Sh2-240
Sh2-240 or Spaghetti Nebula - Click here for full resolution
Sh2-240 (R.A.: 5h 40m 28.88s, Dec: +28º 00’ 44.3”), also known as Simeis 147 and nicknamed the spaghetti nebula, is a super nova remnant on the border of the constellations Auriga and Taurus.. It is located in our own Milky Way, at a distant of 3,000 lightyear from Earth. In the midst of the gaseous left-overs from the exploded star, is its remnant, a rapidly spinning neutron star, a pulsar that emits strong radio signals. The object is extremely faint, and only visible in H-alpha and less so in OIII. In broadband images, the object is pretty much invisible. Due to its low brightness, discovery did not happen until 1954.
Sky-plots with a FoV of 50º (left) and 10º (right). Click to enlarge
Conditions
Images were taken on during three different sessions from the backyard in Groningen, The Netherlands (53.18, 6.54). For an object this faint, multiple sessions are typically needed. Unfortunately during the December session, there was only one clear night. Also in January there was one clear night. Towards the end of February 2022, there was a prolonged period of many clear nights for several weeks. This allowed completing this image set.
Moon was almost full during the first session, but absent during the las session. Sh2-240 is best visible in the winter. The last session at the end of February was just in time for the season, with the object setting behind the visible horizon by about 03:00h.
Visibility charts showing 22:00h altitude throughout the year (left) and throughout the session on February 25, 2022 (right).
Weather was generally good, with low temperatures just above freezing point. For all other weather conditions, see below table. The nearly full moon in the first session resulted in a very low visibility, with SQM value just below 17 mag/arcsec2. In session 2, thick fog rolled in after about 1h of imaging.
Capturing
The Spaghetti Nebula is a very large target. Many people choose to image it as a mosaic to capture the whole object. To increase the field of view, a 0.72x reducer was added to the Takahashi FSQ-106. With reducer, the telescope turns into a reasonably fast wide-field telescope of f/3.6 with focal length 382mm. In combination with the ASI6200MM-Pro full-frame camera, the field of view is more than 5 degrees wide, capturing almost the whole object.
Telescope
Mount
Camera
Filters
Guiding
Accessoires
Software
Takahashi FSQ-106 + 0.72x Reducer, Sesto Senso 2
10Micron GM1000HPS, Berlebach Planet
ZWO ASI6200MM Pro, cooled to -15 ºC
Chroma 2” LRGB unmounted, ZWO EFW 7-position
Unguided
Fitlet2 (Linux 20.04), Pegasus Powerbox Advance, Flip Flat
KStars/Ekos 3.5.6, INDI Library 1.9.3, Mountwizzard4 2.1.2, openweathermap.org
The image was captured using narrowband H-alpha and OIII filters with 3nm bandpass. Sh2-240 is a very faint object, and the originally planned 5 min. subs were too short, even for this fast setup. So narrow-band exposure times were increased to 10 minutes each. The GM1000HPS mount had no problems taking such images unguided. For proper star colours, a separate small set of R, G and B images were made. Below are the frames listed taken in each session that made it to the final image.
During the first session, focus of Blue and Green captures was quite a bit off due to changing temperatures. They were repeated in the third session.
Image
The field of view of the camera/telescope combination is only just large enough to capture the full object. It was difficult to frame the subject. The outer envelope from the nebula is OIII signal, but in individual subs, almost no signal is seen. To make things even more difficult, SkySafari and KStars would center the image at slightly different points in the sky. It would have been better if the camera would have rotated a little more, to get a more horizontal orientation, making it a better fit.
After the first session, interim processing indicated that especially the OIII signal was much too weak for a good edit. So subsequent sessions were planned and a total exposure of 17.6h was achieved. A long exposure is absolutely critical for this object.
Because the images were taken at three different sessions, some small differences in camera rotation caused for some uneven edges on the frames. Normally one would crop this out, but because the object was only just fitting, no crops were applied. So the final image has the full resolution of the camera, with 9576 x 6388 pixels, or 61.1 Megapixels. It covers a field of view of 5,42 degrees horizontally.
Annotated image showing other deep sky objects, stars brighter than mag. 9 and the image’s orientation.
Processing
Processing this image was quite a challenge, for two reasons. First, the object is very faint and even with stacks of 40-50 frames, the signal only exceeds the background by a small margin. Any slight gradient left in the background was persistent throughout the processing and had to be dealt with at some point. Secondly, the images were taken on three different occasions, leading to small differences in camera rotation and frayed edges. The object required every bit of the image to be retained, so cropping around the edges was not an option.
Calibrating the frames was fairly straightforward. Flat frames were calibrated with Bias (100) and Light frames were calibrated with Dark (50) and Flat (25) frames and registered using the WeightedBatchPreprocessing script. Image frames were normalized and scaled using the NormalizeScaleGradient script and integrated using NSG parameters.
