What is best exposure time - IC 342 test

[menardre]

There have been several posts regarding what is the optimum exposure times for the newer CMOS sensors. There are a ton of forum posts and on-line articles regarding exposure times. Some are very complicated with many equations. I decided to perform my own test.
So a few nights ago I took 1 hours worth of images at various exposures. So 120 exposures of 30 seconds, 60 exposures of 1 minute, 30 exposures of 2 minutes and 15 exposures of 4 minutes (I actually took about 10% more exposures of each to account for having a few bad shots). These were all done concurrently using the same equipment (SVX130 refractor and ZWO ASI 071 camera) so the results should speak for themselves. Meridian flip occurred during the 4 minute exposures. I picked IC 342 because it is fairly large and fairly faint. Each set of images was then processed using Images Plus. The first set of images below are simply the identical stacked and ArcSinh stretched result. The second set utilize the same identical image processing tools to the stretched images ... color balance, background flatten, etc. I treated each set of images identically. I used Photoshop only to add the text and compressed the JPG to make the file sizes smaller.

The first set of images were ArcSinh stretched

120 exposures - 30 second subs stretched.

IC 342 30S Subs stretched compressed.jpg (952.53 KiB) Viewed 2658 times

60 exposures 1 minute subs stretched.

IC 342 1M Subs stretched compressed.jpg (980.04 KiB) Viewed 2658 times

30 exposures 2 minute subs stretched

IC 342 2M Subs stretched compressed.jpg (1.37 MiB) Viewed 2658 times

15 exposures 4 minute subs stretched

IC 342 4M Subs stretched compressed.jpg (1.43 MiB) Viewed 2658 times

The second set use the same image processing tools such as color balance, background flatten, microcurves, etc

120 exposures 30 second subs processed

IC 342 30S Subs processed compressed.jpg (1018.11 KiB) Viewed 2658 times

60 exposures 1 minute subs processed

IC 342 1M Subs processed compressed.jpg (1.3 MiB) Viewed 2658 times

30 exposures 2 minute subs processed

IC 342 2M Subs processed compressed.jpg (1.19 MiB) Viewed 2658 times

15 exposures 4 minute subs processed

IC 342 4M Subs processed compressed.jpg (1.13 MiB) Viewed 2658 times

So, draw your own conclusions. Which exposure would you use???

Roger

[Bruce D]

I wonder what 5 x 12 minute exposures would capture. Is there a way to estimate the longest practical exposure before saturation takes over?

[bluemax]

Roger, thanks for a very good study on exposure time! I have always wanted to do just what you did.
I suppose this is pulling the pin on a hand grenade but, does your data mean that Robin's long dissertation on exposure time is incorrect?
On the surface it sure does to me. I'm no physics expert but it seems like your study is quite clear that longer is better. Maybe Robin means really a lot of short ones like 1000+ .

[Bruce D]

Frank i didn't have the nerve to take Robins on and pull that pin, I'm glad you did! Maybe Robins WAS figuring on thousands of short exposures, but given most of us have limited telescope time doesn't a discussion about best exposure time have to factor in total exposure? Rogers experiment clearly shows longer exposures capture far more data than an equal total exposure divided into much shorter subs, looking at the extremes because that often makes a point most obvious, I'd be very curious to know how many 30 second exposures it would take to gather the same amount of data Roger captured with his 25 four minute exposures. Also Roger, I'd like to know if you use a delay between each exposure to allow for downloading and saving the previous shot. Or maybe that's more a DSLR thing. If so, if one is taking a thousand exposures that delay will add up.

Look at me- trying to seem like I know what I'm talking about!
Awesome experiment Roger and i know a lot of time went into it, well done!

[Pete]

Wow! Lot of work and surprising outcome.

Theoretically the longer the stack exposure the more photons captured. And as many shorts are supposedly equivalent to fewer longs it shouldn't make any difference.

Robin's focus is on minimizing noise. Noise shows up when stretching dim areas of the image. SharpCap quantifies both shot noise and sky brightness in units of electrons/pixel/second. And comes up with a suggestion for gain and exposure that minimizes the combined noise. Surprisingly this formula advocates a large stack of relatively short exposures.

With CCDs the camera shot noise is on the order of 10e while it's more like 1e - 5e on CMOS. The shot noise is what we have traditionally overcome with long exposures on CCDs.

So, 10 second stacks vs equivalent 60 second total time?



Amateur Astrophotography Magazine has done what Roger did. And they found that using shorter exposure times will give you limited signal collection. In other words, the fainter parts of the object won’t have been registered onto the cameras chip. This is because the chip will not have had time to collect enough photons emitting from those parts of the object, hence building enough signal to be registered. So no matter how many images you then stack, zero multiplied by any number of subs is still zero data! See [url]https://www.amateurastrophotography.com ... is-better/[/url

So Amateur Astrophotography Magazine duplicates Roger's results. Wow!

