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

#1 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 11 August 2020 - 07:12 AM

Hello,

 

I am currently imaging with a RC 8“ and ASI 1600MM from a light polluted backyard (Bortle 5/6). For narrowband, total integration times to get acceptable results is great (1 object per night, 2 nights in summer). I would like to keep the image scale below about 0.8 arces/pixel to take advantage of the rather good seeing I often observe. At native image scale of 0,6“, I need long total integration times for broadband to ger decent SNR (20+ to 30+  hours, depending on object). I tried the CCDT reducer but it causes image tilt and/or reflection and is a hassle to set up properly in my setup. In-driver binning does not work well with the ASI1600, and I wasn‘t able to reproduce the theoretical SNR-increse by binning in software (workflow: Mure Denoise, Resample, Stretch). I have used an etendue calculator to show which scope/camera combination might improve on the imaging times. As a mount I have a Mesu 200 Mk2 coming in, so weight should not be much of an issue. I am aware traveling to a dark site would be better, but that is not an option (I am very frequently on-call as a physician). The following should, I think, provide substantial speed improvements:

 

CDK 12.5/Kaf-16308:

+ Proven performers

- Price

 

CDK 12.5/Qhy600M or Asi6200MM

+ (Quite) a bit cheaper, but still pricey with new filters and FW

+ High QE

- Needs CMOS binning

 

ONTC Newton 10“ f/4/Qhy 690

+ Lowest price (can keep using current filters)

- Frequent difficult (?) collimation

- Rather small fov (but not by much)
 

ONTC Newton 12“ f/5/Qhy16200A

+ Largest Fov

+ integrated camera/fw

- Frequent collimation, but easier

- Bulk

 

EdgeHD 11/Qhy600M or ASI6200MM

+ Still in budget

+ Large user base

+ Highest etendue

+ Lightest scope

+ Also good for visual/planetary

- Needs CMOS binning

- Scope needs tweaking for auto focus (probably secondary mirror focusing)

 

Although the CDK looks very appealing, at the prices in Europe, money does play a role. 

 

My questions would be the following:

- Am I right in assuming I could decrease total integration time per target by a factor of about 5-6 with the above combinations (or use the time to further increase SNR)?

- Have I overlooked any issues with the above combinations?

- Any bad/good experiences with any of the above with imaging smaller DSOs?

 



#2 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 13,677
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 11 August 2020 - 09:08 AM

Hi Rasfahan:

 

I'm REALLY glad to see this question because I'm traveling the same path right now.  I have an EdgeHD 8 with an ASI071 camera.  The 0.7x focal reducer gives me reasonable exposure times, but introduces unacceptable aberrations.  And I'd love to reduce my exposure time even more.  I too have tried binning and have not produced the theoretical improvements in SNR.

 

I don't have a definitive answer for those scopes without doing a lot of research on the various cameras' pixels sizes and scope focal lengths.  But some things to consider.

 

First, you mention your binning workflow includes "Resample".  To do a true binning, you need to use the IntegerResample process in PI.  Resample applies a different algorithm.  I would be interested to learn if this makes a difference for you.  Again, I have not seen the expected SNR improvements even using IntegerResample.

 

Second, when calculating etendue, you need to consider whether you're interested in overall etendue, or etendue per pixel (or super-pixel if binning).  Some things affect one, others affect both.  For example, increasing your sensor size will increase your overall etendue because more photons will hit a larger sensor than a small one.  But it will not change your exposure time or your SNR because it does not affect your etendue per pixel.  On the other hand, reducing your f/ratio will increase both your overall etendue and your etendue per pixel, so it will reduce your exposure time and increase your SNR.  

 

In my opinion, etendue per pixel is the more important metric.  Anything that affects it should affect your exposure time and SNR.

