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

Continue to oversample or go more optimal?- what would you do?

  • Please log in to reply
9 replies to this topic

#1 pyrasanth

pyrasanth

    Cosmos

  • *****
  • topic starter
  • Posts: 8,400
  • Joined: 08 Jan 2016

Posted 24 June 2025 - 08:50 AM

Consider my C14 at F7.7 using the ASI2600 MM PRO on a 1x1 bin. This has an effective resolution (probably never achievable on Earth) of 0.28 arc secs per pixel, so vastly over sampled.

 

I'm currently guiding at 0.88 arc secs per pixel through the ONAG using the ASI174 mini  which is the same focal length of the main optics on a 2x2 bin.

 

There is no doubt about the need to bin however would you be binning 2 or 3- 4 might be a bit limiting if a spate of superb seeing suddenly hit but even a 4 bin would only be putting the optics at a theoretical 1.12 arc secs per pixel which would need superb UK seeing conditions.

 

It is quite a hard choice to get things right- I'm tempted to think x3 with potential drizzling. This would put my guiding and main image scale around the same.

 

Let me know your thoughts & I can do some tests.

 

It is attractive  to bin, better guiding results and SNR along with smaller file sizes.

 

I wish you all good seeing at the theoretical limit of your optics (when you get to the moon).


Edited by pyrasanth, 24 June 2025 - 08:51 AM.


#2 Higgsfield

Higgsfield

    Viking 1

  • -----
  • Vendors
  • Posts: 792
  • Joined: 10 Sep 2020

Posted 24 June 2025 - 09:42 AM

Consider my C14 at F7.7 using the ASI2600 MM PRO on a 1x1 bin. This has an effective resolution (probably never achievable on Earth) of 0.28 arc secs per pixel, so vastly over sampled.

 

I'm currently guiding at 0.88 arc secs per pixel through the ONAG using the ASI174 mini  which is the same focal length of the main optics on a 2x2 bin.

 

There is no doubt about the need to bin however would you be binning 2 or 3- 4 might be a bit limiting if a spate of superb seeing suddenly hit but even a 4 bin would only be putting the optics at a theoretical 1.12 arc secs per pixel which would need superb UK seeing conditions.

 

It is quite a hard choice to get things right- I'm tempted to think x3 with potential drizzling. This would put my guiding and main image scale around the same.

 

Let me know your thoughts & I can do some tests.

 

It is attractive  to bin, better guiding results and SNR along with smaller file sizes.

 

I wish you all good seeing at the theoretical limit of your optics (when you get to the moon).

Oversampled, no way. Bin, why? You want a decent image do you not? Drizzling when you have the native resolution is not the best idea. Subframe to keep down the file size if need be. BXT works better when oversampled, I and many believe. Take some test shots, these old myths will be dispelled.

 

Your RMS error as I'm sure you know = sqrt(seeing²+ guiding_error² + stack_error²). While seeing is dominant, BlueXTerminator - which I assume you have - can deconvolve a fair amount of it. You likely can get down to 1.2 arc seconds per pixel or even better. The defraction limit of your scope is likely under a third of this value, so you want to use the full power of your optics right! I'm at around 1.02 arc/px after BXT. 0.88 arc seconds per pixel guiding error isn't great, but not unusual. Try going unguided and take short exposures like 5 to 15s for the brighter targets, there are a lot of them. Take 5s for stars, you generally do not need a longer exposure. On many of the brighter narrowband targets, 12 - 15s Ha and Oiii works really well.

 

Here are some links to example images taken on a 16" F8, at the same resolution as your setup, 0.238 arc/px. Binning would have lost noticeable resolution. A compromise is perhaps to bin 2 the colour info, and bin 1 the luminance. UK skies mean you got to get stuff in a hurry so LRGB is not a bad way to go. You have 4x the light gathering power in the L. The detail, as you know is also in the L.

 

https://www.astrobin.com/d1ci91/C/

 

https://www.astrobin.com/qpe1cz/

 

https://www.astrobin.com/b6o690/

 

The proof is in the pudding!


Edited by Higgsfield, 24 June 2025 - 10:00 AM.

  • smiller and licho52 like this

#3 pyrasanth

pyrasanth

    Cosmos

  • *****
  • topic starter
  • Posts: 8,400
  • Joined: 08 Jan 2016

Posted 24 June 2025 - 09:59 AM

Oversampled, no way. Bin, why? You want a decent image do you not? Drizzling when you have the native resolution is not the best idea. Subframe to keep down the file size if need be. BXT works better when oversampled, I and many believe. Take some test shots, these old myths will be dispelled.

