Go Back   IceInSpace > Equipment > Astrophotography and Imaging Equipment and Discussions

Reply
 
Thread Tools Rating: Thread Rating: 5 votes, 5.00 average.
  #1  
Old 15-03-2013, 07:21 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
DSO broadband exposure strategy - Narrowband added - total exposure time added

There are quite a few websites that recommend sub exposures based on optimising SNR. The general assumption seems to be that longer is always better, but that there is a practical optimum sub exposure, which is reached when the read noise is 5% of the total noise (eg the Starizona approach).
edit: here are some web references: http://starizona.com/acb/ccd/advtheoryexp.aspx http://www.starrywonders.com/snr.html http://www.hiddenloft.com/notes/SubExposures.pdf
This is perfectly reasonable if the sole objective is to minimise imaging time. However, if the objective is taking pretty pictures, then dynamic range also comes into it and long sub exposures will not be the best approach. If you wish to retain star image colour, you must minimise star saturation. This can be assisted by taking numerous short subs. For example stacking 10 x 9minute subs which each have 20,000 well depth gives 200,000 effective well depth – stacking 30 x 3 minute subs gives 600,000 effective well depth, so the signal is the same, but the stars have much more headroom. The downside is that 30 subs have more read noise than 10 subs, so dynamic range and SNR are reduced with shorter subs. SNR and good star colour are competing requirements.

The star colour issue can be quantified by modifying the dynamic range calculation to include sky noise, yielding:

Headroom = 2.51 * Log10((N * welldepth) / sqrt(N * readnoise^2 + N * skyelectrons))
where N is the number of subs and dark current is ignored. Skyelectrons is the number of sky background electrons per sub.

This parameter is essentially an indicator of how many mags above the background noise a star can be before it saturates (no account is taken of the star PSF, but the measure is fine for comparative analysis). To illustrate, the attached graphs show SNR (standard formula) and headroom plots for three different systems with ~the same aperture and pixel scale. The sky is 125 photons/s/m2/arcsec2 (fairly dark) and the extended target is 25 photons/s/m2/arcsec2 (about like NGC247). SNR and headroom are those for 3 hours of combined subs – the sub length is shown on the x axes, not the total time.

The table shows the standard recommended sub exposure (read noise 5%) and also the much shorter sub exposure for which read noise equals sky noise – this is probably a marginally useful lower bound at which total exposure time would need to be ~doubled to get the same final SNR as from "5%" subs.

With all of the systems, it is clearly possible to gain at least one mag in headroom by reducing sub exposure length and accepting a slightly lower SNR (which could be recovered by longer overall imaging time). One mag extra headroom would reduce the number of saturated stars to about 1/3 of that at the longer sub exposure. Thus, the standard "5%" approach should probably be taken as the maximum sub exposure length and imagers could consider significantly shorter subs if star colour is an issue.

The table is also very instructive WRT the wide variety of sub exposures required for the various technologies. For the same scenario, recommended KAI11002 system sub exposures are 930 seconds, whereas the icX694 only needs 110 seconds - clearly the choice of appropriate sub exposures is important to get the best out of any CCD technology.

I guess in summary then, long subs are not the way to go if you are looking for pretty broadband colour pictures, particularly when using low noise/low well depth CCDs. Haven’t looked at narrowband yet and would be very grateful for any feedback on this approach.

Thanks for taking the time to read. Regards Ray
Attached Thumbnails
Click for full-size image (summary.jpg)
133.8 KB74 views

Last edited by Shiraz; 28-03-2013 at 07:28 AM. Reason: rewrite and condense
Reply With Quote
  #2  
Old 17-03-2013, 06:13 PM
SpaceNoob (Chris)
Atlas Observatory

SpaceNoob is offline
 
Join Date: May 2012
Location: Canberra
Posts: 268
Makes sense, I tend to use ~300sec sub exposures on my 8300 sensor @ f/5.
Reply With Quote
  #3  
Old 17-03-2013, 07:15 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Hi Chris - thanks, appreciate the feedback - seems like your experience matches quite well with the predictions. unpublished prediction for my QHY8 setup also matched the sub exposure that I had settled on using trial and error, so 2/2 to date . Regards Ray.

Last edited by Shiraz; 17-03-2013 at 07:25 PM.
Reply With Quote
  #4  
Old 17-03-2013, 07:36 PM
RobF's Avatar
RobF (Rob)
Mostly harmless...

