Software binning explained

[jimjam11]

I used this link:

https://pixinsight.com/forum/index.php?topic=7388.0

 

Basically a combination of the Statistics process and Noise Evaluation script...

[david_taurus83]

10 hours ago, jimjam11 said:

I used this link:

https://pixinsight.com/forum/index.php?topic=7388.0

 

Basically a combination of the Statistics process and Noise Evaluation script...

Thanks. I did try that but I must have been using the wrong figures. Do you set to 16 bit mode in Statistics?

[jimjam11]

No, normalised real and scientific notation. They are then directly comparable with the noise evaluation output...

[david_taurus83]

17 minutes ago, jimjam11 said:

No, normalised real and scientific notation. They are then directly comparable with the noise evaluation output...

Thanks. I didn't convert the decimal correctly the first time. My results:

1x1 stack, SNR 6.01, FWHM 2.644px: 3.67"

 

2x2 stack, SNR 10.05, FWHM 1.561px: 4.33"

 

1x1 stack binned, SNR 9.437, FWHM 1.497px: 4.16"

 

1x1 subs binned before stacking, SNR 9.439, FWHM 1.495px: 4.15"

 

So very marginal improvement between binning the subs prior to integration versus binning the integration. It does appear to improve stars and mask the tracking errors. The pixels on the 1600 are small anyway so you only notice the binning if you pixel peep.

[jimjam11]

On 13/10/2019 at 13:27, vlaiv said:

I regularly bin my data in software when I record it with 8" RC (1624mm FL) and ASI1600. Native sampling rate is around 0.5"/px and that is of course way too much. Sometimes I bin it 2x2 and sometimes 3x3 - depending on conditions of the session.

What is the benefit of imaging at such a high resolution if the sky never supports it? I am dabbling with an RC6 which gets me to a similar image scale or 1.15" when binned 2x2 so I am wondering what benefit it has over my 150p which natively comes in at 1.15"/px?

[vlaiv]

13 minutes ago, jimjam11 said:

What is the benefit of imaging at such a high resolution if the sky never supports it? I am dabbling with an RC6 which gets me to a similar image scale or 1.15" when binned 2x2 so I am wondering what benefit it has over my 150p which natively comes in at 1.15"/px?

There is very small benefit in terms of pixel blur if you process your subs right.

However, that is not the main reason why I image at that resolution. I image at that resolution because that is what my setup gives me, and I can bin image to get to resolution that my guiding and sky allow.

When I was choosing my setup, resolution was only one of concerns, and in hindsight, I probably over estimated things - 1"/px is really hard to reach (FWHM of 1.6") even if sometimes I do get seeing in 1" FWHM zone (0.8-1.2" FWHM). I might be able to utilize this resolution more often once I move away from the city and get better mound (aiming at mesu 200). Good thing is that one can use focal reducer to widen the field in this combination (up to 30mm of field will be usable and sensor is only ~22mm diagonal), and I'm working on fractional binning - that will get me optimal resolution for a given night (or target).

There are other reasons why I've chosen that setup, and I'll name a few:

- I wanted 8" aperture that could be used on HEQ5 mount. I tried 8" F/6 newtonian and while it worked - it was not stable platform. This means that I needed either F/4 scope - which is very demanding both on collimation and on coma corrector (design and placement). When I did my research, I found out that coma correctors don't really provide perfect correction - some have issues with spherical aberration, some correct coma only to a certain distance of center. Most are designed with very short back focus suitable for DSLR type cameras (55mm) and not for OAG + filters, etc ... Another option was SCT - but I don't like those in imaging role

- RC on the other hand is very good imaging instrument - it is well baffled, won't dew up easily (just happened once to me - dew on secondary), it is compact so easier to mount and guide. Does not need any corrective optics on sensor the size of ASI1600 - it has fairly flat field and round stars. There are additional things about this scope that I value not related to imaging (but rather as science instrument).

