It’s amazing what you can see if you stare long enough.

I’m lucky to have fairly decent skies at our home. They’re not world-class skies, but they’re not horrible either. Once leaving the nearest “road to town,” you’d drive about 5 miles on winding country roads to find our house. There are no streetlights in our neighborhood or the surrounding areas. Still, we’re close enough to Memphis and a few suburban towns that there are light domes to the south, west and north. The eastern sky through about 45 degrees past zenith to the west is my prime imaging area, so that’s where I concentrate. I’d call it a Bortle 5 sky and my FreeDSM sky quality meter seems to agree.

I’d love to image the Horsehead and Flame with a monochrome astronomy camera coupled with a full set of narrowband & broadband filters. That’s not the gear I have though. My astronomy camera is an SBIG STF-8300C. As best I can figure, its Kodak KAF-8300 CCD sensor was released in 2005. Yeah, 20 years ago. For a while, cameras with the KAF-8300 were all the rage in the semi-affordable range of dedicated astronomy cameras. I bought this camera new in 2012 during SBIG’s second iteration of the camera. While the STF-8300C was drool worthy back then, it’s not without limitations. Its sensor is slightly smaller than an APS-C sensor, which limits the image field of view somewhat. It has an 8.3 megapixel sensor, which means quite a bit less pixels than is common now.

The biggest two limitations for this camera for an image like this are amp glow and noise. While you can find lots of posts online with people saying amp glow isn’t a big deal (“flats will fix it”), the amp glow does impact an image project like this where I’m trying to find every faint detail. Since both sides of this sensor add a glow to each captured image, these are areas where faint details in deep sky objects are not as easy to capture. To get around this somewhat, I rotated the camera such that the main amp glow areas were on areas of the image where they’d have the least impact. In fact, I planned to crop off most of the amp glow areas. With a 2-panel mosaic, this still gave me the coverage I wanted of the Horsehead and Flame.

Noise is a main reason that astronomy cameras have regulated cooling systems. The colder you can make your camera sensor, the less noise there will be. When this camera was new, it was common to hear that for every 5°C you cooled a camera the noise was reduced by half. All the sub-exposures (aka subs) for this project were captured with the camera set to -5°C. Even with cooling, proper calibration images have to be used to help minimize the noise in the final image. The large amount of sub-exposures, each dithered to point at a slightly different spot in the sky, also help drive the noise down thanks to math…lots & lots of math.

Last year I spent some time calculating signal-to-noise (SNR) ratios for different camera temperatures, different lengths of subs and different numbers of subs. Lately I’ve captured around 20 hours of data for most of my projects. This gave a nice bump over the 5-10 hour projects (or less) I was typically doing before. After seeing some group astrophotography projects with over 100 hours of data, I threw the gauntlet down for this nebula-season project to have over 100 hours of usable images. For me, that’s huge.

To get the most out of my camera and my skies, I decided to use an Optolong L-eXtreme filter for this project. It helps a significant amount with light pollution and lets me gather data concentrated in the Hydrogen-alpha (Ha) and Oxygen III (OIII) emission frequencies. For emission nebulae regions, these are typically the two main spectral lines worth imaging. With the filter in place, Ha light fell on the camera’s red pixels and OII primarily filled the green pixels. There was OIII data also in the blue pixels, but the data from the green pixels was much, much better. So, the camera’s blue pixels were pretty much just along for the ride.

I used NINA to frame the mosaic and developing the imaging plans. I decided on a 2-panel mosaic with a camera rotation of 182° and a 20% overlap. Using NASA Sky Survey data to frame it, the plan looked like this in NINA.

Data capture

It was an amazing October for astronomy in the Memphis area. In that one month alone, I was able to capture 177 hours of subs. That let me finish my two previous biggest projects, the Heart Nebula and the Soul Nebula, along with several other projects. On the evening of October 22nd, 2024, I started this Horsehead & Flame project.

I bought my Losmandy G11 mount last spring and I’ve been through a couple rounds of tweaks to improve its periodic error. Shortly after buying it, I measured the periodic error at +/-9 arc-seconds. After my first round of improvements, I was able to get that down to a PEC-corrected periodic error of about +/-4 arc-seconds. After some more work, I was able to get that down to about +/-2.5 arc seconds. While I was ecstatic about these improvements and I love that that’s possible with an 18-year-old mount, my RA autoguiding has never been as tight as I want it. I haven’t finished either an east-weighted system or my own spring-loaded worm mod for the scope, so the extra slop in RA did have some impact to my stars in this image. Improving that is on my to-do list. I think with BlurXTerminator, you’d wouldn’t even be able to tell.

I finished collecting data for this on the morning of January 17th.

You can view the Lightbucket imaging logs here:

• Panel 1: https://app.lightbucket.co/astronomers/grouchoduke/targets/44820f60-d7f1-4e58-ab15-512207f1546b
• Panel 2: https://app.lightbucket.co/astronomers/grouchoduke/targets/ecbfac81-60ab-40d7-a03c-bd130d77ff41

Processing

Along the way, I made a sanity check of the data. At about 16 hours of data gathered for each panel (32 hours total), I put this edit together. I hadn’t heard of Siril’s in-development mosaic tool yet, so I got some help from a friend in the Memphis Astronomical Society who stacked and stitched the mosaic together using WBPP in PixInsight. Even though this was just a quick edit, it was enough to show me that I was on the right track and got me more excited about this project. (This is obviously a different color palette than I chose for the final image. This is a more of an HOO-style presentation.)

