In at the deep end – An introduction to imaging with a peculiar galaxy ARP 227 and a 1m telescope (2/2)
By: Mark Seaton, Flamsteed Astronomical Society
Problems with the data
I very nearly missed the deadline! I knew I needed to leave it as late as possible as the target was getting higher the longer I left it, but I almost forgot all about it and it took a gentle prod from Dirk to get my request in, and LCO only just managed to complete the observations in the nick of time – phew!
I was very excited to have a look at what I had captured but that quickly turned to horror – what had happened to my Luminance data!?! In addition to Dirk’s help, I had contacted Vladimir Vlaisavljević (AKA Vlaiv), a very knowledgeable member of Stargazers Lounge online forum, for advice. Vlaiv suggested to me that given the limitations of one hour of data, even with a huge 1m telescope, I should use at least half the time for luminance data. Luminance captures all the colours at once so is more sensitive and better at increasing the signal to noise ratio and should reveal more of the fainter shells. This sounded like, and should have been, a good idea. I would not have had many colour subs though if I had done that or they would have been too short so I compromised, taking three subs of 280s for each colour filter and three luminance subs of 200s. This comes to 3120 seconds, which is not one hour, the reason being that there is a time overhead for each exposure to allow for the downloading and calibration and this was almost exactly one hours’ worth of telescope time.
Dirk had told me I should take at least 3 subs per filter to minimise the chance of satellite trails and cosmic ray hits which are a real problem for these observatory telescopes at their high altitudes. Unfortunately, the LCO telescopes are not designed for pretty pictures as Dirk repeatedly pointed out to me – they are for Science! Because of this, they do not have a proper Luminance filter. A luminance filter lets through only the visible wavelengths and cuts out the UV and IR that can cause problems. What the LCO telescopes have instead is a clear glass filter, or what they call “Air”, which lets through everything. This is only intended for calibration and experimental purposes, and as far as I could see in the archives, never for what I wanted to use it for. This proved to be a very bad move. All my “luminance” data came out with a peculiar swirling ripple pattern. Dirk called this “fringing” which is a common issue with professional CCDs apparently. It is likely a result of IR light bouncing around within the glass of the filter itself and constructively/destructively interfering with itself to produce this weird pattern. Interesting and quite attractive in a way, but not very helpful for my image!
What to do? I could not really use my Luminance data as it came out, interesting though it was, so it looked like I had wasted a quarter of my allotted integration time. At Dirk’s suggestion I spent a good deal of time searching the archive of the LCO looking for other subs taken with the same camera and the “air” filter to try and rescue the luminance data. I did find some and sure enough, they had the same pattern as mine. Well it was roughly the same but not exact, seeming to vary with exposure length. Under Dirk’s helpful instruction, I attempted to create a kind of “flat” correction image by use of median stacking these subs. This could then be used to subtract the fringing pattern from my data. For this to be successful, it would need at least 15 unique subs taken of different targets so that the stars in the fields would be lost in the stacking process, but I could only find four or five. The result did not really flatten the stars sufficiently and though it did work to an extent, it was not even enough and had residual star shadows, which came through as dark blotches, so did not really work. However, it was a very interesting procedure to learn about and fun to try.
Getting help from Vlaiv on SGL
I reached out to Vlaiv, as he is always tirelessly helpful and very clever with all things astrophotography related on the Stargazers Lounge online forum. He gave me a lot of advice and I experimented on my data for many hours using his suggestions. He immediately suggested to “split bin” the data by 3x as it was by his calculations very oversampled. Oversampling creates higher resolution images of blur, which is not very useful. To do this required downloading and learning to use a piece of free software called Image J (I downloaded a build of it called Fiji). This seems to be an incredibly powerful tool developed for microscopy but equally applicable to telescope images. I think it can do a lot of the pixel manipulation that expensive dedicated astrophotography programs can do, though with a more hands-on approach.
Vlaiv gave me a macro and clear instructions and it was quite straightforward to load all the subs and then to create 9x as many sub-subs by using only one pixel from a square of 3×3 original pixels, with 1/9 the resolution. This would give me 9x as many subs to bin by stacking which would help with noise averaging (27 sub-subs instead of the original 3 for each colour filter) whilst still giving me a respectable ~1arcsec per pixel resolution. This worked quite well. The image had improved contrast, which helped to bring the shells out more. However, the noise did not improve as much as I had hoped. Maybe because of the lower resolution the noise grain was more evident, and the bright centre of the main galaxy looked a bit blocky.
