want the sensor to move a relatively short amount across the sensor during the imaging period. This is to eliminate what Tony Hallas has called ‘Color Mottling’ (he is an American) – the variation in pixel sensitivity on scales of 10 to 20 pixels which can give rise to a coloured rather than a grey background to an image. This movement will obviously give a limitation as to how long exposures can be without giving rise to ‘star trailing’. The result of the small misalignment will cause a single hot pixel to give a ‘trail’ across the aligned and stacked image as seen below. A useful result is that the ‘light’ from the hot pixel is spread out across the image and will thus be far less bright than if the sensor was perfectly aligned throughout the exposure period. An adjustment of the ‘black point’ of the image during the image processing will then often eliminate these pixel trails without loosing the faintest stars. [If there are not too many, one can simply paint them out so no very faint stars would be lost.]
The blue pixel trail was 122 pixels long over an exposure period of 40 minutes; which indicates a movement across the sensor of ~0.5 pixels during the 10 second exposures, so star trailing would not be apparent and longer exposures would have been possible. [With the α7S, I keep exposures shorter than 30 seconds to avoid the Sony ‘Star Eater’ problem. A noise reduction algorithm is applied to ‘bulb’ exposures of longer than 30 seconds which can eliminate very faint stars.] The near linearity of the track shows that the mount tracks well so this aspect of the imaging exercise gave very good results.