After imaging for 3 months, I hit 116.7 hours of total data. Doing intermediate data reviews along the way, I knew this would be enough to have over 100 hours of data for the image plus some extra. Now, with a fresh pile of 700 sub-exposures, each 10 minutes in length, it was time to get to the real data crunching. The processing turned out to have a few more roadblocks than I expected, so I thought I’d share those.

First was the stacking. Normally, you don’t hear about the stacking people do for their astronomy images. They just do it and it’s great. Boring even, right? With this much data, I spent quite a bit of time varying stacking settings to zero in on what data I wanted to include in the final image. I did an initial pass in Nebulosity to automatically grade the images. I mostly used this to get a feel for what I had. I then went through each image manually and discarded ones that were obviously bad. I then tweaked rejection, normalization and other settings in Siril to combine my final stacks of each mosaic panel. For Panel 1 (Horsehead), I kept 203,400 seconds of Ha data and 202,800 seconds of OII data. For Panel 2 (Flame), I kept 203,400 seconds of both Ha and OIII data. Calling Panel 1 202,800 seconds and Panel 2 203,400 seconds, that left only 3.9 hours of data on the cutting room floor. I could have probably cut more, but I was happy with how that result looked. So, I ended up with 112.8 hours of data that I kept.

Like I mentioned earlier, the Siril team is working on a new mosaic feature. Rich at Deep Space Astro did a great sneak peak for it, so I grabbed a Siril nightly build and started playing. This is my first mosaic, so I had some learning to do. I had to do more massaging than the quick clicks Rich did to align and combine his mosaic panels. One of my panels wouldn’t align using the method he did, so I had to use a more manual process to plate solve, register and align the panels. Once I got each frame aligned (you know, green messages instead of red messages in Siril), the mosaic worked. From there, I calculated my pixel overlap and adjusted Siril’s mosaic feather setting until it gave me the result I wanted.

Background gradient extraction is another thing that’s become easy. Not for this image. With the huge amount of nebulosity all over this image, GraXpert’s AI background extraction didn’t do well. No matter what I did, it wanted to remove way too much actual data. I tried the RBF in GraXpert, but no matter what I tried there it looked like it was taking the background gradient and flipping it. I spent time over several evenings trying to get GraXpert to give me what I wanted. I wasn’t happy with the results from Siril’s background extraction tools either. Just as I was about to take some time off in frustration, Franklin Marek released Auto Dynamic Background Extraction (ADBE) in his rapidly-expanding SetiAstro’s Suite. He posted a great video about it, so I gave it a spin. With ADBE, it didn’t take long before I had the background gradient tackled. (Thanks Frank!!)

Deconvolution caused some problems too. (Again, without PixInsight…) I don’t have access to BlurXTerminator for deconvolution. I normally do deconvolution in Siril using Siril’s standard deconvolution tool. Unfortunately, deconvolution in Siril left all kinds of really nasty artifacts no matter what I did (check out the speckled blotchy mess in this crop of the Siril-deconvolved image). I tried GraXpert next, but it wouldn’t get through its object-only deconvolution. Every time I ran it, it hung at about 95% on its progress bar. Deconvolution is really just a baby step in the process, but it’s an important one that sharpens the image and corrects star shapes.

I used about a million small stretches using GHS in Siril to bring out all the detail I could find. Seriously, I don’t know how many individual stretches I did, but it was a lot.

Alnitak is the thorn in the side of anyone who’s processed a Horsehead or Flame image. As the left-most star in Orion’s belt, it’s extremely bright. It’s a massive star roughly 33 times the weight and 20 times the diameter of our own sun. If I did the math right, Alnitak is over 500 times brighter than all the nebulosity combined around the entire Horsehead region. It stands out. Couple that with the L-eXtreme filter and I ended up with some pretty ugly halos and other artifacts around Alnitak, especially in OIII. I used another one of SetiAstro’s shiny new tools, Blemish Blaster, along with some more work using various amounts of stretched data and some tools in Photoshop to make the most of the Alnitak area.

I typically do my near-final tweaks in Photoshop. I used Camera Raw, Nik, and NoiseXTerminator along with a few other filters and adjustments. Done!

Please check out the full-sized image here: http://astronomynightly.com/wp-content/uploads/2025/02/HorseFlame_v2b-share.png

Details:

• Explore Scientific ED127 refractor
• Hotech SCA flattener
• STF-8300C camera with an L-eXtreme filter
• Losmandy G11 mount on a custom BoldMFG pier
• 700x 600sec sub-exposures (116.7 hours) from my Bortle 5 backyard divided across 2 panels (112.8 hours used)
• Captured in NINA
• Processed with Siril, GraXpert, SetiAstro Suite, Photoshop, Nik and NXT