Jump to content

sgl_imaging_challenge_2021_annual.thumb.jpg.3fc34f695a81b16210333189a3162ac7.jpg

Cheapest setup for SNR5 "standard"


Recommended Posts

I'm sort of trying to invent some criteria for "speed" of live stacking rig and I came up with SNR5 "standard".

I'll briefly describe what that "standard" is and how it is calculated and want to see what rigs satisfy it - as being at least SNR5 capable, and how much such rigs would cost - not including computer, but including scope, mount, camera and all bits and pieces needed to tie it up together and make it work (well maybe not power source and things like that?).

Here is what goes into SNR5 "standard" - it is rig that:

1. has at least one degree diagonal coverage of the sky

2. has sampling rate around 1.5"/px (I guess we can say +/- 0.2 or so but I think we really want to say at least 1.5"/px or if we allow it to go higher - at least 1.6"/px - lower res benefits SNR but looses detail, and in reality it is unlikely that we will achieve higher detail than 1.5"/px with these setups).

3. Is capable of doing color rendering of the target

4. Needs to achieve SNR of at least 5 in following conditions:

- one minute total exposure in 4s subs

- in 20mag skies

- of 20mag target

(20/20 - easy to remember)

Calculations include following:

Mag 0 sky / target produce 880000 photons per second. All telescope losses are accounted for - This means we calculate clear aperture (central obstruction is subtracted if present, mirror reflectivity is accounted for, and glass transmission also). QE of sensor is taken to be 40% of that declared to be max QE and final result is multiplied by 1.41 - this is because of the way color sensors work and the way we can obtain Luminance layer from the data (we have two pixels of green color - hence 1.41 factor, green is about 1/3 as sensitive as peak QE, but other two channels contribute as well so we use 0.4 instead of 0.33 - this is rough estimate but enough for "standard"). For mono sensors + filters - only mono is used and "average" QE is used (this means if peak QE is 80% we will use something like 60-65% depending on where endpoints sit).

Read noise which can be achieved is used - not baseline, since anyone doing 4s exposures is likely to use high gain. If there is binning - we will treat it as software binning and read noise is increased with factor of "base" of binned pixels - for 2x2 binning - we have x2 read noise, for 3x3 binning - we have x3 read noise and so on....

Dark current is also taken into account - either at -10C if cooled camera is used or 25C if regular camera is used.

Here is my candidate:

EQ3 class mount

130PDS + CC

ASI294 cooled model.

This combination will achieve SNR5 with no binning.

Working resolution is 1.46"/px, assumed peak QE is 80% (although ZWO says it is TBD, I read somewhere that it is estimated to 80%) - so used QE was 0.32. I used CO of 36% (47mm) and mirror reflectivity of 94%. I also used read noise of 1.4 (although it can go down to 1.2e according to ZWO - I wanted to be safe there).

In any case stack of 15x4s subs gives SNR of 3.583 and multiplying that with 1.41 gives 5.05

All of that for only 1999.99e :D (give or take)

Can we do this cheaper?

 

 

 

Link to post
Share on other sites

HI Vlaiv, I don't think it can be done cheaper or that there are many reasonable setups that can even achieve "SNR5", but I think that some of your criteria are very strict.

Take for example the FOV. Although it would be nice to have 1°, this excludes smaller sensors, while most dso's are not bigger than 0.5°.

Then there is the sampling rate. While 1.5"/px, is where the seeing  limits you for decent AP under moderate skies, I feel that eeva is all about speed, so a bit of undersampling (up to 4"/px works fine with me) could be allowed.

As for the scope, F/5 would be OK, and we should really aim for workable (minimum aberrations) subF/4.

Link to post
Share on other sites
32 minutes ago, R26 oldtimer said:

HI Vlaiv, I don't think it can be done cheaper or that there are many reasonable setups that can even achieve "SNR5", but I think that some of your criteria are very strict.

Take for example the FOV. Although it would be nice to have 1°, this excludes smaller sensors, while most dso's are not bigger than 0.5°.

Then there is the sampling rate. While 1.5"/px, is where the seeing  limits you for decent AP under moderate skies, I feel that eeva is all about speed, so a bit of undersampling (up to 4"/px works fine with me) could be allowed.

As for the scope, F/5 would be OK, and we should really aim for workable (minimum aberrations) subF/4.

Now when I think about it - it is rather strict isn't it?