Blooming: The bright stars in the image suffered severe blooming and even worse, different amounts for different colours caused ugly jagged multi-coloured spikes. Blooming happens when individual CCD cells overflow with charge, which then spills over into adjacent cells in a column. Because they lower the quantum efficiency and increase non-linearity, scientific CCD cameras do not use the anti-bloom gates commonly found in consumer grade astro CCDs which can drain this excess charge to ground, so this is unavoidable. I could have touched it up in Gimp but after trying that, I think I prefer to leave it alone and just accept it.
Banding: The four quadrants of the sensor are not perfectly matched in sensitivity giving rise to a noticeable 2×2 tiled image when extremely stretched. The banding reduction tool in Siril dealt with this quite well.
Dust motes: Several dust donuts were not removed by the LCO calibration processes and are quite prominent. These could probably have been touched up in Gimp but I chose to leave them.
Dirt or pixel defects in the sensor: These manifest as blue marks in the composite image and I could not find a way to remove them. Looking at the individual fits files I could see that whatever caused them was present in all filters so probably on the sensor itself but affected each filter differently. They blocked the red light completely appearing as black flecks, but in the blue and green, they appeared as lighter grey marks leading to the blue flecks in the final image. I could probably blend them in by hand in Gimp.
Reflections: A large and prominent reflection of the telescope’s obstructed aperture dominates the top left of the image caused by the bright star (HD 7991). This was particularly prominent in the red channel. I did not like this at first and tried to minimise it but came to accept and even quite like it – it makes the image interesting in a similar way that lens flares can in stills photography. Interestingly this was far less prominent when I tried combining the Luminance data but that had too many other problems
Gradients: The background gradients were all over the place and not smooth linear ones but with a quite severe band across the top of the frame and circular gradients evident in the green channel for some reason. The target wasn’t that low and checking Sky Safari the moon was below the horizon, however the Sun was about to rise so that may have been the cause. I also suspect there may be an issue with the telescope camera as this was also seen in other subs from the same telescope in the LCO archive. The gradient reduction tool in Siril (background extract) did a good job at reducing this but it was impossible to fully remove other than by really darkening the background.
Noise: With only 45mins of useable data, there is a lot of grainy noise in the image, which is emphasised by the stretching needed to bring out the faint shells. I tried to smooth it with the median filter but that also smooths all the detail. I think more data would definitely have helped here.
Dirk suggested a technique, which was the most advanced thing I tried with my data. His suggestion was to use smoothing to achieve two complimentary effects. Using a median filter to smooth the image would smooth out the noise in the background but at the same time would sharpen edge transitions, which should bring the fainter shell details out. Of course, this would only help where there were edge details. The trouble is applying the median filter across the whole image would also blur the stars and other sharper details so the technique required first removing the stars, then smoothing the remaining background image and finally replacing the stars. I was able to perform this surgery on my image using Image J and thresholds. This showed promise but I felt it gave an artificial feel to the result. It was probably a bit too much for my novice abilities, but I will play with it some more and it was interesting and educational to try.
This was a great opportunity and I have learnt a huge amount and had a lot of fun. I ended up with an image that is… interesting. Some of the distinctive shells around NGC474 have come through but I wish there were more. NGC470, the smaller bluish spiral galaxy, came out quite well as it is a lot brighter and there is a fair amount of detail in there. Very little of the other smaller shell galaxy (NGC467) came through as it is obliterated by the large aperture reflection, but I think the reflection adds something interesting to the image. There are several other smaller galaxies in the image too. It is far from a perfect image but I do like it for all its flaws – they make for an accidental “polaroid” like quality and are part of the image really.
It is important to learn how to use tools properly so that they can serve you well and it is all too easy to blame the tools for poor results. I went into this with no knowledge at all about imaging or processing and I made many mistakes. Dirk gave me lots of advice throughout and I probably should have listened better! If I were to get the opportunity again, I have learnt so much that I think I would have much more chance of getting a better result. I would certainly choose a more appropriate target and forget about using Luminance, as it does not seem to work without a proper luminance filter. Something closer to Zenith where the atmosphere is thinner to maximise the potential for the resolution of such a large aperture. I would probably chose a globular cluster as they are a favourite of mine and a challenge to do well, and the light gathering of this scope should be able to go deep while keeping the target large in the view.
I have enjoyed the whole project so much that I have now purchased a telescope purely for imaging and have been quite intensively experimenting with it and tweaking it with the hope that soon I will be able to take my own good quality images and to contribute my own data to the HOYS project. But this is an ongoing project and I am not quite happy with the results yet.
Thanks to HOYS for this amazing opportunity, and Dirk for your tireless help and for answering all my daft questions, and LCO for letting me play with their big telescope – it was a lot of fun! In addition, thanks to Vlaiv for being so generous with your time and teaching me new things.