 

Your RMS error as I'm sure you know = sqrt(seeing²+ guiding_error² + stack_error²). While seeing is dominant, BlueXTerminator - which I assume you have - can deconvolve a fair amount of it. You likely can get down to 1.2 arc seconds per pixel or even better. The defraction limit of your scope is likely under a third of this value, so you want to use the full power of your optics right! I'm at around 1.02 arc/px after BXT. 0.88 arc seconds per pixel guiding error isn't great, but not unusual. Try going unguided and take short exposures like 5 to 15s for the brighter targets, there are a lot of them. Take 5s for stars, you generally do not need a longer exposure. On many of the brighter narrowband targets, 12 - 15s Ha and Oiii works really well.

 

Here are some links to example images taken on a 16" F8, at the same resolution as your setup, 0.238 arc/px. Binning would have lost noticeable resolution. A compromise is perhaps to bin 2 the colour info, and bin 1 the luminance. UK skies mean you got to get stuff in a hurry so LRGB is not a bad way to go. You have 4x the light gathering power in the L. The detail, as you know is also in the L.

 

https://www.astrobin.com/d1ci91/C/

 

https://www.astrobin.com/qpe1cz/

 

https://www.astrobin.com/b6o690/

Thank you- really useful information. I think I might have confused you with guiding setup vs guiding results. The guiding capabilities are 0.88 arc secs per pixel & I mostly get guiding RMS of around 0.6-0.7.

 

I like the images you have posted especially as I've imaged 2 of these targets myself on a 1x1 bin. I believe from your findings I should continue with the 1x1 bin and use BXT to try & refine the data. It has always produced good results & it may very well be because the data presented to it is oversampled.

 

Here are a few examples of my current efforts

 

https://www.astrobin...lwc/D/?nc=&nce=

 

https://www.astrobin...51wxz/?nc=&nce=

 

https://www.astrobin...opgor/?nc=&nce=



#4 Higgsfield

Higgsfield

    Viking 1

  • -----
  • Vendors
  • Posts: 792
  • Joined: 10 Sep 2020

Posted 24 June 2025 - 10:04 AM

Not too shabby. Some of the difference comes down to processing or perhaps not having quite enough data for a higher SNR. But the details seems to be in the data. Would have to see the raw data to provide more of an assessment.  Nice!


Edited by Higgsfield, 24 June 2025 - 10:10 AM.


#5 SilverLitz

SilverLitz

    Soyuz

  • *****
  • Posts: 3,508
  • Joined: 17 Feb 2018
  • Loc: Louisville, KY

Posted 24 June 2025 - 11:38 AM

Seeing at my home is typically 1.5" per Meteoblue, and I see noticeably more detail down to ~0.3"/px, with no benefit for any finer.  This is getting there either natively or after 2x drizzle.  I also use BXT.  



#6 CygnusBob

CygnusBob

    Messenger

  • -----
  • Posts: 450
  • Joined: 30 Jun 2018
  • Loc: Las Vegas, NV

Posted 24 June 2025 - 11:42 AM

pyrasanth

 

For a 14 inch telescope at 550 nm wavelength, critical sampling for the optics alone would be 0.159 arc-seconds, "1/2 lambda/D".  However bad seeing would probably mean using a plate scale this fine would be a waste of time.  Actually the plate scale is set by the size of the pixels and the effective focal length of the OTA so you could, using a Barlow lens have a plate scale even smaller.  A plate scale of 0.28 is achievable.  No problem! However that does not mean that the image will be high resolution.  Guiding errors of 0.88 arc-seconds RMS will certainly limit the final "resolution".  

 

For guiding, using a plate scale at critical sampling is most likely a bad idea.  First, in addition to fine sampling we need a high SNR.  Short exposures and a fine plate scale means lower SNRs.  The other thing is that poor seeing will dominate the guiding signals so that good position error measurements will be swamped out by atmospheric seeing induced position errors.

 

However for stacking the final imagery a fine plate scale does matter.

 

Bob


Edited by CygnusBob, 24 June 2025 - 11:43 AM.


#7 Higgsfield

Higgsfield

    Viking 1

  • -----
  • Vendors
  • Posts: 792
  • Joined: 10 Sep 2020

Posted 24 June 2025 - 02:12 PM

Oversampled, no way. Bin, why? You want a decent image do you not? Drizzling when you have the native resolution is not the best idea. Subframe to keep down the file size if need be. BXT works better when oversampled, I and many believe. Take some test shots, these old myths will be dispelled.

 

Your RMS error as I'm sure you know = sqrt(seeing²+ guiding_error² + stack_error²). While seeing is dominant, BlueXTerminator - which I assume you have - can deconvolve a fair amount of it. You likely can get down to 1.2 arc seconds per pixel or even better. The defraction limit of your scope is likely under a third of this value, so you want to use the full power of your optics right! I'm at around 1.02 arc/px after BXT. 0.88 arc seconds per pixel guiding error isn't great, but not unusual. Try going unguided and take short exposures like 5 to 15s for the brighter targets, there are a lot of them. Take 5s for stars, you generally do not need a longer exposure. On many of the brighter narrowband targets, 12 - 15s Ha and Oiii works really well.