RobF is offline
 
Join Date: Jul 2008
Location: Brisbane, Australia
Posts: 5,716
Great overview of a complex subject Ray. I don't think I've ever kept up with a mathematical explanation for sub time better. I have a 8300 chip (QHY9) and usually shoot 5mins subs, sometimes down to 3, so intuitively seems close to what you calculate.

Would be interested in the NB (particularly Ha) calcs.
Pulling up seat in this room to listen in.....
Reply With Quote
  #5  
Old 17-03-2013, 08:51 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Ray,

What about the option of using HDR to combine a small stack of short exposures with the "normal" data? Since you're only capturing the bright areas that saturate on a longer exposure you don't need much data to get good SNR, so the time to collect the short subs is not a great impost. And then you don't need to compromise on the length of the normal subs.

Cheers,
Rick.
Reply With Quote
  #6  
Old 17-03-2013, 09:08 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
hi Rick. What a neat idea. so by non-linear blending of short exposure data for the bright bits you get better star colour yet retain full sub lengths - yep, that works. The only downside I can think of is that the short exposure stars would be smaller than the saturated regions they replace, but that would not be impossible to fix.

Of course, there are some secondary reasons why it is useful to reduce sub exposures below the SNR optimum - easier tracking, fewer satellites, wind problems reduced etc, but there is still a strong case for minimising imaging time with full length subs.

Anyway, many thanks for introducing such a great idea. Do you think commerical HDR software would do the job?. Ray

Last edited by Shiraz; 17-03-2013 at 09:25 PM.
Reply With Quote
  #7  
Old 17-03-2013, 09:12 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by RobF View Post
Great overview of a complex subject Ray. I don't think I've ever kept up with a mathematical explanation for sub time better. I have a 8300 chip (QHY9) and usually shoot 5mins subs, sometimes down to 3, so intuitively seems close to what you calculate.

Would be interested in the NB (particularly Ha) calcs.
Pulling up seat in this room to listen in.....
Hi Rob, that's a very generous comment - thanks.
good to hear that your experience fits as well - 3/3.
Am fairly well along with the Ha model and will publish soon - added in dark current, but it didn't do exactly what I though it would - want to be sure I understand why before posting. regards Ray
Reply With Quote
  #8  
Old 18-03-2013, 12:29 AM
naskies's Avatar
naskies (Dave)
Registered User

naskies is offline
 
Join Date: Jul 2011
Location: Brisbane
Posts: 1,865
Thanks for writing that up, Ray - very interesting read. Along Rick's suggestions, I'd be interested in a theoretical optimal sub exposures for say the Orion nebula where you need (with most CCDs) to take multiple exposures to capture both the faint background dust and the core.
Reply With Quote
  #9  
Old 18-03-2013, 07:42 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by Shiraz View Post
hi Rick. What a neat idea. so by non-linear blending of short exposure data for the bright bits you get better star colour yet retain full sub lengths - yep, that works. The only downside I can think of is that the short exposure stars would be smaller than the saturated regions they replace, but that would not be impossible to fix.

Of course, there are some secondary reasons why it is useful to reduce sub exposures below the SNR optimum - easier tracking, fewer satellites, wind problems reduced etc, but there is still a strong case for minimising imaging time with full length subs.

Anyway, many thanks for introducing such a great idea. Do you think commerical HDR software would do the job?. Ray
Thanks, Ray. I have done this with the HDRComposition process in PixInsight successfully. I haven't tried any of the normal photographic HDR programs to see if they would work. I prefer to process in FITS format right up until I export a final image.

Cheers,
Rick.
Reply With Quote
  #10  
Old 18-03-2013, 11:25 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by naskies View Post
Thanks for writing that up, Ray - very interesting read. Along Rick's suggestions, I'd be interested in a theoretical optimal sub exposures for say the Orion nebula where you need (with most CCDs) to take multiple exposures to capture both the faint background dust and the core.
stars are not in the model Dave, but can do a quick and dirty back of the envelope on M42 -

Assuming that M42 is about mag 21 in the outer reaches – same as a fairly dark sky.

The optimum SNR calculation for the nebula calls for sub exposures of about 300 secs with the 254f4.75 8300, which yields a headroom of 9.4 mag. From available info on the distribution of surface brightness in M42, this headroom should cover the hot bits of the nebula, but not the brightest stars.