 

With my EdgeHD 8, I have been considering dropping the focal reducer, moving to an ASI6200, and then binning 2x2.  That would provide me the same FOV as my current ASI071 with the FR.  And if I bin 2x2, it would give me an effective pixel size of about 7.5u, with a pixel scale of around 0.7".  In theory, binning to that size should give me 2 stops improvement in SNR.  But before spending all that money on the camera, I decided to test binning on my 071.  So far, it seems like I'm seeing less than one stop improvement.

 

I'm also considering the TS Optics ONTC 10" f/4.  I'm not familiar with most of the rest of the scopes, but I don't believe the EdgeHD 11 with the ASI6200 will give you any improvement in exposure time.  The long focal length, high f/ratio, and small pixels of the 6200 will not create much etendue per pixel.  If you bin, it should improve things, but again I've not seen that in practice.

 

-Dan

 

-Dan


Edited by Midnight Dan, 11 August 2020 - 09:12 AM.

  • rockstarbill likes this

#3 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 11 August 2020 - 09:47 AM

Thanks for your thougts, Dan. The scope/camera combinations were actually selected to give a comparable image scale (sometimes with 2x or 3x binning, as for the ASI6200, where noted) of about 0.6-0.8 arcsec/pixel with a good fov for the objects I plam to image.

I will certainly try your tip about IntegerResample, I was not aware of that distinction, that is valuable to know.

 

I was referring to pixel etendue, because only that should affect total integration time, if I understood the concept correctly. In the spreadsheet I used, the above combinations give a per pixel etendue between 200-350, whereas my current combination scores a meager 37 (it‘s actually the sheet from fellow CNer roofkid, taking into account aperture, f/ratio, obstruction, qe and pixel size).

 

- Torben


Edited by Rasfahan, 11 August 2020 - 09:47 AM.


#4 sn2006gy

sn2006gy

    Viking 1

  • ****-
  • Posts: 909
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 11 August 2020 - 09:59 AM

It's been my experience that moving to faster scopes in light polluted areas doesn't necessarily translate to reduced integration time. It would translate to shorter exposures and more exposures for nearly identical integration time with less noise and more fidelity for the given skies. I've come to realize shooting a single target all night or over multiple nights work best so chasing the speed demon many not translate to much but more images per integration - which is great. (but longer integrations). 

 

I believe if you chase the multiple-target a night because your scope is technically twice as fast as it used to be (on paper), you would be spending a LOT of money only to have similar quality as you have now.

 

I'd also add, that moving to a 6200, you would be trading scope time for processing time. "Integration time" on your PC will be exponentially longer.

 

I "only" have an APS-C and processing a single night of 70 300 second subs takes 50-100gb of storage and a bit of time to integrate on my 16 thread Ryzen with 32gb of ram on an NVMe.


  • Craig H, ezwheels and MontyC like this

#5 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 11 August 2020 - 11:12 AM

Thanks, Byron,

 

I actually have started to target fainter fuzzies, which I quite enjoy. I do not mind overcoming the obstacles to get high resolution images. Considering what time I need for objects with higher surface brightness, I estimate at least 6-7 nights per object with my current setup (my last image of M64 took 12h integration time and I would judge it as mediocre in SNR, needing at least 24h). This makes every image a project spanning months, and, depending on weather, years. My thinking was that better per pixel etendue might decrease this to maybe 2 nights. I have no interest in finishing more than one target per night.
 

An ASI6200 would need to be binned 3x to fit with the above scopes, resulting in 7 MP images, effectively reducing my processing time considerably (longer possible exposures and less MP compared to unbinned ASI1600 which is grossly oversampling even the RC8). It is, of course, unfit for the RC8 because of its image circle.

 

- Torben


  • psandelle and sn2006gy like this

#6 rgsalinger

rgsalinger

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 7,122
  • Joined: 19 Feb 2007
  • Loc: Carlsbad Ca

Posted 11 August 2020 - 11:42 AM

I switched to a QHY600 about 6 months ago.  I've been able to reduce my imaging time to achieve a 20 plus SNR by about half when compared to my previous QHY16200. I usually bin the camera 2x2 which gives me a .6 as/pixel image scale. That's almost perfect sampling at the dark sky site where seeing seems to be just under 2 arc seconds most good nights. 

