SharpCap Forums

This post is the result of an Argentine/US/UK collaboration. Many thanks to Leandro (@leandro, Argentine) for capturing the data and to Brian (@oopfan, US) for developing the methodology to use Robin’s Sky Background Electron Rate Calculator with the Antlia Quad Band Filter.

This image is a combination of the R, G, B channels extracted from the captured data.

Data: Lights 96x75s, gain=562, black_level=70; bias 50x1s; flats 50x1s; darks 20x75s; total integration time=2h. Captured 5th December 2024 under a 30% waning moon.
Equipment: Montura NQH 3 Orion GoTo System; Telescope Sky Watcher 200 x 1000 Reflector Newton; Filter Antlia Quad Band 1,25"; Camera Altair Astro 183C Pro; Guiding SkyWatcher Evo 50 x 240 mm + camera Altair 130 M + PHD2; Focus by Bahtinov Mask and SharpCap Tools
Software: SharpCap 4.1, Astro Pixel Processor v2.0.0.0-beta29, GIMP
Location: Argentine, Latitude 24°S, altitude 1287m, Bortle 6.5

Processing
Astro Pixel Processor was used to extract the R, G, B, Ha and OIII channels during the calibration/integration process. The following processing steps were taken

• Combine RGB

• Crop

• Remove Light Pollution

• Calibrate Star Colours

• Sharpening, contrast, saturation, background neutralisation applied

• Final adjustments in GIMP

Data was also captured on 15th November 2024 under a ~98% waxing moon in close proximity to the nebula. The capture was over 2 nights using gain=177 and exposure=120s giving a total integration time of ~6h. Processed as above with APP but with use of HSL Selective Colour.

The extracted Ha, OIII, R, G, B channels were processed in PixInsight to produce a HOO + RGB combination using the NBRGBCombination (Narrow Band plus RGB) script. HDRMultiscaleTransform was used to compress the Ha, OIII and RGB images before combining. This was to avoid over saturation of the core of the nebula.

Methodology for calculating the exposures to account for the Antlia Quad band filter.
Thanks to Brian (@oopfan) for developing/providing this.

The above shows the transmission graph for the Antlia quad band filter. The right hand peak represesnts the IR bandwidth. Because the Altair 183C camera has an IR filter on the sensor window, the IR bandwidth is blocked and was not included in the transmission calculations.
The steps are:

• Estimate the total bandpass of the filter

• Calculate the Sky Electron Rate Sky Background Calculator modified to account for the filter.https://tools.sharpcap.co.uk/

• Use the sensor analysis results to decide on read noise to be used in the minimum exposure formula (Read Noise) * (Read Noise) * (Factor) / (Sky Electron Rate).

• Build the spreadsheet using rounded minimum exposures.

Estimate the total bandpass of the filter

• The transmission graph image was loaded into PixInsight which was used to measure the pixels between 400nm and 700nm (bandwidth of interest is 400 – 700nm.

• The Scale = 300nm/(X2 – X1). X2 was read off as 546 pixels. X1 was estimated in the same way.

• The values of A, B, C, D, E, F (pixels) were read off from the 50% transmission line. G and H were not used as the IR is blocked by the camera sensor window IR filter.

• The formula used was Total Bandpass = [(B - A) + (D – C) + (F – E)] * Scale

• The result was Total Bandpass = 83nm. This was the value used in the Sky calculator.

Calculate Sky Electron Rate
To calculate SER for a COLOR sensor using an LP filter:

• Select MONO sensor on Robin's calculator.

• Select CUSTOM filter.

• Type in the filter bandpass.

• Note that Robin's calculated SER is 3x larger than it will be for a COLOR sensor.

• Therefore, for a COLOR sensor, Exposure = RN^2 / (SER / 3) * 10.

• For a MONO sensor, then Exposure = RN^2 / SER * 10.

Apply the Minimum Exposure formula [(Read Noise) * (Read Noise) * (Factor) / (Sky Electron Rate)]. A Factor of 10 was used.

Read Noise was decided by looking at gain and full well, the decision was to use a low gain of 177 (brighter objects) and a high gain of 562 (fainter objects).