That is just a combination of scope and camera that I'm most pleased with and it forces me to use 0.5"/px when shooting and bin later, and that is fine with me. I do have another scope that I use with ASI1600 for wide field imaging and resolution of 2"/px - a small frac.

 

[jimjam11]

Thanks, I thought that might be the answer. How are you deciding what level of binning to go for, is that 1/3 or 1/4 of your star FWHM so you maintain well sampled stars or something cleverer?

 

 

[vlaiv]

1 hour ago, jimjam11 said:

Thanks, I thought that might be the answer. How are you deciding what level of binning to go for, is that 1/3 or 1/4 of your star FWHM so you maintain well sampled stars or something cleverer?

 

 

Very good estimate of optimum sampling rate is about FWHM / 1.6. That is based on Nyquist sampling theorem and Gaussian approximation of star PSF (and it's Fourier transform).

If you for example have star FWHM in arc seconds be 2.8" in your stack (individual subs will differ on average FWHM, but you can do "test round" of stacking to measure FWHM in resulting stack - just convert to arc seconds based on your initial sampling rate), then optimum sampling rate is 2.8 / 1.6 = 1.75"/px.

Most of the time you won't be able to bin to exact value based on FWHM, and I advocate going "a bit over" rather than "a bit under".  If you for example have 0.5"/px sampling and your ideal sampling rate is 1.2"/px - I would rather bin x3 and go for 1.5"/px then bin x2 and go for 1"/px.

First - it will raise SNR by larger factor, and second - image will need less sharpening if any.

[jimjam11]

Thanks, what range of fwhm do you see with the rc8?

Have you ever been tempted with the ccd47 reducer?

[vlaiv]

Ok, so here are couple of measurements that I took from some of my subs:

Ha filter, night of good seeing - 4.7px at ~0.483"/px (I round that up to 0.5"/px when talking about it, but it is closer to 0.48"/px), so that gives:  ~2.27" FWHM

OIII filter, night of good seeing - 4.63px - ~ 2.24" FWHM

Ha filter on a night of poor seeing - 7.87px = ~3.8" FWHM

Lum filter (actually IDAS LPS P2), poor seeing - 5.91px = ~2.86" FWHM

We could say that it ranges from 2.2" FWHM to 3.8+" FWHM, or expressed in effective resolution - 1.375"/px to over 2.275"/px. One could probably fare worse if seeing is really bad - but who would want to image in those conditions?

Just for a comparison, here is theoretical expected FWHM for my conditions - 8" aperture, around 0.5" RMS guide precision, and let's say seeing in 1.2" - 2" range.

Star FWHM for ideal optics under 1.2" seeing and 0.5" RMS guiding should be 1.78" FWHM, while for 2" seeing is 2.39" FWHM. My values are a bit more than this, and while I get rather OK seeing forecast, I think it's down to local thermals that bump up my FWHM values (surrounded by houses and bodies of water - Danube river is right in my imaging path - about 1-2Km away).

This is what meteoblue forecasts for this evening, for example (gray column is seeing in arc seconds):

Like I said, even in best conditions, I'm still in 1.3"/px zone. Maybe things will change once I move to the countryside and to a bit higher elevation, and get better mount.

I have CCD47 reducer, tried it once, did not like it, but that could be due to tilt. In the mean time I upgraded my focuser on RC to 2.5" one with threaded connection. I will have to try again, but I'm not expecting much from it. It turns out that it will not correct fully if placed at distance that makes x0.67 reduction, so most people place it closer to get x0.7-x0.72 reduction. It's not corrector and if you apply x0.67 reduction, then 22mm diagonal will be close to 33mm circle, and anything above 30mm on that scope requires field flattening and correcting, even below 30mm might not be very good.

If I were to look at reducer for RC again (and I will at some time in future) - I would look at x0.75 Riccardi FF/FR. It seems to work with this scope as well and I think it works good (at least that is what I read about the combo).