 

The scope that I use is a PlaneWave CDK12.5. It is by far the best purchase that I have ever made in this hobby. The PW scopes are much more than just an OTA. My scope has computer controlled heaters, fans and focuser along with the best focusing software that I've ever used. The imaging circle is full frame. 

 

Lower_Right.JPG

 

is a zoomed in shot of the extreme lower right of a 5 minute exposure with only a screen stretch so that you can see the stars. I can't imagine getting better data than this. 

 

Rgrds-Ross

 

 

 

 

 

 

 

 

 

 

 

Attached Thumbnails

  • Lower_Right.JPG

  • Alterf, psandelle, Craig H and 3 others like this

#7 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 11 August 2020 - 04:29 PM

Thanks, Ross, that is an amazing star quality in the corner and very little noise. I am more and more coming to the conclusion that the Planewave/Qhy600 combo is the most hassle free and effective path to go. The cheaper alternative might be the imaging newton, but I have no experience with newtons, and fear to be standing outside in freezing cold every winter session collimating the scope. The RC8 takes me 30min to collimate, but I only had to do it twice so far, once after a 300mi road trip.


Edited by Rasfahan, 11 August 2020 - 04:30 PM.

  • psandelle likes this

#8 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 13,677
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 11 August 2020 - 05:35 PM

Hi Rafahan:

 

Just an update on the binning vs. SNR issue.  (I told you I was on the same path :-). Last night I did a test with my Edge 8 see if I could quantify the effect of binning with my ASI071 camera.  I took about 15 subs of a star field at 15s, 30s, 60s, and 120s.   I also went to 240s, but those got clouded out so I had to toss them

 

I tried a number of different workflows using DSS and PI and got varying results.  It was difficult to determine what was really happening in each workflow so I decided to just use PI and keep it simple and straightforward.  Here's what I did with each set up subs:

 

1) Use the Debayer process on the subs, twice. Once was using the VNG method which does not bin.  The second time, I used the SuperPixel method which creates one large pixel from the 4 bayer matrix pixels.  So I had 8 groups of subs, 4 exposure times of binned and 4 of non-binned.

 

2) For each group, I used the StarAlign process to register the subs.

 

3) For each group, I used the ImageIntegration process to stack the aligned subs.

 

4) I loaded each of the 8 resulting images into PI's SubframeSelector process to measure the SNR Weight.

 

Here's the results.  Note that the 120 second images should actually be a little higher in SNR Weight than what is shown.  I lost several subs at the end of the 120 session to clouds, so those stacked images do not have double the integration time of the 60 second images.

FInal bar chart.jpg

 

I'm not exactly sure how the vertical scale for SNR Weight works, but you can see the jump from one blue bar to the next and use that as a reference for one stop of improvement in SNR.  You can see that the difference between a red and blue bar is larger than one stop, but not 2 stops.  Seems like roughly 1.5 stop to me.  

 

In theory, a 2x bin should get you 2 stops of improvement (double the SNR).  While I'm not quite seeing that here, it certainly seems like enough of an improvement to make binning a viable option.

 

-Dan


  • psandelle, Peregrinatum and Rasfahan like this

#9 rgsalinger

rgsalinger

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 7,122
  • Joined: 19 Feb 2007
  • Loc: Carlsbad Ca

Posted 11 August 2020 - 06:49 PM

Two stops is 4x improvement.

 

You are getting two stops of signal but you also double the noise when you bin a cmos camera. Those numbers are exactly what you are seeing in the longer exposures --- a 2x improvement in SNR for the same time.  I did the same (not so nice but essentially the same) experiment using a set of 180 second subs binned and un-binned. Signal went up (as expected) by 4x and noise went up by 2X. So, I've been shooting binned for quite a while now. This works out really well as the subs are just 30mB and not 120mB when they come off the camera.

 

The next frontier in CMOS has to be full frame with something like 6 micron pixels for those brave souls using a really big telescope. With an Edge 8, and a camera like the QHY600, you effectively have a "free" binning option unless you put the telescope into outer space. 

