How to expose for Narrowband - Forums

Topic starter

Apr 5, 2021

Hi to all users out there,

during the last week, I took the opportunity to spend some nights taking pictures of part of IC1805. I was able to combine the Ha data into an image that can be found here:

https://www.astrobin.com/61hd3h/?nc=user

Last night, I took some pictures in OIII and SII as well to combine them with the Ha to a color image. But here I ran into some problems that I need help with.

I own a licence of PixInsight for a few days and I am obviously new to that software. It turned out, that I had problems stacking the SII data. I took the images using my ZWO ASI183MM camera with a gain setting of 178 (unity = 110), binning 1x1 and an exposure time of 300 seconds. When I try to stack these images with the WBPP2 script in PixInsight, I am not able to get a master light. After I did some research, it seems that the SNR of the subframes is too bad to stack them properly. I also tried DSS without any success. If I use PIs ScreenTransferFunction to make the content visible, I do see really a lot of noise. So I guess I may be right with my assumption. The OIII images (about 36 frames with 300s each) were good enough to generate some data. Here, I had really bad reflections on some brighter stars, which seems to be a problem of the cheap ES OIII filter. But that's another topic.

I guess, that exposing 300s in SII with the given gain and binning settings seems to be not enough. I may try some longer exposure times. But the guiding (ES MN-152 with 731mm focal length as main scope and 240mm guide scope with Altair 178M) had to be really spot on. Its quite difficult to ensure this quality. So I want to use the opportunity to ask you guys, what your settings for Ha, OIII and SII are. May a binning of 2x2 in one of the channels be an option? I had some "big" stars in the given Ha image that may appear smaller with shorter exposure times. But I'm afraid that I will have problems stacking the data, again. Does someone experience the same problem in PI that you are not able to stack a series of images?

Thanks for your help.

Clear skies

Christian

Helpful Engaging

Björn Arnold

Apr 6, 2021 · Edited: Apr 6, 2021

Hi Christian,

Let me send this head: I haven't been trying narrowband imaging yet but maybe this may help you anyways:
I own a Altair 183M Pro (uncooled) which uses the same sensor. Since I had issues capturing good subs in LRGB, I got into some CMOS "research" and learned a few things:
First I dropped the idea of unity gain. From what I've read and seen, I cannot see anymore what it shoud be good for. I recently captured subs going below unity gain (higher read noise) to increase dynamic range.

Binning won't help you a lot. From my experience, you need the same min. exposure time for all binning settings (for CMOS).

For my camera I would need an exposure time of about 1200 seconds for a 12nm narrowband filter. That's very long and I would then try to increase gain to reduce it. With increase gain, I would still believe your exposure time needs to be something like 600 seconds for narrowband.

CS!
Björn

Helpful

Brian Boyle

Apr 6, 2021 · Edited: Apr 6, 2021

Hi, I am pretty much a n00b at NB imaging, but this is what I have gleaned from others and my own calculations

1) The KISS principle i.e. Keep it Simple, Stupid. Just starting out, I use the the same exposure length in each narrow band, same gain on my ASI1600MM Pro (139), same temp (-20C).
2) I interleave H,S, and O subs - my filters all have same focus point - to minimise seeing/weather differences between bands.
3) Due to RN, I now give each sub image a minimum of 600secs, you run the risk of being RN dominated otherwise. In bright moon (or bright skies) you might be able to dial back to 300sec, but I shoot in Bortle Class 3.
4) Take at least 15frames in each band, to improve rejection.
5) And dither every group of 3 frames.
6) Get good calibration images. Flats every night. Darks more occasionally (its what bad nights are for!). I have a library of darks are various gains, temps, and exposure lengths. But mostly I use T=-20C Gain=139 and t=30, 60, 120, 300 and 600sec. The last is for NB, rest are for broadband.
7) I also find PI very good for NB images. I use SHO-AIP to combine, in the scripts section.

Now, there are people out there vastly more experienced than me, and they no doubt have much better prescriptions. But these are my guidelines just starting out in NB (particularly number 1). No doubt I will get better, but it has got me to first base.

CS

Brian

Helpful Engaging

Bogdan Borz

Apr 6, 2021 · Edited: Apr 6, 2021

Hi Christian,

This is weird. It is not a problem of signal if you don't have a rejection algorithm for SNR for example. One issue that comes up is that the subs cannot be star aligned and then they won't be integrated. But if they are star aligned they can be integrated. Did you try to do it manually? Do you have the registered folder with the S2 subs created by WBPP? Of course you can always integrate them manually.

PS: You can stack even very faint signal subframes this is not a problem. Sometimes in narrowband you stack subs where you can barely see the signal on the individual subframe and I had subs where the LP¨ gradient was comparable with the O3 signal for example. So I would try to register them on a good reference (take your Ha sub for example). Once registered, just integrate them manually, the process won't stop or abort even if you signal is faint or noisy.

Best regards

Bogdan

Helpful

Christian Großmann

Topic starter

Apr 6, 2021

Hello and thank you for the help.

@Bjorn and Brian
I think, that your answers both subscribe to my thoughts. At least, its good to know that I am thinking in the right direction. So I may increase my exposure time and do some more investigations.

@Bogdan
The idea of having a problem with the star alignment is a new thing. I may try to integrate the frames manually. But I have to look how it's done. As said, I am really a newbie to PI. But if that works, my frames are maybe usable.

CS

Christian

Christian Vulpescu

Apr 6, 2021

Hey Christian,

Does the registration work on your SII channel?
If yes, than try to just calibrate and aline the frames by unchecking the check mark at "Image Integration" in WBPP2.
So you can Integrate the subs manually.
Therefore go to "All Processes" and choose "Image Integration".
Leave all as defold, change only in "Pixel Rejection 1" , "Rejection Algorithm" Winsorized Sigma Clipping.
And in "Pixel Rejection 2" , at "Rejection high", go with the sigma on round about 2,0 till 2,8 standard deviations.
After the integration process you shold see a mean rejection rate of 1% - 2% in the process console.

If you have a problem with alining the subs, than you can go first to "Cosmetic Correction" and correct the subs for hot pixels.
The registration tool, called "Staralinement", sometimes tryes to aline on hot pixels.

With the cosmetic corrected subs go to the "Staralinement"-Process.
There you put in first a reference image, than the others and than set an output directory.
You will see below "Star detection".
There you can try to raise the number of "Detection Scales). Larger stars will be considered.
In addition you can open the "Star Matching" below "Star Detection" and raise the "RANSAC tolerance" to maximum.
That can help.

CS

Christian

Helpful Supportive

dkamen

Apr 6, 2021

Hi Christian,

If you can see stars at the subs then PixInsight can almost certainly register and stack them, just not with the default settings.

If you look at the process console when registering, it tells you how many stars it detects on each image

The number should match what you are expecting visually. If you see a few tens of stars and the console reports hundreds or thousands, it means noise and hot pixel are mistaken for stars. If it reports a very low number of stars (or no stars found), it means detection is not sensitive enough. Either way, you most likely will need to adjust the parameters in the Star Detection section of StarAlignment:

If your images are too noisy the algorithm is overdetecting. So you want to increase the sensitivity slider to maybe 0.80 or more (I know, it's a little counter-intuitive that you increase something called "sensitivity" in order to reduce the sensitivity) and maybe increase noise scales to 1, which will make it ignore things with radius of 1 pixel.

If your images have too low signal and the stars have abnormal shapes, you could try increasing peak response (which makes it detect "flatter" features) and loweing maximum distortion to perhaps 0.45 (which makes it accept less perfectly shaped stars).

Now, if your sub has only 4-5 large (tens of pixels wide) fuzzy stars and the rest is noise, I agree that this is a hopeless situation and you should expose better (or even check focusing).

With regards to your other questions:
-binning 2x2 does help. It doubles SNR and makes stars exactly half in size (less flat), also reducing the possibility that . With a CCD it quadruples SNR, but even with a CMOS the effect is important.
-With the 178 chip and a 12nm Ha filter, I've had pretty good results by exposing 60-120 seconds and using a high gain (240 to 300) and a high offset (200) at bin 2x2. Dark calibration is paramount so you need the high offset to avoid clipping and the high gain reduces read noise which is very comparable to signal at narrowband. However, at 300 seconds I seriously doubt insufficient signal is your problem. I think it is just a matter of adjusting the detection parameters.