I'm running SharpCap have attempted to use/apply Robin's exposure calculations. But the SharpCap align routine requires a minimum of 10 bright alignment stars in the image and that doesn't happen with his recommended very short exposures or recommended very low gain. The camera sensor analysis has determined that shot noise drops as gain increases and once past a gain level of 200 it's minimized. So I've standardized on the gain of 200 and have been using exposures of 15s, 30s, 60s, 90s, 2m,3m and 5m. In theory Robin's approach is optimal but in practice its pretty much unusable.

I'm not in the habit of contradicting PHds in their area of specialty, but both Amateur Astrophotography Magazine's and Rogers work suggests that exposures of 60 X 10 seconds does not equal 10 X 60 seconds. And that there's a problem processing low signal levels.

SharpCap's built in align & stack function is very precise. And my particular experience is that SoftCap's pre-aligned pre-stacked image works far better than stacking and aligning individual frames in ImagesPlus. Less work too.

Practice overrides theory.

[mark.m]

First, please take everything that I say with a grain of salt, because I'm no astrophotographer -- I'm a scientific imager, and the two are different. I'm not ready to discount "Robin's analysis", because it's quite consistent with my own analysis and it makes good, reliable predictions about the images that I get from my camera.

Several months ago, I built something of a monster spreadsheet to help me figure out what I needed for my next camera. And I captured all of the imaging math that Robin talks about: read noise, light pollution, dark current, saturation, gain, quantum efficiency, pixel scale, and so on. In addition to its original purpose (helping select a camera), I realized that this math could do an interesting job of replacing my crude thumbrules for choosing subexposure duration and number of subexposures to stack. I wrote some new software around that, and now have several thousand nighttime images that were obtained using those predictions. (The new software chooses subexposure quantity and duration based on predictions of SNR and onset of saturation.) And the predictions have been spot on.

There are several different reasons why Roger's results could differ from Robin's numbers:

• Stacking: There is more than one way to stack a set of images (mathematically). The different techniques can be simplified to the same underlying mathematical results (as far as signal-to-noise goes), but give different (looking) stacked sequences. (There's an ambiguous pixel scaling factor that the programmer must choose.)
• Representation: Stacking increases dynamic range, and this may exceed the original image's file encoding technique, in which case the stacking program will probably (hopefully) use a different pixel encoding technique. (When I stack my images, my stacker shifts from using 16-bit unsigned integers to using floating point numbers for the output file.)
• Calibration: Where in the stacking sequence bias and dark images are included has little effect on the math of the noise in the final image but has a big effect on the pixel ranges in the resulting file. When we custom process an image, we quickly correct for this. But when we apply the "same" scaling and post-processing steps, we may be unintentionally hurting the resulting image quality.

One of the differences between astrophotgraphers and scientific imagers is how we treat image noise: astrophotgraphers try to get rid of it and scientific imagers try to preserve it. All of "Robin's math" is about maximizing SNR, and sometimes it's hard to estimate that from a processed image that has intentionally tried to suppress noise.

And perhaps one of the take-aways from this is that the "best picture" is not necessarily the picture with the "best SNR." (By the way, there's a big difference between pixel-level SNR and star-level SNR -- I care about star-level SNR and most astrophotography seems to focus on pixel-level SNR.) At any rate, Roger, I think it'd be interesting for you to take each of the different results and process them differently, trying to make the best picture possible out of each. And see how those final products compare... (although that's probably hours of additional work).
- Mark M

[NGC7000]

@ Pete- Love the "Exposure" illustration. Brilliant!

Tom

[Bruce D]

Tom- you beat me to the punch! Pete you are a hoot!

I figured there would be some difference between exposure times, and I understand (more at an intuitive rather than physics level) about wanting to limit noise, I guess what set me back about Roger's experiment was the magnitude of the difference. Mark I wonder if the short exposures work better for you because you are usually looking at and measuring point source stars where low noise is most critical to your measurements, and aren't concerned with wispy details Pete points out may not register at all on short exposures? IDK, but I find the results of Roger's experiment quite dramatic

[menardre]

I am glad to hear everyone's response to my test.

I have read many articles related to how long you should expose objects. Based on that, my own tests, I agree with what Pete said in summary ...

when you have a dim object there just are not that many photon's. When you the include the quantum efficiency you end up with what Pete said ... multiply anything by zero is still zero. You have to expose long enough so that you capture a 'significant' number of photons in order for any stretching to work.

My rule of thumb is fairly simply .... I take a sample image of the object and I look at the result of a single (autostrectched) image in SGP. If I can see no trace of the object, then I increase the exposure time until I can at least see a trace of the object. That way I am sure that I am capturing sufficient number of photons. Of course I also look to make sure I am not saturating something else. This may not be the most scientific approach, but it allows me to visualize the outcome better.

Thanks for everyone's input ..... I will study everyone's comments in more detail since there are a lot of good thoughts.

Roger

[menardre]

Regarding any delay between exposures...

My ZWO ASI 071 takes about 1 sec (actually a little less) to download the image and start the next.... so not much of a delay.

Roger