 

Rgrds-Ross


  • psandelle, rockstarbill and Gene3 like this

#10 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 13,677
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 11 August 2020 - 07:04 PM

Two stops is 4x improvement.

 

You are getting two stops of signal but you also double the noise when you bin a cmos camera. Those numbers are exactly what you are seeing in the longer exposures --- a 2x improvement in SNR for the same time. 

 

Rgrds-Ross

Hi Ross:

 

Two stops is 4x improvement in signal.  But as you point out, noise goes up by 2x as well.  So the SNR improvement for 2 stops is 2x as I stated.

 

-Dan



#11 rgsalinger

rgsalinger

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 7,122
  • Joined: 19 Feb 2007
  • Loc: Carlsbad Ca

Posted 11 August 2020 - 08:07 PM

Well, the signal improvement is 2 stops - 4x more light (photons) but the noise goes up by 2X. So if I take signal/noise it's a 2x improvement. We seem to agree on that. I would call this a one stop improvement in signal to noise ratio.

Rgrds-Ross



#12 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 13,677
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 11 August 2020 - 10:22 PM

Well, the signal improvement is 2 stops - 4x more light (photons) but the noise goes up by 2X. So if I take signal/noise it's a 2x improvement. We seem to agree on that. I would call this a one stop improvement in signal to noise ratio.

Rgrds-Ross

Ah I see what you're saying. 

 

The way I look at it, if you increase your exposure by one stop, you get about a 42% increase in SNR.  If you increase your exposure by two stops you get a 100% increase in SNR (or 2x).  So to me, 2x improvement in SNR is equivalent to 2 stops of exposure.  

 

-Dan



#13 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 12 August 2020 - 06:17 AM

Thanks, Dan, that is good information. I m currently traveling and will try the IntegerResample process when I am back next week. I will report the result back here.

#14 lambermo

lambermo

    Viking 1

  • -----
  • Posts: 999
  • Joined: 16 Jul 2007
  • Loc: .nl

Posted 12 August 2020 - 12:23 PM

Here's some performance indicators to compare your current setup (RC 8“ and ASI 1600MM) with the first in your list (CDK 12.5/Kaf-16308) :

 

 

# compare-telescopes.py --s1 TSRC8 --c1 asi1600 --s2 CDK12.5 --c2 KAF16803
# Telescope 1 f/8.00  l=1624mm D=203mm O=42% res=0.48"/p FOV=37'x28'= 0.43x eoi= 1.00x poi= 0.39x e= 0.17x pe= 0.17x ps= 0.17x os= 0.39x
# Telescope 2 f/7.99  l=2541mm D=318mm O=37% res=0.73"/p FOV=50'x50'= 2.35x eoi= 1.00x poi= 2.58x e= 6.03x pe= 5.89x ps= 5.89x os= 2.57x

The eoi,poi,e,pe,ps,os are explained here https://github.com/d...ance-indicators

You can make your own comparisons either with the Python version or the online web version at https://lambermont.d...telescopes.html

(I'm using a not yet released python version with a gear database which eases input, but the results are the same)

 

Having said all that I'm in the same boat and have not decided on new gear yet ;-)

-- Hans



#15 Peregrinatum

Peregrinatum

    Apollo

  • *****
  • Posts: 1,496
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 12 August 2020 - 05:00 PM

Hi Rafahan:

 

Just an update on the binning vs. SNR issue.  (I told you I was on the same path :-). Last night I did a test with my Edge 8 see if I could quantify the effect of binning with my ASI071 camera.  I took about 15 subs of a star field at 15s, 30s, 60s, and 120s.   I also went to 240s, but those got clouded out so I had to toss them

 

I tried a number of different workflows using DSS and PI and got varying results.  It was difficult to determine what was really happening in each workflow so I decided to just use PI and keep it simple and straightforward.  Here's what I did with each set up subs:

 

1) Use the Debayer process on the subs, twice. Once was using the VNG method which does not bin.  The second time, I used the SuperPixel method which creates one large pixel from the 4 bayer matrix pixels.  So I had 8 groups of subs, 4 exposure times of binned and 4 of non-binned.