 

Here are some links to example images taken on a 16" F8, at the same resolution as your setup, 0.238 arc/px. Binning would have lost noticeable resolution. A compromise is perhaps to bin 2 the colour info, and bin 1 the luminance. UK skies mean you got to get stuff in a hurry so LRGB is not a bad way to go. You have 4x the light gathering power in the L. The detail, as you know is also in the L.

 

https://www.astrobin.com/d1ci91/C/

 

https://www.astrobin.com/qpe1cz/

 

https://www.astrobin.com/b6o690/

 

The proof is in the pudding!

I did not at pickup that you are using an ONAG. I also use an ONAG. At this focal length, PHD2 may be doing more harm than good. 0.5 to 0.6 arcsec rms error is very typical and maybe putting a floor under your system. Not to say you will not on occasion do much better like 0.35 - 0.4", but those are likely to be rare events given your typical seeing. This problem is with how PHD works, and how the control algorithm interacts with the dynamics (moment of inertia, etc) of your mount given the payload you have - settling time for any corrective pulse. My suggestion is to see just how long an exposure you can take before you get obvious non round stars - 0.6 eccentricity is OK! You may be surprised how long you can go if properly polar aligned. May also vary with sky position.  On my setup I can go 60s easy and upwards of 400s unguided. This reduces your rms error = sqrt(seeing² + stack_error²). Bear in mind that the stacking image registration error can be significant at your focal length, hence why I added it to the equation. Note when you use BXT apply first in "Correct Only" mode, this fixes any eccentricity in your stars, then apply BXT proper. Because of your plate scale you may be able to push the NonStellar Correction beyond 0.5. As for the guide camera we Bin 4. Here we are looking to maximize the SNR especially since the image is in the IR.



#8 pyrasanth

pyrasanth

    Cosmos

  • *****
  • topic starter
  • Posts: 8,400
  • Joined: 08 Jan 2016

Posted 24 June 2025 - 03:01 PM

I did not at pickup that you are using an ONAG. I also use an ONAG. At this focal length, PHD2 may be doing more harm than good. 0.5 to 0.6 arcsec rms error is very typical and maybe putting a floor under your system. Not to say you will not on occasion do much better like 0.35 - 0.4", but those are likely to be rare events given your typical seeing. This problem is with how PHD works, and how the control algorithm interacts with the dynamics (moment of inertia, etc) of your mount given the payload you have - settling time for any corrective pulse. My suggestion is to see just how long an exposure you can take before you get obvious non round stars - 0.6 eccentricity is OK! You may be surprised how long you can go if properly polar aligned. May also vary with sky position.  On my setup I can go 60s easy and upwards of 400s unguided. This reduces your rms error = sqrt(seeing² + stack_error²). Bear in mind that the stacking image registration error can be significant at your focal length, hence why I added it to the equation. Note when you use BXT apply first in "Correct Only" mode, this fixes any eccentricity in your stars, then apply BXT proper. Because of your plate scale you may be able to push the NonStellar Correction beyond 0.5. As for the guide camera we Bin 4. Here we are looking to maximize the SNR especially since the image is in the IR.

Thank you- good tips. My camera currently can only bin 2 max ( ASI 174 mini) I will need to change the camera to get a more substantial binning level.

 

I will see how far I can get with unguided imaging.



#9 Higgsfield

Higgsfield

    Viking 1

  • -----
  • Vendors
  • Posts: 792
  • Joined: 10 Sep 2020

Posted 24 June 2025 - 03:45 PM

Thank you- good tips. My camera currently can only bin 2 max ( ASI 174 mini) I will need to change the camera to get a more substantial binning level.

 

I will see how far I can get with unguided imaging.

Bin4 is a selection option in pdh2. Cameras are generally Bin 2 in the hardware. 



#10 WadeH237

WadeH237

    Voyager 1

  • *****
  • Moderators
  • Posts: 12,109
  • Joined: 24 Feb 2007
  • Loc: Ellensburg, WA

Posted 24 June 2025 - 05:37 PM

If it were me, I would always shoot at the camera's native resolution and actually process it that way.  Then, near the very end of processing, I would examine some stars to see if there was "empty resolution".  At that point, I would resample the image to achieve whatever pixel scale you find appropriate.

 

In terms of quality of data, there is nothing to be gained by binning with CMOS cameras.  CMOS cameras are always native at 1x1.  The down sampling actually happens in software (the driver).  Pushing it to later in processing allows some of the processing steps to work better.

 

The only reason to ever bin a CMOS camera is to reduce the size of the raw files.  For example, if your scope is hosted at a remote site and you need to download the data over a slow link, it's better to have smaller files.  But even then, being over sampled is usually desirable.

 

The guide camera is a different story.  In that case, the guide software doesn't get the benefit of post processing.  So binning the guide camera may help you to use dimmer guide stars.


  • sc02492 and steveincolo like this


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






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics