Messier 51 / Whirlpool Galaxy (OSC RGB)My apologies in advance for what turned into a rather long ramble below, as you might imagine I've had plenty of time to muse on things over the past five months whilst cursing the poor weather and waning darkness.Processing / Data QualityUnfortunately this is another instance of "image everything and just throw it at WBPP", while on paper this image is almost 40 hours of data (once WBPP rejected frames are taken out), it's not 40 hours of
good data. Generally I just image everything during nautical dark hours (if I can guide on it, I'll image it), but this time I really didn't do myself any favours. The issues with this data set run the gamut - bad flats, full moon, light pollution gradients from multiple angles, high cloud, aurora.... the works! The final stack is an absolute mess of complex gradients and while GraXpert has done an admirable job, leaving a visibly flat background, I can't help but wonder about the damage that's done to the SNR and fine detail in the process. I'm interested to try this stack on
MARS when it releases as a comparison.
A simple filter in SubframeSelector such as the below would have led to ~25% of frames being discarded, reducing the exposure time to nearer 30 hours instead. While I'm not convinced the SNR would change much, I suspect a number of those bad subs are contributing to the gradients I mentioned above.
PSFSignal > PSFSignalMax * 0.1
The ultimate
aspiration behind going for a deeper integration was to try and bring out the tidal tail that extends off the galaxy. However, due to both the aforementioned data quality issues and the following discussion on L-Pro signal loss, I've been unable to extract as much of it as I would have liked. Luckily there is a secondary advantage to having such a large number of sub-exposures, that the data will support a 2x drizzle integration with few to no artefacts. I was relatively surprised to find a real detail improvement in the 2x drizzled stack given that I'm actually pretty well sampled for my seeing, although when checking on Russell Croman's
MTF Analyzer tool it does appear to suggest I would benefit from drizzling.
I've lightly and identically processed, cropped and blinked the difference in the extracted luminance (L*) of the drizzle 1x and 2x stacks below. As you'll note, the 2x stack is noticeably sharper at the expense of some signal (both stacks were linear stretched to the same background median before cropping).
EquipmentAround half way through capturing the exposures for this image I upgraded the 200P-DS again, this time installing a CNC milled spider from
Backyard Universe. It is significantly more rigid than the rather flimsy spider that came with the scope, as a result I've seen a noticeable improvement in both collimation stability. A nice bonus is also that, because the vanes are thicker than the stock spider, the quality and quantity of diffraction spikes has also increased.
This will probably be the last image I shoot using the Optolong L-Pro, at least from my location, as I'm now fairly convinced that it's doing more harm than good. I've been doing some
very rough estimation and, based on my 533MC and an "average" spiral galaxy, I figure I'm losing nearly a third of the possible signal to the filter. To put it another way, if I'd shot this image using just a UV/IR cut filter I estimate I might only need around 25 hours of data to achieve a similar SNR.
Continuing from the
estimation I was doing above regarding SNR, I also attempted the calculate the impact of imaging in mono instead of OSC. Specifically I looked at the necessary luminance imaging time only (ignoring any extra RGB that would be required), which to match the SNR of this image would have been around 12-13 hours. Now assuming a 1:1 ratio L:RGB that would end up being a similar integration time as with the UV/IR cut filter, but crucially without any of the issues associated with bayer matrices (sampling issues, artefacts, etc). All of that was a
very long-winded way of saying - I'm looking to invest in a mono setup going forwards!
Signal-to-Noise Ratio (SNR) / Image QualityOver the last year or so I've really struggled with imaging low surface brightness galaxies from my Bortle 6 location, most notably
NGC 4236 and
M101. This has led me down a path of attempting to better understand how location and equipment choices impact on the resultant images - to this end I've started to use the
U235SNR calculator to (very roughly) investigate both past and future broadband targets.
When running my previous images through U235 I calculated the following approximate SNR values (bear in mind that U235 suggests something in the 20-30 range would be considered "good"):
While I wouldn't read too much into the exact numbers, you can see only three of my published galaxy images would fall into that "good" category, and it falls off rapidly as we move to the images of fainter objects or with reduced integration time. Also, for all of the reasons discussed above (OSC, LP Filter, Poor data habits), these numbers are probably the upper bound values.
Having this tool and information available also allows the approximation of a few other interesting statistics, for example if I wanted to "upgrade" my image of NGC 4236 to an SNR of around 20 - it would have required on the order of
400 hours exposure time with my light pollution and setup at the time - It's no wonder I struggled to process that image!
Something else I looked into was the impact of light pollution, specifically how much imaging time would I require to achieve a certain SNR with different levels of light pollution - below is a pair of charts that demonstrate this for each of the galaxies I've imaged so far:
The above values are calculated from the v-band magnitude of each object using 300s exposures, a target SNR of 20 and the following equipment:
- ZWO ASI 533MM (-10c / Gain 100)
- Sky-Watcher 200P-DS
- Starizona Nexus (0.75 Reducer)
- No Filter
I think the main take away from this (ignoring the exact values somewhat) is the exponential growth in required integration time as the level of light pollution increases.
To drive the above point home, I used some free trial credits with
Telescope Live to acquire a small amount of data for comparison purposes, the image below is made up of around 50 minutes each in R, G and B filters without any dedicated luminance. The data was from the
SPA-2-CCD telescope, a
substantially more capable instrument than my own, albeit much slower. The telescope is situated in a location with approximately 21.8 magnitude skies.
To be clear, I have only used the remote data for this example image, it is not included in my own version.
I'm not claiming this is the best possible edit for the data, if anything it's a rather lazy attempt, I simply wanted to understand the difference that "clean" data makes. This is less than 10% of the integration time of my own image, with a slower instrument to boot, and the resultant image is
broadly similar.
Weather / SeeingFinally (yes really), I would like to allow my inner Britishness to show through .... and complain about the weather!
During the last few months it has felt as if I've had far fewer clear skies than in previous years, and I've certainly managed to publish far fewer images. This feeling seems to be backed up by assorted comments I've seen, both here on Astrobin and elsewhere, from other users in and around Europe. I decided to see if I could find some kind of data to back this up, I found the
NEO Cloud Fraction (1 Month - Aqua / MODIS)data product.
I used the nearest 0.1 degree grid square to my location and extracted the cloud fraction for each month for the last few years. It does appear that this year has on average had the highest coverage, with April being especially bad and January being an improvement when compared to previous years.
Finally (for real this time), I decided to take a look at the impact of seeing on the resultant image. Because of the large number of sub-exposures I captured for this image there is quite a large range between the FWHM values for the very best and very worst frames. I took 100 each of the highest and lowest FWHM frames and stacked them separately, below I've blinked the resultant stacks. The best frames are in the range of 1.7"-2.2" and the worst 2.9"-3.7".
Comments