This was the resultant spreadsheet. For low gain (177) an exposure of 120s (rounded) was calculated, for high gain (562) an exposure of 75s (rounded) was calculated.

Further testing will be carried out against targets of varying brightness - to capture the best quality image in the least amount of time and by acquiring better quality data reduce the amount of post processing required.

Dave, Leandro and Brian

This is a 'no-stars' version of the above HOO_plus_RGB created with Starnet++

Dave

I thought I’d use this opportunity to explain my motivation for what you see in Dave’s post. I’m sure you are wondering why are we making it so complex? Why not just use The Brain to measure the Sky Background and have SharpCap calculate the Minimum Exposure? I’m to blame.

Over the summer I bought a Seestar S50. I realized very quickly that I needed SharpCap to help me with exposure determination. I imaged the heart of The Heart Nebula with the built-in dual-band light pollution filter. (The filter passes Ha and OIII for a total bandpass of 45nm.) I chose a 20-second exposure based on experience. (I really only have a choice of 10s and 20s. Thirty seconds is available but it is useless because of the high rejection rate due to field rotation with the Alt-Az mount.)

I was very happy with the 3-hour stack but I was expecting to see OIII. I knew it was there based on AstroBin images. What happened? OIII is fainter than Ha in this nebula. As I learned later I could have picked it up if I increased the exposure to 60 seconds. That was the Minimum Exposure for the Seestar and my skies: Bortle 5.3 at the zenith.

My punishment for failing to image at the Minimum Exposure was severe noise where the OIII should be. No de-noising filter worked but I did find one algorithm in Siril that was specifically designed to combat photon-starvation. All of the other algorithms adroitly mitigated Gaussian Noise. However, what I was dealing with is Poisson Noise as a result of photon-starvation when the exposure is below the minimum.

I regularly communicate with Dave. He liked that the math was working and suggested he try the same. One thing he discovered is he needed to increase his exposure from 10-minutes to 20-minutes using a 7nm narrowband filter of M27 with the Atik 314E CCD camera. What a phenomenal improvement!

In my opinion having a spreadsheet with the math programmed in removes a lot of the pressure when you are under stars, polar aligned, and ready for imaging. I’d rather know ahead of time what I’m getting into.

By the way, I did buy this to help with LP measurement:
http://unihedron.com/projects/sqm-l/

I came across https://www.starfixer.org/ on my internet travels. An AI application to fix star elongation, perform star reduction and denoise. Free of charge and easy to use. Top image is the original, bottom image is processed (took ~2m to process).


Dave

Hi folks,

great collaboration and nice posts explaining what is going on and the reasons that made you do things that way - I always enjoy reading this sort of thread

Dave, could you share a higher res/larger copy of the star fixer comparison? I'm struggling to see any marked difference in the fairly low resolution version included above.

cheers,

Robin

Thanks Robin. Higher res as requested. I see star reduction (useful in a narrowband image) but no loss of background nebula.

Dave

Hi Dave,

yes, those images make it much clearer what differences there are between the two versions, and it's very noticeable when flicking between them that the stars (particularly in the middle of the image) are much tighter in the processed version. Thanks for that and sharing the link to starfixer, which looks like a useful resource

cheers,

Robin

At least to my eye, Starfixer gives a 3D look to the nebula by pushing back the stars. I like it.

Robin

I tried the tool on my Double Cluster - not a good idea. It works well with the right sort of image - big nebula or narrowband.

Dave

Dave,

I hope you don't mind me showing off Starfixer. Here is a before & after shot of my heart of The Heart Nebula taken with my Seestar S50. It is a stack of 345x20s subs taken with the internal LP filter at Bortle 5.3. Total integration 115 minutes. Processing thanks to Siril. Note that the 20s exposure is far below the 60s Minimum Exposure which caused the loss of much of the OIII. Thankfully Siril has a denoising algorithm that helped tame the Poisson Noise due to photon-starvation (left image). The right image is the result of applying Starfixer to it. Processing time in the Cloud about 1.5 minutes.
Brian