A bit noisy

[irtuk]

Morning all  Hope anyone else in the south east is enjoying the recent clear nights we seem to have been blessed with.

I decided to have a crack at the Elephant Trunk Nebula yesterday and my 10 minute subs are looking like this, through HA filter, does it look a bit too noisy?

Its a ZWO 1600MM mono cooled to -10 but I didn't have 10 minute darks in my library so I ran a single frame through APP this morning using 5 minute darks, I am shooting some 10 minute darks right now.

So, is this noise or is it nebulosity?

The tracking is a bit off as well, something else to be working on

Ed

[vlaiv]

Calibrating with proper darks will make a difference, also, from the file title - it looks like you had you gain set at 0?

Set it at least to unity when doing NB images - that will significantly reduce read noise (from 3.4e down to 1.7e - halving it).

You also seem to use long focal length scope? You are oversampling by quite a bit - binning your data will again improve SNR, and yes, single sub will be noisy, stacking them improves SNR again.

[irtuk]

Thanks for the input.

What's "unity" gain ? I am using the ZWO ASIImg software and the gain options are low, middle and high. 

/edit/ let me just google that for myself shall i found some excellent explanations...

The scope is an Altair Astro 102mm ED and the wife would be simply overjoyed to hear that I require a different type of telescope for this work

 

Edited September 7, 2021 by irtuk

[david_taurus83]

Was this last night or Sunday night? As Sunday night was very bad for high hazy clouds. I was trying to setup and focus an OAG on Sunday! Subs were also very noisy. Last night was much better as should tonight be.

[vlaiv]

2 hours ago, irtuk said:

The scope is an Altair Astro 102mm ED and the wife would be simply overjoyed to hear that I require a different type of telescope for this work

What mount are you using? Sampling rate is 1.1"/px - both from 714mm of FL and when "platesolved" from image. That is not very high sampling rate, however - your stars in the image are huge for some reason.

Maybe there was issue with focusing or perhaps seeing was particularly bad on given night.

In any case - I'd recommend that you bin your images x2 in software (2.2"/px is much better resolution for 100mm of aperture) - and this one, maybe even x3.

[newbie alert]

6 hours ago, vlaiv said:

Calibrating with proper darks will make a difference, also, from the file title - it looks like you had you gain set at 0?

Set it at least to unity when doing NB images - that will significantly reduce read noise (from 3.4e down to 1.7e - halving it).

You also seem to use long focal length scope? You are oversampling by quite a bit - binning your data will again improve SNR, and yes, single sub will be noisy, stacking them improves SNR again.

Vlaiv.. what's your take on over sampling limit, I've read before that you say sampling at 1.6 is optional , nyquist theory says between .67 and 2 being optimal

 

[newbie alert]

Irtuk, you're not to far from me.. Are you a member of SEKAS in Dover at all?

[vlaiv]

32 minutes ago, newbie alert said:

Vlaiv.. what's your take on over sampling limit, I've read before that you say sampling at 1.6 is optional , nyquist theory says between .67 and 2 being optimal

 

Nyquist sampling theorem is very precise in what it says - it does not say anything between two values being optimal. It says that one should sample band limited signal at twice the highest frequency component.

In order to apply Nquist sampling theorem - we need to have certain conditions - namely band limited signal. In general case, when telescopes are concerned, this condition is strictly met only for planetary imaging. Aperture of telescope and its effects on diffraction of light create image in focal plane that is band limited.

Here is this visually represented (or rather in graphs):

Left image is Airy pattern cross section and right image is MTF. Left is effect of aperture on single point of light and right is representation of that filter in frequency domain. In right image we can see that graph hits zero at one point (most right point) - that is our cut off frequency and it shows that we have band limited signal - image produced by telescope aperture does not contain higher frequencies that that and hence - Nyquist sampling theorem can be fully applied to that.

If we are talking about deep sky imaging - we no longer have above condition and we have to make number of assumptions (which are very reasonable assumptions).

I'll first show you equivalent graphical representation and then talk about approximation and numbers involved.

