First use of the Optolong L-eXtreme filter - strange pattern in Ha and OIII image

Have you never seen this before? I would almost assume it has to do with the modification of the camera.

@vincent-mod
I have been using the DSLR (EOS-750Da) for about 2 years. Also I have already used UHC filters. These strange patterns have never appeared before.

Mmmm, would you mind sharing some of the raw data? Like 20 light frames and the master calibration files?

Go to https://upload.astropixelprocessor.com and use upload1 (or upload2 till 5) as the username and the password will then be the same as the username.

Create a directory named “stefan-b-weirdpattern” and upload in there. Thank you!

Hello, Vincent,
thank you very much in advance for the help.

I have uploaded 20 lights from three sessions and the corresponding calibration files.
I had already integrated the frames of the first session (29.07.2020) (H-Alpha and OIII). I could not find this strange pattern. It seems to appear only when you integrate all three sesions.

I'm curious what they find out and what is the reason for it.

Greetings
Stefan

Excellent, I’ll have a look and report back within 2 days. Thanks

Ok so that took a bit longer, sorry about that.

I just loaded one session and had a look at the calibration data. Wat strikes me is that the masterbias has a weird pattern in it. I'm investigating now if that might be the cause.

Thanks for the information.I have also noticed the pattern. The horizontal stripes in the masterbias are probably a peculiarity of the sensor of the Canon EOS-750D. This is not only the case with my camera, but with all models of this series.
So far I never had problems with the integration of several sessions.
The only difference now is that I used the alogrithm Ha-OIII Extract Ha or Ha-OIII Extract OIII for three sessions (multi-session processing) for the first time.

I have been told for quite a while now (over three years) that DSLR cameras and bias frames don’t go very well together. The short exposure time basically is too short for the sensor. The same is true for many new CMOS sensors (as found in for example ASI cameras).

Have you tried taking dark frames, flat frames and dark flat frames (i.e. dark frames with the same length and iso/gain as the flats) and only use those? That’s what I do for all my ASI cameras and that seems to work wonderfully well for me.

HTH, Wouter

@wvreeven

I have always calibrated with biases, darks, flats and dark flats. I will try it without the biases. I am very curious about the result, if there is a difference.

Thanks for the advice!

Greetings

Stefan

Not sure if DSLR's have problems that much in general, if I'm not mistaken it's mainly the new CMOS sensors that seem to have that. But in any case, a bias frame of about 0.5 seconds per sub works as well and I'd advice to do that instead of the fastest possible setting.

So I'm processing the data, forgot to actually use the extract features, must be the heat. 😉 But funny is that it doesn't show the artefacts, I didn't use the masterbias though.

I'll reprocess tomorrow with your workflow.

@vincent-mod

Currently I am creating BIAS frames with 0.5s exposure time. After that I will start a new image processing with APP using the new BIAS-frames.

My Workflow:

1. separate creation of all master frames for session 1, 2 and 3

2. integration with multi-session processing using Ha-OIII extract Ha

3. integration with multi-session processing using Ha-OIII extract OIII

4. Combine RGB --> HOO

Cool! Let's see what that does.

I processed your 60 frames just all in 1 session (extract Ha, LNC and MBB), used masterdark, BPM and masterflat also from session 1 and this popped out;

Can't see any artefacts. And darks don't have to be taken every session, only flats basically. Beautiful signal!

ps. I see your setup is undersampling, so if you dither well you can use drizzling to increase resolution.

@vincent-mod

That looks great. 

Thanks for the tip for the drizzle integration.
All my light frames are dithered, so I can use the drizzle mode. Which settings should I use?

Kernel: topHatKernel
droplet size: 0.5
mode: drizzle
scale: 2

Is this correct?

Here I made a combo of the Ha and OIII;

@stefan-b

I've never used it myself actually, as I have no undersampling. So nice to try it on your data. 🙂 Those standard settings are good to start with or are even fine as they are, a smaller drop size will increase resolution. You will get twice the resolution, but also more noise, that is a trade-off with this technique.