What do you think should be reasonable criteria?

Should we still maintain SNR5 20/20 in one minute or relax it in terms of SNR, or time achieved or something else? And what should be criteria on FOV and resolution?

I mean - I've figured out really fast setup but it fails on couple of my points above. AzGTI + Samyang 135mm F/2 + ASI178mc cooled.

It is 3.67"/px and FOV is something like 3° x 2°

That one is SNR8 for 20 / 20 / 15 x 4s exposures. If we swap cooled for regular ASI178mc (as cooled is no longer produced and to save some cash) we are looking at about 1300e and SNR7 (assuming dark current as high as 1e/s/px - although I have no idea what it might actually be - I doubt it is as high as that).

 

Link to post
Share on other sites

Hi Vlaiv, sorry for the late response.

On 02/02/2020 at 23:20, vlaiv said:

And what should be criteria on FOV and resolution?

I've seen very good results with resolution as low as 3-4 "/px , and certainly a  FoV of 0.5° is pretty much decent.

So, one can use small sensitive sensors ( like zwo asi224, zwo asi290, lodestarX2c) with small focal lengths (telephotos/ 60,70,80mm refractors (preferably ed) with or without 0.5 reducer/ small newts 114p/500 or 130p/650 or heritage) on a small mount (like the azgti or eq3goto).

Or go to a camera with a bigger sensor (like zwo asi294 or atik infinity) with a bigger focal length (<4" triplet refractors/ sct's with 0.6 focal reducers / 6"f5, 8"f4 or even faster newtonians like the sharpstar F2.8 or the Tak ε-180) on a bigger mount.

On 02/02/2020 at 15:05, vlaiv said:

4. Needs to achieve SNR of at least 5 in following conditions:

- one minute total exposure in 4s subs

Although my lodestar shows almost everything in a minute, it is a bit noisy and have to give it 2-3mins for faint targets. Also the cmos cameras, even the very sensitive 224, needs more time to build up a decent image, so I think the total exposure limit, would be better at 2-3mins. Still very fast and still eeva in my book but as they say, YMMV.

As for the subexposure length, again I find it varies greatly depending on the target and usage of UHC or NB filter, so no need to restrict that as long as you've set a total integration time.

  • Like 1
Link to post
Share on other sites
  • 3 weeks later...

Missing from these cost calculations are laptop or mini-computer.  I know that is often considered a standard requirement with any camera because we have long been accustomed to (say) 1.2 megapixel CCDs. However, I urge caution about this requirement with data hungry large sensor, high resolution CMOS like ASI294, ASI1600 or Atik Horizon (BTW, the Horizon read noise  is astonishingly low).

To comfortably run a ZWO ASI294 (10.7 megapixel) or my Atik Horizon (16 megapixel) even in 1920 x 1200 resolution (n.b. both cameras can handle a 4K UHD output), my advice is that you are best served by a computer with a decent CPU and at least 8 GB RAM if processing stacked 4 second exposures. It can be done 'slower' with lesser CPU power/RAM, but if as a result, processing time per exposure exceeeds actual exposure time, then a minute total integration time (24 x 4 seconds) inevitably becomes a lot longer. That might be fine for 'imaging' (........some folk wait all night for long exposures!), but can be a tad frustarting for 'observing' and EEVA now seems to embrace enthusisats at each end of the spectrum. I am an observer and seek 'real time'. Probably without much justification in truth, but don't you just hate the stacking progress bar being so slow?

A problem is that laptops and minicomputers tend to 'steal' system RAM if under intensive graphics load causing camera sofware to run more slowly. Much depends on graphics card, but their on-board graphics are often inferior to desktops. Frankly, I found a 7i5-2600 processor with Iris Plus 640 Graphics would often crash due to inadequate resources with only 4 Gb RAM. It would run with 8Gb RAM, but still quite slowly, and it took a 8i7-8550 with 16 Gb before it ran in 'real time'. Some mini-computers (like cheaper NUCs) are supplied with only 4GB RAM so it is easy to be caught out. 

A contributory factor is Windows 10 demands signifcant system resources. LINUX and can reduce this load. My point is that when considering budget in any ultimate system, with the advent of CMOS, choice of computer is becoming more significant. Hence, that battered old laptop might not suffice, albeit that if running (unsupported) Windows 7 it is probably faster than if running Windows 10.

 

  • Like 1
Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. By using this site, you agree to our Terms of Use.