Here I don't even need to create composite image - I found one online because it is well known relationship.

In long exposure imaging point spread function of telescope looks like Gaussian curve - and that is our first approximation. If we consider central theorem - it is very sound approximation, and in fact - you can do fitting of Gaussian curve on star profile and it fits nicely most of the time (that is how we get FWHM, eccentricity and so on when we measure what our stars look like, it is also used to find peak for alignment of images and so on ...).

By the way - Central limit theorem says that sum of many random variables tends to normal / Gaussian distribution - and that is what long exposure makes - integration / sum of all seeing disturbances and mount tracking errors.

Back to Nquist - Gaussian star profile means that Fourier transform of it is also Gaussian in shape - that sort of poses problem for Nquist. That means that we don't have cut off frequency. Gaussian goes on forever and never reaches zero. It gets smaller and smaller - but never reaches zero - not even at "infinity" (it does tend to zero as x tends to infinity).

First approximation was Gaussian shape as PSF in our images.

Second approximation is to say that Gaussian curve falls low enough after a while so it does not make any impact on final image. We can say - after some frequency - all higher frequencies are simply too small to make significant contribution to image. This is where we approximate things for second time and we have choice of this number limit. I've chosen this number rather arbitrarily - but not quite so. I've chosen it so that you can still sharpen your data - but only "sensible" data - one with SNR > 5.

If you choose this limit in such way and you do the math and connect that to FWHM of our Gaussian profile - you get simple relationship that sample rate needs to be FWHM of star profile / 1.6. It takes all frequencies that are still at least 10% or original frequency (or was it 1%  - I can't remember now, but I can crunch the numbers again to see which one was it at the time I derived this).

I then did simulations to confirm that my estimations and approximations make sense - and they do. That is how I derived factor of 1.6 with respect to FWHM.

If you wish - we can do the math step by step so you can see the actual derivation?

 

[newbie alert]

Cheers Vlaiv...too much for my little brain I'm afraid...most of it was whoosh, and it's over my head...like I've said before, on another level...

[vlaiv]

39 minutes ago, newbie alert said:

Cheers Vlaiv...too much for my little brain I'm afraid...most of it was whoosh, and it's over my head...like I've said before, on another level...

 

Ok, so to simplify things - when doing DSO imaging - you should be around FWHM / 1.6 for your sampling - meaning if you measure FWHM of 3.2" then you should be at 2"/px as 3.2 / 1.6 = 2. This is only approximation but very good one.

This is not the same as planetary imaging where you adjust your F/ratio based on frequency that you work with (say about 520nm for color images or 656nm for Ha - or 850nm for example for IR - depends on what filter you are using) and pixel size of your camera and there is exact "solution" based on Nyquist without any approximations.

[irtuk]

On 07/09/2021 at 18:07, newbie alert said:

Irtuk, you're not to far from me.. Are you a member of SEKAS in Dover at all?

No I'm not as it happens, but I think I bought the scope from the SEKAS webmaster, down Herne Bay kinda area.

 

Ed.

 

[irtuk]

On 07/09/2021 at 15:07, vlaiv said:

What mount are you using? Sampling rate is 1.1"/px - both from 714mm of FL and when "platesolved" from image. That is not very high sampling rate, however - your stars in the image are huge for some reason.

Maybe there was issue with focusing or perhaps seeing was particularly bad on given night.

In any case - I'd recommend that you bin your images x2 in software (2.2"/px is much better resolution for 100mm of aperture) - and this one, maybe even x3.

I'm using a SkyWatcher NEQ-6, ZWO ASI 1600MM, ZWO Filterwheel, this package from FLO as it happens, there's a field flattener in there, Pegasus focus cube, Pegasus Powerbox, small 50mm guidescope with an altair camera.

[newbie alert]

6 hours ago, irtuk said:

No I'm not as it happens, but I think I bought the scope from the SEKAS webmaster, down Herne Bay kinda area.

 

Ed.

 

Yes, I know who you mean....cracking bloke