Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Shorten total integration time with new equipment

astrophotography equipment
  • Please log in to reply
133 replies to this topic

#101 ChrisWhite

ChrisWhite

    Fly Me to the Moon

  • *****
  • Posts: 6,142
  • Joined: 28 Feb 2015
  • Loc: Colchester, VT

Posted 19 August 2020 - 04:24 PM

I may be wrong, but I believe SNR on subs has the same diminishing returns.

 

Um, reread my posts.  This is the point I've been trying to make.... Short or Long... the diminishing return is the SAME.  Its a function of the rate of signal acquisition and the rate of noise acquisition whether in long subs or short subs, many or few....  You must use TOTAL integration time as a reference here. 

 

Thats why many of us prefer longer subs than shorter as long as we are not clipping stars.  You get to have your cake and eat it too... and in some cases, bring out fainter signals. 


Edited by ChrisWhite, 19 August 2020 - 04:26 PM.

  • rockstarbill likes this

#102 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,516
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 19 August 2020 - 04:31 PM

I totally agree in regards to maximizing exposure time per sub, as much as possible... for me this is the most efficient use of my imaging time.



#103 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,516
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 19 August 2020 - 04:33 PM

Just so people know, when I say shorter, I mean, less than the hour cited here smile.gif

 

My subs are typically 300 seconds on my F5.5 scope... (with a triad ultra)

 

I have an Epsilon coming in at F3.3, so i'll do the math there and see when diminishing returns kicks in and experiment with that.

 

I myself, haven't gone so crazy to try shorter subs and massively more files to integrate, but.. .i may give it a whirl sometime. Just want to make it clear that the crazy example I cited to me, was just in opposites of the other end of time with 1 hour subs

 

SGP download times are criminal.

 

With NINA it asynchronously downloads, dithers and settles before SGP finishes downloading. 

Acutally, SGP downloads quickly, it's everything else that takes it time to accomplish: dither, temp comp, guider settling, etc.



#104 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 04:35 PM

Um, reread my posts.  This is the point I've been trying to make.... Short or Long... the diminishing return is the SAME.  Its a function of the rate of signal acquisition and the rate of noise acquisition whether in long subs or short subs, many or few....  You must use TOTAL integration time as a reference here. 

 

Thats why many of us prefer longer subs than shorter as long as we are not clipping stars.  You get to have your cake and eat it too... and in some cases, bring out fainter signals. 

I still question this.

 

Lets talk read noise and signal:

 

A 1e read noise camera overcomes the read noise at < 10 seconds (period.. swamped)

 

A 3e read noise camera overcomes (swamps) the read noise at 100 seconds by diminishing returns but 1000 seconds in total.. you can run at unity gain and trade some well depth to drop this down on cmos to something closer to 1.3

 

A 10e read noise camera doesn't swamp readnoise until > 1000 second subs.

 

 

So a CMOS will capture more photons over the readnoise earlier than a CCD that is still swamping readnoise.  I think this entirely supports my SNR measurement of time/efficiency

 

Do i need to cite more?

 

So swamping read noise on a CMOS doesn't have the same impact to SNR as it does on a CCD



#105 ChrisWhite

ChrisWhite

    Fly Me to the Moon

  • *****
  • Posts: 6,142
  • Joined: 28 Feb 2015
  • Loc: Colchester, VT

Posted 19 August 2020 - 04:40 PM

bangbang.gif

 

Im out...


  • Midnight Dan likes this

#106 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 04:40 PM

I totally agree in regards to maximizing exposure time per sub, as much as possible... for me this is the most efficient use of my imaging time.

This is exactly what SNR math does, does it not? 

 

This works for me too - with the caveat i only shoot OSC so i dither or die... that's time i'll spend anyway :)



#107 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 04:41 PM

bangbang.gif

 

Im out...

too bad.

 

Like i said, maybe i don't understand it... i could learn.. but instead of people answering my questions they get mad.  My example of the electron read noise in sensor and its relationship to Sub time and SNR was an open question that is worthy of a response.


Edited by sn2006gy, 19 August 2020 - 04:42 PM.


#108 rockstarbill

rockstarbill

    Cosmos

  • *****
  • Posts: 7,564
  • Joined: 16 Jul 2013
  • Loc: Snohomish, WA

Posted 19 August 2020 - 04:41 PM

The one hour sub was just an example of two data sets taken with the same system. Beyond that, it can be ignored. No one is saying go take hour long subs. 



#109 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,516
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 19 August 2020 - 04:41 PM

This is exactly what SNR math does, does it not? 

 

This works for me too - with the caveat i only shoot OSC so i dither or die... that's time i'll spend anyway smile.gif

For total integration time yes, but not for the exp time per sub.