 

2) For each group, I used the StarAlign process to register the subs.

 

3) For each group, I used the ImageIntegration process to stack the aligned subs.

 

4) I loaded each of the 8 resulting images into PI's SubframeSelector process to measure the SNR Weight.

 

Here's the results.  Note that the 120 second images should actually be a little higher in SNR Weight than what is shown.  I lost several subs at the end of the 120 session to clouds, so those stacked images do not have double the integration time of the 60 second images.

attachicon.gifFInal bar chart.jpg

 

I'm not exactly sure how the vertical scale for SNR Weight works, but you can see the jump from one blue bar to the next and use that as a reference for one stop of improvement in SNR.  You can see that the difference between a red and blue bar is larger than one stop, but not 2 stops.  Seems like roughly 1.5 stop to me.  

 

In theory, a 2x bin should get you 2 stops of improvement (double the SNR).  While I'm not quite seeing that here, it certainly seems like enough of an improvement to make binning a viable option.

 

-Dan

Great work Dan, I love it when people on here do experiments like this...

 

For the 2x2 binning "red bars" did you use the IntegerResample process?  and... not sure what you did with he Debayer... you did it twice for each set of data?


Edited by Peregrinatum, 12 August 2020 - 05:01 PM.


#16 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 13,677
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 12 August 2020 - 05:36 PM

Great work Dan, I love it when people on here do experiments like this...

 

For the 2x2 binning "red bars" did you use the IntegerResample process?  and... not sure what you did with he Debayer... you did it twice for each set of data?

For the red binned bars, I didn't use IntegerResample.  Instead I selected the SuperPixel method while using PI's Debayer process.  So yes, I debayered each set of data twice - once using SuperPixel, which debayers and bins at the same time, and once using the VNG method, which does not bin.

 

-Dan


  • Peregrinatum likes this

#17 Peregrinatum

Peregrinatum

    Apollo

  • *****
  • Posts: 1,496
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 13 August 2020 - 11:48 AM

OK I see... when I tried to bin previously I would do it near the end of processing with IntegerResample, but I didn't realize it adds noise... so it makes sense to do the resample with the Debayer process using the Superpixel option and then denoise in the linear state...



#18 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 13,677
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 14 August 2020 - 11:10 AM

OK I see... when I tried to bin previously I would do it near the end of processing with IntegerResample, but I didn't realize it adds noise... so it makes sense to do the resample with the Debayer process using the Superpixel option and then denoise in the linear state...

The bit about "adding noise" can be a little misleading.

 

With a mono camera, binning can be done in-camera which, for a 2x2 bin would simply add the 4 pixels together.  This doubles the noise, but quadruples the signal, so the SNR goes up by 2.  One issue with doing this is that clipping can occur if the pixels add up to more than the dynamic range of the camera allows.  

 

With software binning, it's an average rather than just an addition.  The 4 pixels are added, then divided by 4.  That eliminates the possibility of clipping.  So in this case, the end result is NOT an increase in noise, but a 2x reduction, while signal stays the same.

 

-Dan



#19 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 14 August 2020 - 01:53 PM

I just returned from home and found that using IntegerResample with a downscaling factor of 2 does, indeed, increase SNRWeight by a factor of exactly 2 - in raw frames as well as processed frames for the ASI1600MM Pro (as according to theory it should, of course). Also, it allows me to stretch harder, so replacing the in-driver binning without the disadvantages and allowing me to sample at a much better 0.96 arcsec/px instead of 0.48. This has allowed me to capture a very nice (as I think) HaO3RGB Crescent Nebula in only 5 hours at f/8, I could see no loss of detail due to the downsampling (seeing was mediocre). I will go for a small galaxy on the next clear night and reevaluate, a factor of 2 for free is a good start.

 

Edit: down- for upsampling above.

Attached Thumbnails

  • 602E268B-55BE-4630-A953-BFBA221229E6.jpeg

Edited by Rasfahan, 14 August 2020 - 02:57 PM.

  • Peregrinatum likes this

#20 Peregrinatum

Peregrinatum

    Apollo

  • *****
  • Posts: 1,496
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 14 August 2020 - 02:21 PM

I just returned from home and found that using IntegerResample with a downscaling factor of 2 does, indeed, increase SNRWeight by a factor of exactly 2 - in raw frames as well as processed frames for the ASI1600MM Pro (as according to theory it should, of course). Also, it allows me to stretch harder, so replacing the in-driver binning without the disadvantages and allowing me to sample at a much better 0.96 arcsec/px instead of 0.48. This has allowed me to capture a very nice (as I think) HaO3RGB Crescent Nebula in only 5 hours at f/8, I could see no loss of detail due to the upsampling (seeing was mediocre). I will go for a small galaxy on the next clear night and reevaluate, a factor of 2 for free is a good start.

Looks great!  at what point in the processing did you apply the IntegerResample for the image above?



#21 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 14 August 2020 - 02:55 PM

Looks great!  at what point in the processing did you apply the IntegerResample for the image above?

The full process was preprocessing, stacking and narrowband-RGB combination with Astro Pixel Processor. Then in Pixinsight DBE, TGV Denoise (w/ custom lightness mask as local support), IntegerResample, MaskedStretch, ACDNR (w/ strong lightness mask, lightness only), ArcsinH-Stretch, Multiscale HDR Transform (Scale 6), very slight sharpening with MMT, Curves Transformation. I am very satisfied with the faint details, not quite so with the stars, which I might process again seperately.



#22 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 13,677
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 14 August 2020 - 03:02 PM

Excellent, Rasfahan!  Glad to hear it's working for you.  How did you measure SNR to determine the 2x improvement?

 

-Dan



#23 Rasfahan

Rasfahan

    Sputnik

  • -----
  • topic starter
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 14 August 2020 - 06:43 PM

Excellent, Rasfahan!  Glad to hear it's working for you.  How did you measure SNR to determine the 2x improvement?

 

-Dan

I used the subframe selector script on frame before and after IntegerResample, also compared the unscaled frames with a frame at the same scale after IntegerResample(downscale) -> Resample (upscale). And I reprocessed the above image that showed lots of residual noise in the O3-filaments before.

 

- Torben



#24 rgsalinger

rgsalinger

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 7,122
  • Joined: 19 Feb 2007
  • Loc: Carlsbad Ca

Posted 14 August 2020 - 08:41 PM

The documentation for Integer Resample says pretty clearly that you will get 1/2 of the noise if you use it as well as 4x the signal. That's because it treats each of the four pixels that will make up the down sampled image as if was part of a separate exposure. So,  it's like have taken 4 images. It's pretty cool. However, bear in mind that if you aren't oversampled enough you will lose details. That's also in the documentation. Since I can image at .3 arc seconds, I can't wait to try this out. (I'm going to hate those 120 mbyte images though.)

 

Rgrds-Ross



#25 hobbyknipser

hobbyknipser

    Mariner 2

  • -----
  • Posts: 239
  • Joined: 23 Jul 2019
  • Loc: Germany Orange Zone

Posted 15 August 2020 - 10:17 AM

Hi, Torben,

 

if you want to catch small weak objects and reach a very good depth, mostly in only 2h... what about the parameters of the collague Rick_J? It's only an idea! So please don't be angry! blush.gif

He had had darker skies (and you brighter skies) , but at f/10 it seems not to be sooo important, or am I wrong?

You could use a camera with smaller pixels (but with less light) and so on...but a big aperture (14") is relevant for the depth in little time.

Here a link: #2 !

 

https://www.cloudyni...t-does-it-take/

 

cheers

Andreas


Edited by hobbyknipser, 15 August 2020 - 10:31 AM.



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