With a sky of mag 21 and mag 9.4 headroom, stars will saturate at mag 11.6 if all of the star energy is contained in 1 pixel. In 2 arc sec seeing, star energy will be distributed over about 4 pixels, so saturation will occur for stars of 4x mag 11.6, or mag 10.1 for 300 sec sub exposures.

Since the brightest star in the trapezium is mag 5.1, we have to reduce the exposure by the equivalent of mag 5, which is about 1/100 to keep that star from saturating. So, for broadband imaging, the trapezium exposure would need to be less than 3 seconds. The energy in each colour channel would be roughly 1/3 that in broadband – so the trapezium stars should remain unsaturated up to ~9 seconds exposure in the colour subs, depending on the star colours.

I haven’t got any of my old M42 data easily to hand to check if this is reasonable – is anyone able to comment on the basis of their data? Bear in mind though that this result is specific to the chosen equipment setup and sky/nebula brightness assumptions. and as a "quick and dirty" it might have glossed over an important factor and be just plain wrong

Your question also raises the prospect of doing some genuine hi res lucky imaging on the trapezium and similar - these stars could be imaged at well below the saturation level at high framerates (~10% of saturation in 0.3 sec for example). will have to follow this up - could be fun.

Last edited by Shiraz; 18-03-2013 at 11:23 PM.
Reply With Quote
  #11  
Old 25-03-2013, 08:13 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
As a complement to an earlier broadband analysis, have completed a Halpha analysis to determine optimum narrowband sub length. Turned out to be a bit of a nugatory exercise, since the basic rule of thumb for narrowband is “make the subs as long as you possibly can” – and that’s it. The limiting factor that determines how long your subs can be will probably be one of:
• Tracking consistency,
• Polar alignment accuracy/drift,
• Meridian flip/altitude,
• Satellites, clouds, wind, seeing variability
• Cable snags etc.

In doing the analysis to come to this conclusion, the starting assumptions were:
• the narrow band sky brightness is 5 p/s/m2/arcsec2 , which corresponds to about a 10nm filter and a fairly dark sky,
• the narrow band brightness of the target nebula in Ha is 10 p/s/m2/arcsec2 - this is probably realistic, but the value is not critical to the analysis in any case.

Three approaches to determining appropriate sub lengths were investigated:
1. Initially tried the standard broadband sub-exposure strategy of “read noise = 5% of total background noise” (See the earlier broadband discussion and the following websites: http://starizona.com/acb/ccd/advtheoryexp.aspx http://www.starrywonders.com/snr.html http://www.hiddenloft.com/notes/SubExposures.pdf ). This strategy can be applied to narrowband imaging with the proviso that total background noise includes dark current as well as Poisson noise. This approach can lead to extremely long optimum Ha sub exposures for low QE CCDs – to the point of impracticality.
2. As a more practical alternative, the second approach is based on the idea that sub exposures can be significantly reduced from the “5% noise” optimum by accepting a tolerable 20% increase in overall imaging time - if you use the specified shorter subs you will have to image for 20% longer to get the same SNR as from optimum subs. This is called the “20% more” sub length in the following comments. You could use some other SNR/time penalty, but 20% seemed reasonable.
3. The third measure shows the effects of a much shorter short sub length, where read noise equals other noise – this is the point where overall imaging time must be doubled to restore the “5%noise” SNR. This is called the “double” sub length.

The SNR and headroom plots vs sub exposure for the three demo systems used previously and also a KAF3200ME based system (also ~1 arc sec) are shown in the attached figures. Note that the SNR used here is the ratio between the nominal object signal and the background noise (rather than the object noise) – this measure was chosen to represents the visual quality of an image. As an illustrative analogy for the measure, imagine a grove of trees that is flooded for a new dam – the high SNR systems will have a low water level (noise), with most of the trees (nebulae) sticking out of the water, whereas the low SNR systems will have high water level and only the highest treetops (brightest nebulae regions) will be visible. ie the SNR used here indicates how deep a system will image before it runs into the noise floor.

The attached table shows the”5% noise”, the “20% more” and the “double” sub exposure times for each system - plus the associated headroom and SNR after 3 hours total exposure for the “20%more” option. In summary:
• The KAI11002M system has a relatively low QE at Ha and relatively high noise, so it requires long sub exposures – the standard “5% noise” sub is 3.9 hours, the “20% more” sub length is over an hour and the “double” sub time is 24 minutes.
• The KAF8300 system has moderately low noise and intermediate QE. The “5% noise” sub exposure is 2.2 hours, and the “20% more” sub is 47 minutes. The “double” sub is 18 minutes.
• The icx694 system has low noise and high QE. The “5% noise” sub exposure is 0.9 hours, but the low read noise of the CCD brings the “20% more” sub length back to 23 minutes and the ”double” sub time is 5 minutes.
• The KAF3200ME system has very high QE, but relatively high dark current. It requires 1.1 hours for “5% noise” subs, 28 minutes for “20% more” subs and the “double” sub time is 7 minutes.

Any of the systems will benefit from subs of an hour or so and the 11002 system would do best if the nominal 3 hour total exposure was a single sub. Of course, you could use short subs with any of the systems, but at the expense of significantly increased overall imaging time or decreased imaging depth (in effect, short subs throw away some of the advantage of narrowband imaging – narrowband gets rid of most of the background Poisson noise, but multiple short subs just put read noise in its place).

Note that the 11002 and 3200 systems are both heavily influenced by dark current at the assumed sky background level. I took the dark current specs from commercially available cameras, but better performance would be possible if these two chips were more deeply cooled. The 3200 system has the best headroom, but all systems probably have sufficient for most purposes due to the greatly reduced star brightness in narrowband.

I would greatly appreciate any feedback from anyone who has done some narrowband imaging with systems anything like those mentioned.

Thanks for taking the time to read. Regards Ray
Attached Thumbnails
Click for full-size image (summaryhalpha.jpg)
128.3 KB22 views
Click for full-size image (tablehalpha.jpg)
42.3 KB20 views

Last edited by Shiraz; 25-03-2013 at 09:22 AM.
Reply With Quote
  #12  
Old 25-03-2013, 09:51 AM
SpaceNoob (Chris)
Atlas Observatory

SpaceNoob is offline
 
Join Date: May 2012
Location: Canberra
Posts: 268
With advice from SBIG I have been using this calculator for my subs:
http://www.ccdware.com/resources/

I have found it useful. You'll need to adjust values to match your actual camera and sky data. Instructions are available on the top of the page.
Reply With Quote
  #13  
Old 25-03-2013, 10:31 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by Shiraz View Post
Three approaches to determining appropriate sub lengths were investigated:
1. Initially tried the standard broadband sub-exposure strategy of “read noise = 5% of total background noise” (See the earlier broadband discussion and the following websites: http://starizona.com/acb/ccd/advtheoryexp.aspx http://www.starrywonders.com/snr.html http://www.hiddenloft.com/notes/SubExposures.pdf ). This strategy can be applied to narrowband imaging with the proviso that total background noise includes dark current as well as Poisson noise. This approach can lead to extremely long optimum Ha sub exposures for low QE CCDs – to the point of impracticality.
Hi Ray,

I don't have time to dig through your post in detail right now, unfortunately, but it looks like a very interesting analysis.

In case you haven't seen it before, here's another take on determining optimal sub-exposure length: http://www.cloudynights.com/item.php?item_id=1622

Cheers,
Rick.
Reply With Quote
  #14  
Old 25-03-2013, 07:38 PM
RobF's Avatar
RobF (Rob)
Mostly harmless...

RobF is offline
 
Join Date: Jul 2008
Location: Brisbane, Australia
Posts: 5,716
And reading another post here by Peter Ward, and owning an 8300 chip camera myself, guess the other issue of oversaturation of bright stars on some fields if you go crazy with long NB sub times.....
Reply With Quote
  #15  
Old 26-03-2013, 07:54 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by SpaceNoob View Post
With advice from SBIG I have been using this calculator for my subs:
http://www.ccdware.com/resources/

I have found it useful. You'll need to adjust values to match your actual camera and sky data. Instructions are available on the top of the page.
thanks for the link Chris. Do you use the calculator for narrowband and if so, what sub exposure lengths does it suggest?

Quote:
Originally Posted by RickS View Post
Hi Ray,

I don't have time to dig through your post in detail right now, unfortunately, but it looks like a very interesting analysis.

In case you haven't seen it before, here's another take on determining optimal sub-exposure length: http://www.cloudynights.com/item.php?item_id=1622

Cheers,
Rick.
thanks Rick - yeah, sorry its so wordy. thanks for the link to the CN article, need to go through it in depth, but it looks quite similar to what I did

Quote:
Originally Posted by RobF View Post
And reading another post here by Peter Ward, and owning an 8300 chip camera myself, guess the other issue of oversaturation of bright stars on some fields if you go crazy with long NB sub times.....
Yes Rob, seems there is always a tradeoff between headroom and SNR and the systems I looked at all had similar headroom - the 3200 was best. Narrowband should help though - reduces the star brightness by maybe a factor of 30 or more. What sub lengths do you use for narrowband?

regards Ray
Reply With Quote
  #16  
Old 26-03-2013, 09:03 AM
SpaceNoob (Chris)
Atlas Observatory

SpaceNoob is offline
 
Join Date: May 2012
Location: Canberra
Posts: 268
Quote:
Originally Posted by Shiraz View Post
thanks for the link Chris. Do you use the calculator for narrowband and if so, what sub exposure lengths does it suggest?

Suggested exposure lengths have been 20.7 minutes for Ha 5nm with the FSQ at f/5. While this has been fine for most of my targets, there would likely be issues with brighter targets, given limited well depth. Calculators can be used as a suggested guide for sub length; while respecting the rate of saturation for brighter objects. Like everything in this hobby, its about finding a balance.... only to find that there is some other variable you need to account for. You're dealing with different issues for narrowband.
Reply With Quote
  #17  
Old 27-03-2013, 07:35 PM
naskies's Avatar
naskies (Dave)
Registered User

naskies is offline
 
Join Date: Jul 2011
Location: Brisbane
Posts: 1,865
Very interesting, thanks Ray. As an extension of this, how about predicting the overall integration time needed (with optimal exposure duration) to reach some arbitrary SnR for an object of a given magnitude, under a specific skyglow magnitude, and with a certain sensor/f-ratio scope?
Reply With Quote
  #18  
Old 27-03-2013, 11:52 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by SpaceNoob View Post
Suggested exposure lengths have been 20.7 minutes for Ha 5nm with the FSQ at f/5. While this has been fine for most of my targets, there would likely be issues with brighter targets, given limited well depth. Calculators can be used as a suggested guide for sub length; while respecting the rate of saturation for brighter objects. Like everything in this hobby, its about finding a balance.... only to find that there is some other variable you need to account for. You're dealing with different issues for narrowband.
thanks for the info Chris. yes, it always seems that there is another tradeoff... In fact, the whole reason for doing the modelling was to minimise the cost - I wanted to try many different tradeoffs before I decided what to buy.

Quote:
Originally Posted by naskies View Post
Very interesting, thanks Ray. As an extension of this, how about predicting the overall integration time needed (with optimal exposure duration) to reach some arbitrary SnR for an object of a given magnitude, under a specific skyglow magnitude, and with a certain sensor/f-ratio scope?
model already does that Dave - bit awkward the way it is structured and needs a rewrite though. Results are attached - with bright sky, they largely reflect the differing Ha QEs of the various chips, with the 11002 having roughly half the QE of the others and requiring a bit more than twice the imaging time. Dark current also strongly influenced the 11002 and 3200 results, particularly at low sky brightness. The systems all have ~the same aperture and pixel scale and "5%noise" optimum subs were used regardless of how practical they were - systems were:
KAI11002M 254mm f8
KAF8300 254mm f4.75
KAF3200ME 254mm f6
icx694 254mm f4.

I could not find data on the Ha sky brightness in polluted skies and have used a dark sky measure (10nm BW) of 5 p/s/m2/arcsec2, with guesses as to how much extra in-band flux there might be in polluted sky. Also included 2.5p/s/m2/arcsec2 flux to represent a 5nm filter/dark sky. Looks like it will pay to get a 5nm Ha filter for my new 694 - the low dark current of this CCD (and to a lesser extent the 8300), gives a distinct advantage with very narrow filters.

also, does anyone have any feel for how much SNR is "enough" - it would be nice to be able to say what image quality SNR=30 represents

regards Ray
Attached Thumbnails
Click for full-size image (time.jpg)
41.3 KB15 views

Last edited by Shiraz; 30-03-2013 at 11:51 PM. Reason: update figure
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 09:28 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Testar
Advertisement
Bintel
Advertisement