Page 1 of 1

A Scientific Approach to Narrowband Imaging

Posted: Fri Dec 18, 2020 6:56 pm
by oopfan
It is that time of year again for the Tadpole Nebula in Auriga. Last year was my first attempt at it. Sara Wager inspired me with her exceptional image using Ha and OIII.

Like most people, I chose a single exposure and then captured an equal number of frames. It became clear after the first few frames that Ha was much stronger than OIII. After the second session, I created an interim image in the HOO palette. I discovered that the OIII stack was too weak. I made a mid-course correction and devoted the third session to capturing more OIII. Well, it wasn't enough.

I searched Sara Wager's blog and discovered that she stretches the OIII stack before combining. That worked, but the downside was that my final image was noisy. I later learned that the process of stretching magnifies the noise as well as the signal. There is no such thing as a free lunch, as they say.

Fast forward to today. With 20/20 hindsight, I spent too much time capturing Ha and not enough time for OIII, but how much is enough? I decided to use my knowledge of photometry to help answer that question. The other night I captured one 1000-second frame in each of Ha, OIII, and SII. I calibrated them with AstroImageJ and performed photometry. Here is what I discovered. Ha is 3.5 times stronger than OIII, and OIII is twice as strong as SII.

Every emission nebula is different. The Jellyfish nebula has nearly equal amounts of Ha and SII but only a trace amount of OIII. Fortunately, it is a characteristic of the nebula, so you need to do the photometry only once. Furthermore, I normalized the numbers so that you can use any telescope and camera.

I am in the process of adding this feature to my website. In the meantime, Dave and I and another gentleman are working on expanding the database. I will follow up with some images when the weather clears.


Re: A Scientific Approach to Narrowband Imaging

Posted: Sat Dec 19, 2020 2:37 pm
by admin
Hi Brian,

thanks for sharing – I guess that means that you need seven times more total exposure time for SII than for Ha to get the same sort of noise levels in the two channels in the stacked data. On top of that, the sky brightness due to light pollution in the different channels will also differ (and possibly by different amounts), so you may end up wanting to take different sub lengths in the different channels, in turn leading to the need for various different dark frames.

It's a good job we mostly do astrophotography for the challenge, isn't it?!

Cheers, Robin

Re: A Scientific Approach to Narrowband Imaging

Posted: Sun Dec 20, 2020 5:39 am
by oopfan
Hi Robin,

Yes, the total integration time of SII will be seven times longer than Ha but only if we choose a long exposure for SII so that the per-sub SNR is the same as Ha. In reality this is impractical, so we need a longer integration time for it.

Regarding light pollution, I am assuming that it is panchromatic and uniform in intensity across the spectrum. The amount of LP signal and therefore noise is proportional to the filter's bandwidth, therefore a 3nm will have less LP than a 12nm filter. I may have to employ a more complex model as results roll in.

After looking at many Astrobin images, it is clear that most people are not making optimum use of their time at the telescope. This is why we are seeing very long integration times. They capture A LOT of data and then sort it out in post-processing. My theory is if they knew what the nebula's composition was ahead of time, then they could tailor their integration times accordingly. I believe they could achieve similar results in half the time.

Thanks for your interest!


Re: A Scientific Approach to Narrowband Imaging

Posted: Sun Dec 20, 2020 1:55 pm
by oopfan

The only problem with the photometric method is that the ratios change with the bandwidth of the filters. For example, a 3nm Ha filter will isolate the Ha emission line, but a 7nm Ha filter will also capture two nearby NII lines. The NII lines are just as strong as Ha in some nebulae. The best solution is to build the database from the results of high-precision spectroscopy. Unfortunately I've not found a source for regular emission nebulae, only planetary nebulae. So, using the photometric method we need to specify the bandwidth of the filter. We just can't say that the strength of "Ha" is "x", instead we say the strength of "3nm Ha" is "x".

Another challenge is where do we sample the flux on the nebula? For planetary nebula, I place the circular aperture around the nebula since most PN are circular. Regular emission nebulae are irregular. I've tried two methods: (1) sample the same spot in each filter, (2) sample the spot that represents the mid tone. That second method usually means that the aperture is placed over different parts of the nebula.

The good news is that the photometric results are virtually the same for a calibrated frame versus a non-calibrated frame, but I urge people to use a long enough exposure to get a good signal in the test images. If you want to image a nebula that isn't in the database, then I recommend capturing the test frames first, and then immediately start capturing OIII or SII while you run photometry on the test images. The reason is that Ha is strong in most nebulae, so it is best to begin with the weaker filters while you evaluate your test images.

One last point about photometry, AstroImageJ is designed for exoplanet research. Under normal usage the background signal is sampled at the same time using the "outer annulus" of the aperture. For our purposes we can not use that value for the background, so we define the aperture so as to exclude the outer annulus, or in other words, force the area of the outer annulus to zero. So what do we do for the background? We find an area of the frame devoid of nebulosity, and sample it. My website's calculator will subtract that background reading from the nebula readings. Be careful sampling the background near the edge of an uncalibrated frame. Vignetting can be severe. It is best to calibrate your images, if possible. AstroImageJ makes calibration easy.


Re: A Scientific Approach to Narrowband Imaging

Posted: Sun Dec 20, 2020 3:34 pm
by turfpit
An example to support the posts from Brian. M27 processed as RGBHOO in Astro Pixel Processor. RGB are each 6x300s, Ha & OIII are each 3x600s. Total integration for the 24 frames = 2.5h The other 2 attachments show the results of the analyses of single NB frames in AstroImageJ, where for this object the NB frames are a reasonable balance (7.17 and 8.15). With 30m capture for each of Ha and OIII it is possible to project how much narrowband would be required to show the outer shell.

Captures with an FLI 16803 CCD camera with 100k Full Well Depth.

I am sure Brian will be along soon with the gory details.

M27-P1-RGBHOO_R-6x300s_G-6x300s_B-6x300s_Ha-3x600s_OIII-3x600s-reduced.jpg (239.85 KiB) Viewed 589 times

Ha apertures.jpg
Ha apertures.jpg (101.47 KiB) Viewed 589 times

OIII apertures.jpg
OIII apertures.jpg (104.47 KiB) Viewed 589 times

I also have a capture of IC434 in SHO. With the same analyses we found that with the region being strong in Ha, that I had to reduce the number of Ha frames used as these were causing an imbalance with the false colour. The analysis exercise would need to be done on each object as the strengths of Ha, OIII and SII vary wildly between objects.


Re: A Scientific Approach to Narrowband Imaging

Posted: Sun Dec 20, 2020 6:30 pm
by oopfan
Thank you, Dave. Excellent images!

The Dumbbell Nebula is a planetary nebula, so we have access to its spectrum, courtesy of Professor Karen B. Kwitter, Williams College: ... /index.php

I've zoomed in on the area around the hydrogen-alpha (Ha) line:
M27 Ha NII lines.jpg
M27 Ha NII lines.jpg (21.61 KiB) Viewed 579 times
Notice that the Ha emission line (middle) is surrounded by two NII lines. The telescope that Dave used has a 7nm Ha filter. It was wide enough to capture all three emission lines. So, although the strength of the Ha line is only one-third of OIII, the addition of the NII lines using a 7nm Ha filter makes the total flux approximately equal to OIII. The photometry shows this. Now, if we were to switch to a 3nm Ha filter, then the Ha emission line would be isolated, and we would need to increase the integration time. I wish we had spectra for all emission nebulae, but absent that we must rely on photometry.

Dave mentioned increasing integration time in Ha and OIII to capture the outer shell. Given the results of the photometry, and confirmed by spectroscopy, Dave's two stacks have the following characteristics:

Ha: 3x600s, SNR 39
OIII: 3x600s, SNR 43
Total SNR: 58
Total Integration Time: 1.0 hour

I can't predict when the outer shell will reveal itself, but my rule of thumb is to double the SNR, therefore I recommend:

Ha: 16x600s, SNR 90
OIII: 14x600s, SNR 92
Total SNR: 129
Total Integration Time: 5.0 hours

So, Dave needs to capture another 4 hours of data to reach a total SNR of 129. Unfortunately M27 is gone for the season but it will be coming around again soon.