#110 rockstarbill

rockstarbill

    Cosmos

  • *****
  • Posts: 7,564
  • Joined: 16 Jul 2013
  • Loc: Snohomish, WA

Posted 19 August 2020 - 04:44 PM

I still question this.

 

Lets talk read noise and signal:

 

A 1e read noise camera overcomes the read noise at < 10 seconds (period.. swamped)

 

A 3e read noise camera overcomes (swamps) the read noise at 100 seconds by diminishing returns but 1000 seconds in total.. you can run at unity gain and trade some well depth to drop this down on cmos to something closer to 1.3

 

A 10e read noise camera doesn't swamp readnoise until > 1000 second subs.

 

 

So a CMOS will capture more photons over the readnoise earlier than a CCD that is still swamping readnoise.  I think this entirely supports my SNR measurement of time/efficiency

 

Do i need to cite more?

 

So swamping read noise on a CMOS doesn't have the same impact to SNR as it does on a CCD

Please cite the math used to determine the exposure times please. 



#111 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 04:52 PM

Please cite the math used to determine the exposure times please. 

 

I started typing it all out... but have at it here

 

https://docs.google....dit?usp=sharing



#112 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 04:53 PM

You can save the spreadsheet to your local google drive and mess with the numbers. Scroll to the right see the graphs. or click the tabs on the bottom to see examples.



#113 rockstarbill

rockstarbill

    Cosmos

  • *****
  • Posts: 7,564
  • Joined: 16 Jul 2013
  • Loc: Snohomish, WA

Posted 19 August 2020 - 04:55 PM

So here are calculations on my imaginary ASI6200 that I can change the read noise on, to whatever I want it to be, without affecting anything else on the camera.

 

To swamp the read noise and do nothing else (so basically 1xRN^2) here is what that looks like:

 

1e Noise - 504 Median ADU

3e Noise - 536 Median ADU

10e Noise - 900 Median ADU

 

For broadband, none of these camera configurations are going to need 1000 second images to achieve the swamping goal of 1x.



#114 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 04:57 PM

So here are calculations on my imaginary ASI6200 that I can change the read noise on, to whatever I want it to be, without affecting anything else on the camera.

 

To swamp the read noise and do nothing else (so basically 1xRN^2) here is what that looks like:

 

1e Noise - 504 Median ADU

3e Noise - 536 Median ADU

10e Noise - 900 Median ADU

 

For broadband, none of these camera configurations are going to need 1000 second images to achieve the swamping goal of 1x.

 

Broadband on bright targets is easy math and almost "just follow what everyone else does" works regardless.

 

I pretty much assume complexity comes in with narrowband objects or broadband + massive light pollution.

 

In my case with OSC+light pollution, there is a point i've hit in which the signal rate of my faint object per sub isn't going to increase with as much signal until i take another sub..  i guess one could add the math for full moon and see how that would impact narrowband too if they want. (ha is ok in full moon but i've see si/oiii suffer... stuff is so faint anyway)

 

again.. a lot changes in dark skies.


Edited by sn2006gy, 19 August 2020 - 05:03 PM.


#115 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,516
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 19 August 2020 - 05:06 PM

You can save the spreadsheet to your local google drive and mess with the numbers. Scroll to the right see the graphs. or click the tabs on the bottom to see examples.

You have done a good job with that model, but that model is not based on a function of total integration time, you have fixed T at 10 hours.... what is most relevant to me is what is the most efficient total integration time for the target I am after in my skies with my imaging train, to capture the faintest signals I am after...  your model doesn't answer that.  Once I know the total integration time required (the minimum) then I can determine my sub times based on RN dilution and maximizing exp time without clipping or guiding limitations... thats how I see it.


  • rockstarbill likes this

#116 rockstarbill

rockstarbill

    Cosmos

  • *****
  • Posts: 7,564
  • Joined: 16 Jul 2013
  • Loc: Snohomish, WA

Posted 19 August 2020 - 05:08 PM

Broadband on bright targets is easy math and almost "just follow what everyone else does" works regardless.

 

I pretty much assume complexity comes in with narrowband objects or broadband + massive light pollution.

 

In my case with OSC+light pollution, there is a point i've hit in which the signal rate of my faint object per sub isn't going to increase with as much signal until i take another sub..  i guess one could add the math for full moon and see how that would impact narrowband too if they want. (ha is ok in full moon but i've see si/oiii suffer... stuff is so faint anyway)

 

again.. a lot changes in dark skies.

Well, in that case -- diminishing returns are not something anyone should consider with narrowband imaging. At all. So I am still confused as to what your point is? The goal posts seem like they keep shifting. lol.gif



#117 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 05:09 PM

You have done a good job with that model, but that model is not based on a function of total integration time, you have fixed T at 10 hours.... what is most relevant to me is what is the most efficient total integration time for the target I am after in my skies with my imaging train, to capture the faintest signals I am after...  your model doesn't answer that.  Once I know the total integration time required (the minimum) then I can determine my sub times based on RN dilution and maximizing exp time without clipping or guiding limitations... thats how I see it.

