Best Sub Settings for a RASA8 and ZWO2600MC Pro - To Bin or Not to Bin (Renamed subject)

[Catanonia]

Over a few of my posts, I have had contrasting statements about what is best to use with my RASA 8 and ZWO 2600MC Pro setup.

Some say a few seconds imaging at 1x bin as I will flood the camera with background and loose resolution.

Others are saying I should use 2x or more bin as I will not loose resolution at this FL and setup.

I am currently using 2xBin and 2 - 4 min subs. - I like the quick and deep captures I get compared to 1xbin.

 

So now I am totally confused.

So with  

• RASA8 at 400mm FL
• ZWO 2600MC Pro 
• lEnhance 2 inch filter
• Bortle 5 - 6 skies

What should I be using for optimum images ?

Thanks

 

Edited September 16, 2021 by Catanonia

[vlaiv]

For capture use Bin 1 mode.

You can always (and should) bin in software. In my view, bin x2 is right way to go with RASA8 and ASI2600MC due to pixel size and Rasa spot diagram.

As far as exposure length - that is something you should measure rather than rely on prescribed values. This is because exposure length will be different for people with different setups or even same setup in different conditions.

I'd personally go for gain of 100 with that camera to bring read noise down. That will bring e/ADU to 0.25 or there about (according to ZWO published graph).

In order to get good exposure - take one sub of exposure length that you currently have. Measure mean background value in say blue channel (which is the least sensitive). Multiply measured value with e/ADU value for your gain to get background value in electrons.

Take square root of that to get background noise and then divide it with read noise (1.5e if using gain 100). That is your current ratio of read noise to background noise. You want that number to be at least 5. If it's less - increase exposure length. It is more than 5 - you can use same exposure length as you are currently using or lower your exposure length down.

You can even calculate how much to increase/reduce exposure length - divide 5 with ratio that you obtained by measuring things. Square result and use that to multiply current exposure length to get needed exposure length.

If you run into problem of over exposing either star cores or parts of target - there is quick fix for that. Take small number of very short exposures at the end of the session (say 5-10s exposures). Stack those separately and multiply linear values with ratio of exposure lengths (say 120s / 10s =12 if you use 2 minute long exposures and 10s short exposures).

Replace all over exposed pixels in long exposure stack with values from short exposure stack.

[The Lazy Astronomer]

42 minutes ago, Catanonia said:

Over a few of my posts, I have had contrasting statements about what is best to use with my RASA 8 and ZWO 2600MC Pro setup.

Some say a few seconds imaging at 1x bin as I will flood the camera with background and loose resolution.

Others are saying I should use 2x or more bin as I will not loose resolution at this FL and setup.

I am currently using 2xBin and 2 - 4 min subs. - I like the quick and deep captures I get compared to 1xbin.

 

So now I am totally confused.

So with  

• RASA8 at 400mm FL
• ZWO 2600MC Pro 
• lEnhance 2 inch filter
• Bortle 5 - 6 skies

What should I be using for optimum images ?

Thanks

 

I have to admit one thing I hadn't factored in on your other thread was the spot diagram of the RASA - that's my mistake, sorry.

[Catanonia]

41 minutes ago, vlaiv said:

For capture use Bin 1 mode.

You can always (and should) bin in software. In my view, bin x2 is right way to go with RASA8 and ASI2600MC due to pixel size and Rasa spot diagram.

As far as exposure length - that is something you should measure rather than rely on prescribed values. This is because exposure length will be different for people with different setups or even same setup in different conditions.

I'd personally go for gain of 100 with that camera to bring read noise down. That will bring e/ADU to 0.25 or there about (according to ZWO published graph).

In order to get good exposure - take one sub of exposure length that you currently have. Measure mean background value in say blue channel (which is the least sensitive). Multiply measured value with e/ADU value for your gain to get background value in electrons.

Take square root of that to get background noise and then divide it with read noise (1.5e if using gain 100). That is your current ratio of read noise to background noise. You want that number to be at least 5. If it's less - increase exposure length. It is more than 5 - you can use same exposure length as you are currently using or lower your exposure length down.

You can even calculate how much to increase/reduce exposure length - divide 5 with ratio that you obtained by measuring things. Square result and use that to multiply current exposure length to get needed exposure length.

If you run into problem of over exposing either star cores or parts of target - there is quick fix for that. Take small number of very short exposures at the end of the session (say 5-10s exposures). Stack those separately and multiply linear values with ratio of exposure lengths (say 120s / 10s =12 if you use 2 minute long exposures and 10s short exposures).

Replace all over exposed pixels in long exposure stack with values from short exposure stack.

Thanks, looks like in the previous thread I mis took 2x bin for capture and not software binning.

I am familiar with differing exposures for blown out cores (aka m42) as done that before.

Thanks, will experiment some more and stop capturing at 2x bin

 

Edited September 16, 2021 by Catanonia

[vlaiv]

3 minutes ago, Catanonia said:

Thanks, will experiment some more and stop capturing at 2x bin

With CMOS sensors - capturing bin x2 format gives almost the same results as doing it in software.

It is exact same process - except for some edge cases - like when you have signal that is stronger and clips. If you bin in firmware - it will be limited to 16bit (or any other bit range camera uses) - but in software you can have higher bit precision (like doing 32bit floating point).

One advantage of binning in firmware is that you save smaller files - that is handy in some cases. On the other hand - if you bin in software, you have flexibility to decide when and how to bin. You can bin in the end - stacked image (probably the way to go), or you can do different way of binning - like split bin for example (which takes data from one frame and instead of producing "already stacked" binned version - instead of produces 4 smaller split frames that you stack with other frames in main stacking - better for sub pixel offset accuracy and pixel blur).

[tomato]

An ‘in the field” practical response to your query based on my experiences with a set up very similar to yours:

I have a RASA8 and the QHY version of the OSC sensor, the 268c, imaging under a Bortle 5 sky.

The first image is 20 x 3 mins bin 1x1 with the IDAS NBX dual band filter, camera in Photographic DSO mode.

The second is 150 x 1 min bin 1x1 with an IR/UV cut filter, camera again in Photographic DSO mode. The star shapes are horrendous in the corners of this image, but that’s nothing to do with exposure times or camera settings.

[Catanonia]

29 minutes ago, vlaiv said:

With CMOS sensors - capturing bin x2 format gives almost the same results as doing it in software.

It is exact same process - except for some edge cases - like when you have signal that is stronger and clips. If you bin in firmware - it will be limited to 16bit (or any other bit range camera uses) - but in software you can have higher bit precision (like doing 32bit floating point).

One advantage of binning in firmware is that you save smaller files - that is handy in some cases. On the other hand - if you bin in software, you have flexibility to decide when and how to bin. You can bin in the end - stacked image (probably the way to go), or you can do different way of binning - like split bin for example (which takes data from one frame and instead of producing "already stacked" binned version - instead of produces 4 smaller split frames that you stack with other frames in main stacking - better for sub pixel offset accuracy and pixel blur).

Like shooting a dslr in raw is recommended. You then have the raw data to do with what you like in processing which you otherwise would not have with a JPG  (not binning, but the best example I can think of)

That said, me binning 2x in my set = roughly what I would get with software binning, albeit smaller files and less exposure lengths needed for the same result (yes I don't have the raw 1x bin data and the ability to finer process it)

It all then comes down to exposure lengths, something I need to look more into with my data / seeing

Thanks

Edited September 16, 2021 by Catanonia

[Catanonia]

Ok did some tests on IC1396 just now whilst setting up

With the RASA 8 and ZWO 2600MC Pro,

• at 1x bin, I could separate the 2 stars near the tip of the trunk. 
• at 2xbin, the stars were merging together (same exposure length)

So that determines it for me on at least tonight's seeing, bin 1 is they way to go and probably keep it there in processing.

 

 

[vlaiv]

10 minutes ago, Catanonia said:

Ok did some tests on IC1396 just now whilst setting up

With the RASA 8 and ZWO 2600MC Pro,

• at 1x bin, I could separate the 2 stars near the tip of the trunk. 
• at 2xbin, the stars were merging together (same exposure length)

So that determines it for me on at least tonight's seeing, bin 1 is they way to go and probably keep it there in processing.

 

 

Do you have that bin x1 sub?

I don't think that star separation will be affected by bin x1 / bin x2 on ASI2600 and RASA8 like that.

[Catanonia]

2 minutes ago, vlaiv said:

Do you have that bin x1 sub?

I don't think that star separation will be affected by bin x1 / bin x2 on ASI2600 and RASA8 like that.

Yes noticeable difference between x1 and x2 bin on a 10 second shot.  One the x2 bin the stars were merging and x1 bin they were separated. Was quite a noticeable difference.

Will do a quick image test and post, but will be screen grabs from ASI Pro

Edited September 16, 2021 by Catanonia

[vlaiv]

1 minute ago, Catanonia said:

Yes noticeable difference between x1 and x2 bin on a 10 second shot.  One the x2 bin the stars were merging and x1 bin they were separated. Was quite a noticeable difference.

Will do a quick image test and post, but will be screen grabs from ASI Pro

I asked because I think that it might be a firmware binning issue more than anything.

But don't worry about that for now - I advocate acquisition at bin 1 mode and binning later. I wanted to check if there is indeed such resolution difference as I don't think it should be if data is binned regularly.

[Catanonia]

Here you go.

x1 bin on left, x2 bin on right. Screen grab from PI before debayer, both about 5 seconds.

I repeated the test about 5 times and got the same results each time. 

Obviously differing zoom levels in PI to match scale

Edited September 16, 2021 by Catanonia

[vlaiv]

2 minutes ago, Catanonia said:

Here you go.

x1 bin on left, x2 bin on right. Screen grab from PI before debayer, both about 5 seconds.

I repeated the test about 5 times and got the same results each time.

Ah, ok, yes.

Well - color camera. Bin in firmware needs to be done on bayer matrix so this is in effect 4x4 bin - or no, it's not, but rather 2x2 bayer matrix element which consists of 4x4 pixels is binned into 2x2 bayer matrix element.

That is another reason why you should bin in software - it avoids things like that and let's you control things.

In reality things are not as bad as it might seem from above images. After debayering and when using proper rescaling algorithm (not nearest neighbor like in preview) things can and will be different.

However - that is complex topic that is not quite related to this thread (we can discuss that if you wish as you started the thread and might be OK with that being discussed as well).

Here is just a quick example of that I mean:

I made simple "two star" system - two dots blurred just tiny bit by gaussian blur. Left image is "zoomed in" image at 800% using nearest neighbor resampling. Right is 800% zoom using cubic interpolation.

Left is what happens when people say "my stars look blocky". They just look like they are made up out of little squares - that people often confuse for pixels - but pixels are dimensionless points - they are not squares (image pixels).  Squares is just artifact of used resampling method.

Similar thing happens when you zoom in image that is still bayer matrix and using nearest neighbor resampling (often used in preview because it is much faster computationally then other methods).

 

[Catanonia]

2 minutes ago, vlaiv said:

Stuff I am still reading and digesting

 

Thanks buddy. I have changed my ways for this session and all subs are 1x bin and I have dropped my exposure down to 2min guided to see what results I get.

Yes I will be loosing a lot from 2xbin (firmware) and 4mins, but I will be getting 2x as many frames, albeit only 1/4 of the effective exposure.

Shall see how it progresses.

What I love about this hobby, always learning and playing

[Catanonia]

Same 2 images debayered in PI using defaults that I have always used. 

Certainly a topic for a new thread.

[vlaiv]

4 minutes ago, Catanonia said:

Same 2 images debayered in PI using defaults that I have always used. 

Since you are using PI - well, do two things to binned frame:

1. Enlarge it x2 using Lanczos resampling

2. See that ADU values are the same between two images (binned version might have x4 as high ADU if it was binned additive instead of average) and stretch them the same

then compare them.

[Catanonia]

The best I could get with playing with Lanczos settings.

Same Histogram Stretch on both Debayered images

Looks quite definitive, 2xbin (firmware) is not the way to go on the ZWO2600 is you want resolution on the RASA 8  (not unless you are a wizz with processing)

Edited September 16, 2021 by Catanonia

[vlaiv]

3 minutes ago, Catanonia said:

The best I could get with playing with Lanczos settings.

Same Histogram Stretch on both Debayered images

It does improve overall shape of things, but you are quite right, with these settings - image on the left is clearly better.

[Catanonia]

5 minutes ago, vlaiv said:

It does improve overall shape of things, but you are quite right, with these settings - image on the left is clearly better.

Would gain affect this, even on a few second exposure ?

 

[vlaiv]

1 minute ago, Catanonia said:

Would gain affect this, even on a few second exposure ?

 

If both subs are stretched the same - then it should not.

Different level of stretch can impact star separation a bit.

Depends where you put your white point in stretching and how much you stretch. In image above - if you apply weak stretch (top line represents white point level) - you'll get two separate star profiles (orange bars), but if you do strong stretch and set white point low (bottom red line) - you'll get "joined" star profile - bottom orange line.

Even if you have same exposure length - if binning is done by summing - then ADU values will be higher in binned image (sum of 4 pixels) - and that will make two peaks higher in the image. If you set white point to same value - it can be lower than in bin x1 image with respect to above profile and one can end up having joined stars while other can have them separate.

Above example is mainly due to bayer type binning which is slightly different than what I'm suggesting, and also because of debayering method used.

When you finish your capture and stack your bin x1 images - you can experiment with binning that stack while still linear to see if there will be difference in resolution in that case. In PixInsight - binning is done when you select integer resample and set mode to average.

[Catanonia]

8 minutes ago, vlaiv said:

If both subs are stretched the same - then it should not.

 

Got me thinking on a new imaging method (still thinking it through in my head)

• 1. Gather data at x2 bin
• 2. Gather the stars at x1 bin  (take some separate subs for the stars
• 3. Process the x2 bin frames with drizzle to get a 2x size, the same size as x1 bin
• 4. Remove stars with starnet and replace with the ones from the x1 bin.

I my head it works, but I bet there would be terrible ringing around the stars and it would look fake.

[vlaiv]

1 minute ago, Catanonia said:

Got me thinking on a new imaging method (still thinking it through in my head)

• 1. Gather data at x2 bin
• 2. Gather the stars at x1 bin  (take some separate subs for the stars
• 3. Process the x2 bin frames with drizzle to get a 2x size, the same size as x1 bin
• 4. Remove stars with starnet and replace with the ones from the x1 bin.

I my head it works, but I bet there would be terrible ringing around the stars and it would look fake.

You can really achieve the same or even better result by doing the following:

1. Gather data at bin x1

2. Separate into starless and stars version by using starnet (create starless version and then subtract from original - that will create stars only version)

3. Bin starless version x2 in software

4. Resize starless version to original size by using some fancy resampling method (like Lanczos or Mitchel or similar)

5. Add stars back in.

What you've done is basically wavelet type denoising selectively applied on the nebulosity and backround and there is no need to acquire two sets of data.