Kindest regards,
Dimitris

Well Written Helpful Engaging

Björn Arnold

Apr 6, 2021

Christian Großmann:
Hello and thank you for the help.

@Bjorn and Brian
I think, that your answers both subscribe to my thoughts. At least, its good to know that I am thinking in the right direction. So I may increase my exposure time and do some more investigations.

@Bogdan
The idea of having a problem with the star alignment is a new thing. I may try to integrate the frames manually. But I have to look how it's done. As said, I am really a newbie to PI. But if that works, my frames are maybe usable.

CS

Christian

@Christian Großmann ,
Hi Christian,

I searched the web and found this information on CloudyNights: https://www.cloudynights.com/topic/616524-sub-exposure-tables-for-the-zwo-asi183mm-and-qhy183m-and-colour-versions/

Given this information, I would say that you need to increase sub exposure time.

Cheers,
Björn

Lynn K

Apr 6, 2021

Hi Christian, you have gotten a lot of tips on your camera and processing. I have been doing primarily narrow band imaging for well over 10 years. I am a CCD imager and know little about gain settings . I use MaximDL v6 and Photoshop V5. So, I have nothing to add to your processing approach.

Processing is always important, but I would like to comment on Narrow Band acquisition. Imagers often say that narrow band needs longer exposures to gain signal. This is not true. The same amount of Ha at 656.281 nm will come through a Ha filter as a luminance filter. The sub-frames of the Luminance filter will appear brighter because of all the other bandwidths it captured. But it will contain the same amount of Ha. However, the Ha filter has filtered out the unwanted bandwidths and therefor appears darker. The background will certainly appear darker. One must remember that pixel brightest is not our goal, but pixel contrast is. It's not that the area/pixels in the Ha area is brighter, but the that the non-Ha areas or darker. Any filter will block transmission of some photons. The more expensive will boast a higher transmission.

The above applies equably well to OIII and SII.

In my opinion, your 5 min subs should be enough to gain a descant signal. IF, other factors are not adding un-wanted noise/signal. Noise is signal also, but it is unwanted signal from sky glow etc. The most disruptive factor on the night you images was probably the moon. It was 91% illumination. Ha filters will block out moon light very effectively. BUT OIII will not. The wider the bandwidth, the more moon light will be let in. A 12nm will let in more than a 3nm OIII filter. And the 12nm OIII will also let in more unwanted bandwidths.

Why the additional background noise?? All pixel have bias. Meaning that the vary in how affective thay convert photons into electrons. Some will do it better and be brighter. Some will do it poorly and be darker. This shows up in a Bias frame. I have concluded that if the pixel is exposed to very little light it is not activated much and stays pretty dark, no madder what it's bias. However when it is exposed to faint light, the the pixel bias appears. Read noise. EQ etc complicates this.

The ZWO QE chart shows your camera's highest QE at 550nm. OIII is 500nm. So the camera should be best in the OIII bandwidth. Most cameras or better in the Ha.
However, OIII can be very weak in most emission nebula. The part of the nebula you imaged is the richest part in OIII, but still a lot weaker than the Ha. And consequently the difference in you Ha subs and your OIII subs. I have taken OIII sub of some nebula and have gotten very little to almost no signal. However the Veil nebula will show as strong OIII as Ha.

I have used Astronnomic 12nm filters (mainly C11 Hyperstar), but moved to 5nm Astrodons. I then moved to a 3nm Astrodon OIII and saw my background noise improve greatly. I am staying with the 5nm Ha because the signal in usually strong enough and I get some NII in the mix.

In my experience the OIII can require a narrower band pass and longer exposures. There is no set rule because of the varying QE of the cameras and nebula.

If you are processing a nebula with poor OIII using the Hubble pallet, it will dominated by the green channel and difficult to process. Often those nebula are best processed using Ha as Red, OIII as Blue with synthetic green.

I hope my comments help a bit with understanding narrow band imaging. Again, I think the biggest issue was the moon light not being filtered out by the OIII.

Lynn

Helpful Insightful Respectful Engaging