Both the H-alpha and OIII channel had to be treated with multiple background extractions to even out as much as possible of any gradient left. The first round started with a main subtracting removal using a grid of sampling points. Then several next rounds were applied of hand-picked sample points around problem areas and removed by division. Between each round a new screen stretch helped to exaggerate the differences.
Next up was noise reduction, using the efficient MMT-based method, followed by combining the two channels to an HOO image. In the wide-field image, lots and lots of stars were present. It was decided to best work on the nebulosity in a starless image and then later add the stars back in. The recent update of the StarNet plugin for PixInsight, now called StarNet2, is considered to be a major upgrade. StarNet2 does not come automatically with PixInsight, so had to be installed separately. This turned out to be more difficult than anticipated. PixInsight had to be upgraded, some library files had to be manually placed in the application directory, they had to be granted permission within the OS, etc. Over on Galactic Hunter, there is a nice step by step instruction on how to install it. Once installed it could be used, and the results were really great. The improvement over the earlier iteration is really astonishing.
Rescuing stacking artefacts
When you stack images taken at different sessions, chances are that the rotation/alignment is not 100% perfect between sessions. Add to that the effect of dithering and stacking artefacts at the edges of the final image are hard to avoid. In most cases this is not an issue. A simple crop will just cut off the rough edges. But sometimes cropping would remove important parts of the object and cannot be applied. In such cases you may want to reconstruct some of the missing information. This image of Sh2-240 is such a case. The object only barely fits within the field of view, and any cropping would cut into the object. Still the stacking artefacts around the edges were significant. For a regular photograph you might want to use something like a content-aware fill in Photoshop or so. But such an approach essentially just makes up information and that is what we’d like to avoid in Astrophotography. So instead, some careful reconstruction efforts took place to save as much of the full field of view as possible.
In most cases the best starting point is to work on the starless and starry images separately. In this case a careful use of the CloneStamp tool was able to smooth out the background all the way to the edges nicely. As the nebula was everywhere so close to the edges, the nebula was protected using a range-mask. Some leftovers from the background were clone-stamped out of this mask to really only cover the nebula. With this mask in place, the background could be restored without touching much of the signal.
Stacking artefacts in the background can be corrected by using the CloneStamp tool. See the before/after comparisons of using the CloneStamp tool in the top-left (left images) and top-right (right images) part of the image.
The stars were a bit more complicated. The start of the processing of the RGB data was straightforward. The individual R, G and B channels were prepared in the usual way. After color calibration, two stretches were applied, first a mild stretch using ArcsinhStretch to preserve colour, followed by a regular HistogramTransformation stretch. The stars had a bit of a green cast around the edges which was easily removed using SCNR. Care was taken that the background was really dark, as later this image would be added to the starless image. Under the protection of an inverse star mask, signal strength in the shadows was reduced. Normally at this stage, the star image could be added to the Starless image. But it turned out that some of the edges were only covered by the red-filter, so stars in those areas of the RGB image were completely red. Somehow I needed to replace these red stars with properly coloured ones, but how to get properly coloured stars? In this case, the narrowband channels came to the rescue. The stars-only versions were combined into an RGB image and used as the base for the RGB stars.
Star repairs in the top left corner using a mask created in Photoshop. Each other corner had similar issues.
In order to blend the stars from these two different sources, the background of the HOO-star image should be made comparable to the RGB-star image. This was done using BackgroundNeutralization, with default settings, but with a target background setting equal to an average background signal from the RGB-star image.
Also a mask was needed that would cover the slightly rotated rectangular frame. It was easiest to create this mask in Photoshop and bring it into PI as a jpeg. The mask was applied to the HOO-star image and with Pixelmath, the RGB-star image was added onto the HOO-star image. This worked remarkable well. The only way to tell the difference is that in the frayed edges that first had the red stars, the stars were now more white and less colourful than in the rest of the image. Since the edges were only small, this is hardly visible in the final image.
With some careful considerations, small reconstructions can often be made applying various tools and techniques. And while the result may not always be as good as when the data was collected better in the first place, often the errors can be made small enough to be hardly noticeable and rescue an otherwise sub-optimal image.
As mentioned before, this wide-field image contains a lot of stars. And if not careful, the stars become a bit too distractive and take focus away from the nebula. Therefore the stars were shrunk a bit using MorphologicalTransformation, under a star mask protection. A structuring element of 5 was used and two rounds of size reduction with the amount set to 0.3, dialed in a nice star reduction.
Both images could now be added together using PixelMath, creating the final image.
Processing workflow (click to enlarge)
This image has been published on Astrobin.