@vincent-mod

Thanks a lot Vincent for your support.

The PC is working right now. I will try the above mentioned drizzle setting and see what happens. May take a while with 106 lights. 😉 I will get back to you and post my results here.

Have a nice evening and a nice weekend.

Greetings
Stefan

One final screenshot, bayer drizzle (you have to select that when working with color sensor data) and your settings above. You can clearly see that the stars improved and are less square; this crop is at 100%. There is a pattern coming up though, so might need either more data or slightly different settings.

@vincent-mod

Wow! This looks promising. To test it and make sure it doesn't take that long, I'm gonna
set "crop" for the composition mode, select a small section in the reference frame and experiment a little.

Ok, been practicing a bit with it. 🙂 Scale 1, droplet size 0.6 gives this (at 100%, using a crop of the image for testing). There is a bit of a pattern when you zoom in to 150 or more %, but I think that's just what it is. The stars being rounder is much more noticable and maybe it improves with more dithered data.

Full version, I like it. 🙂

You can right-click and open it in a new tab for full resolution

I did notice that for OIII it may differ, so it's good to test both.

@vincent-mod

Hello,
sorry for my late feedback. I did some more tests with different APP settings and generated some GByte of data 😀 This took some time 😉 

1. Strange pattern

Among other things I created again new BIAS frames with 0.5s exposure time and did the integration with it. But this did not bring any improvement. The disturbing patterns were still in the image.

Next I worked without darks, but with bias, flats and BPM. The patterns were also clearly visible here.

Finally I tested the variant without bias, but with darks, flats and BPM. Finally I could achieve a significant improvement. The disturbing patterns are hardly noticeable anymore.

My settings in APP

0) RAW/FITS   CFA algorithm: Ha-OIII color

6) INTERGRATE   average, quality, LNR adaptive rejection 6.0 /3.0, 1st degree LNC iterations 3, MBB 5%, droplet size 0.6,

                         Mode drizzle, scale 1.0

Screenshot full frame

Screenshot 100%

It seems, that when using the Optolong L-eXtreme filter, the bias of the DSLR cannot be used.

2. Drizzle-Mode

You have noticed an undersampling in my pictures.
So I experimented a little bit with the drizzle function and found the following:

• "Ha-OIII extract Ha" and "Ha-OIII extract OIII" only works in the "Drizzle" mode, NOT with "Bayer/X-Trans Drizzle".
• In Bayer/X-Trans Drizzle mode, only a Ha-OIII mono is created, not a Ha or OIII image. The generated images for "extract Ha" and "extract OIII" look the same. I cannot say if this is a mistake in the program!
• An image calculated with the algorithm Ha-OIII-Color using the mode "Bayer/X-Trans drizzle" appears much darker and has less dynamic than an image created in the mode "Drizzle".

Maybe you can give some more information about "Ha-OIII extract Ha" in connection with the mode "Drizzle" or "Bayer/X-Trans drizzle".

Thanks again for your support!

Greetings

Stefan

FYI my girlfriend has an Optolong L-eXtreme filter as well and has taken pics of the Western Veil Nebula as well. No strange pattern is visible. She used an ASI2600MC camera with only lights, flats, darks and darkflats. Both the flats and dark flats were exposures of 10 seconds due to the small bandwidth of the filter. I know, it is not a DSLR camera but I remain that with any CMOS camera BIAS do more harm than good, even if no small band filter is used.

Wouter

Very interesting I've never had an issue with bias on my older DSLR's, it definitely improves my results (and darkflats are basically bias frames anyway, just a little longer exposed, which didn't work here either). Maybe it simply differs per older DSLR sensor, first time I see it having such a big effect, so good to know!

Let me get back to you on your drizzling questions tomorrow. 🙂

Good point Vncent @vincent-mod. I have only tried with (fairly) new Canon DSLRs so your mileage may vary.

Wouter