10 hours is just a random normalization time.

 

Most of the graphs if you scroll to the right on the SNR examples tab show enough data to see that 5 hours is good enough.

 

This was theory crafting with a lot of people when we were competing to image the infamous squid that takes 10-40 hours depending on your gear.



#118 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 05:10 PM

Well, in that case -- diminishing returns are not something anyone should consider with narrowband imaging. At all. So I am still confused as to what your point is? The goal posts seem like they keep shifting. lol.gif

Well, i feel like we're talking about one thing, then its about something else. On one post you mention a KAF, on another you mention 6200. On some we're talking about a 1 hour narrowband image, on another, we're talking about broadband.

 

I'm just trying to keep up.

 

Check out the spreadsheet.. my wife is wondering what i'm still doing on the computer. Time for me to go grill up some dinner smile.gif

 

The worksheet lets you calculate your sensor, you Light Pollution, Your read noise and if you know how much your filters impact your signal - also graphs your approximate SNR over time... it's pretty cool and helpful.

 

It does support for a lot of things that the generally accepted times work, but that doesn't make these numbers any less helpfull.

 

But again... i could be all wrong.  My math could be wrong. My assumptions could be wrong. I help people go on vacation by running massive distributed compute systems so this is just me applying my passion for metrics and observability to astronomy.


Edited by sn2006gy, 19 August 2020 - 05:12 PM.


#119 rockstarbill

rockstarbill

    Cosmos

  • *****
  • Posts: 7,564
  • Joined: 16 Jul 2013
  • Loc: Snohomish, WA

Posted 19 August 2020 - 05:10 PM

You have done a good job with that model, but that model is not based on a function of total integration time, you have fixed T at 10 hours.... what is most relevant to me is what is the most efficient total integration time for the target I am after in my skies with my imaging train, to capture the faintest signals I am after...  your model doesn't answer that.  Once I know the total integration time required (the minimum) then I can determine my sub times based on RN dilution and maximizing exp time without clipping or guiding limitations... thats how I see it.

And this, is correct. 



#120 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,516
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 19 August 2020 - 05:14 PM

10 hours is just a random normalization time.

 

Most of the graphs if you scroll to the right on the SNR examples tab show enough data to see that 5 hours is good enough.

 

This was theory crafting with a lot of people when we were competing to image the infamous squid that takes 10-40 hours depending on your gear.

What your model shows is that sub exp time is a function of total integration time, and for me that is not the best approach.



#121 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 05:15 PM

actually i may have deleted the graphs over time.. most are on sub length. i'll re-add those



#122 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 05:17 PM

What your model shows is that sub exp time is a function of total integration time, and for me that is not the best approach.

huh?

 

With the graphs, I can clearly see when diminishing returns kicks in for total integration time. If it's still a positive growth and i'm not hapy with one nights integration, i'll go for 2.  If going for 2 is a waste of time for my conditions, i'll schedule that for a dark site.

 

I had to have some time to use as a basis, that is all.

 

This stuff helped me find out imaging iris in my backyard was a lost cause... so i did that at the astronomy club site.

 

I think for that kind of exploration it makes perfect sense.


Edited by sn2006gy, 19 August 2020 - 05:17 PM.


#123 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 05:19 PM

I'm fine if we agree to disagree here.. it works for me.

 

It's not a law of acquisition.. it's just a useful tool to know if i'm hitting a light pollution wall, sensor wall, filter wall, integration wall, sub time wall.

 

I experiment every night if i'm not happy with what i forecasted doing the math. My neighbors will turn on their flood lights and blow everything out of the water or we'll have a random "above average night" and i'm done.


Edited by sn2006gy, 19 August 2020 - 05:20 PM.


#124 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 942
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 19 August 2020 - 05:21 PM

To close this back to the topic, the information in this spreadsheet could be used to calculate shortening of integration time.

 

I broke my brain trying to incorporate an amazing etendue spreadsheet and the optimal scope/sensor config... maybe another day.



#125 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,516
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 19 August 2020 - 05:24 PM

why does a globular cluster typically require much less total integration time than a faint galaxy?  two things I wanted to learn about:

 

1)  how long should total integration time be for a particular target?

 

2)  how long should each sub exposure be?

 

(1) and (2) are independent as far as I am concerned, but your model links them together.

 

We are just using different approaches, no worries.

 

and....

 

I am sure Frank is going to log in and show us how it is done!!




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics





Also tagged with one or more of these keywords: astrophotography, equipment



Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics