Sub Exposure Selection

[novalab]

Hi! I apologize for writing in English, but my native language is only good enough for non-technical discussions, and talking about the girls.

There has always been a lot of stigma attached to using short sub exposures. Until recently, this was always a topic of heated discussions, and people throwing up bunch of equations, personal experiences, but none of them were able to capture and quantify the actual system performance, and configure it for optimal AP/deep sky operation.

The truth is that long exposures do indeed afford the best SNR and deepest limiting magnitude, they also have serious drawbacks. If your sub will be 600s, your tracking/guiding better be near perfect. Can you say "Mount/OTA mechanical flexure? And call the air traffic control to re-route all the planes. And needless to say, any brighter stars will saturate, and possibly start to bloat the image. Shorter SUBs are cheap to throw away, but 600s SUBs hurt.

Today's cameras sport a very low read noise, so stacking images is not very detrimental. The important factor to consider is the ratio of Read noise to background Poisson noise (sqrt. of background signal). Larger this ratio, longer subs you will need.

So let's look at an example of a quintessential C8-XLT with 0.63x reducer and QHY268M camera (IMX571M chip, and a totally random choice). We will use 1 hour total exposure for this example, and will simulate this scenario with no filters, using the Astrophotography CAD application which numerically simulates the entire system, star to fits file. I will note that results will be different for different setups, and the observing conditions set can be seen in the pic below.



The chart above shows how the SUB duration affects the image. As you can see, the limiting magnitude using a 5s sub is ~20.349 mag, and the ultimate sub exposure of 3600s yields 21.412 LM. The sky of 19.5MPAS used in this example will not saturate the background at 3600s exposure in this system, and only fill the well to ~17%. So, about 1 magnitude difference, which is very significant.

However, if you look very carefully you will notice that there is very little change in LM (Blue curve) after 30s SUB duration. So, for this particular system, with the sky and the seeing conditions and the camera used, you will be just fine with 30s – 60s subs. Again, different systems will behave differently: for example my 24"RC likes SUBs od 180s or longer, and my C11 prefers 60-120s. Change the sky, seeing and everything changes.

Now look at the purple curve – Saturation Magnitude. A 5s sub will have unsaturated (linear) dynamic range of 8.3mag-20.4mag. A person using, say, 600s sub will operate in 13.5mag-21.4mag range. A big difference! Again, the 30-60s sub seems like a good compromise here.

The above is done using 1x binning. Since CMOS cameras always perform Off chip binning, this will artificially increasing the binned pixel read noise, so the results will be somewhat different at increased binning. The CCD cameras allow for ON-chip binning, and this will not be a factor, unless you need to bin some crazy number. 

In summary, although using long SUBs always yields the best LM, most systems will operate in near optimal range with 30 – 180s SUBs, without having to deal with serious technical limitations inherent to long guiding. In general, bigger the scope, longer the sub.

You can learn more about apCAD at https://nlobs.com/astrophotography-cad-info/

[tome]

Thank you for details and explanation.

I noticed that you are referring to mono cameras. Would you apply the same logic to OSC cameras?

From urban sky I tried to expose for longer but I don't see any difference past 60 sec subs.
Only for dual band filters. With newtonian 6'' f6.7 and imx533C.

[Astro Lux]

To simplify all of this im always going for this formula and this reference chart:

t = 50 x (read noise)² ÷ electrons/pixel/sec   

the 50 is regarding to reach top 1% over the minimal level  and if you want top 5% change the 50 to 10.

Read noise can be evaulated in sharpcap or by checking camera manufacturers spec sheet/graph .. modern CMOS on LCG usually 4e and HCG from 1.5-1e

Electrons/pixel/sec can be calculated via this site: https://tools.sharpcap.co.uk/

And for everyone that doesnt want to do calculations you always have this image from Dr Robin Glover(creator of sharpcap)

[novalab]

Hi Tome,

You are quite correct. For each system and operating environment there is a minimum optimal sub exposure. Beyond this duration, there is very little improvement. In the case I showed above that is about 30 seconds. Yours will differ, and I can't tell you what that is without knowing more about it. In your case, this seems to be below 60s, so I would stick with 60s as a simple to keep in mind. Also, shorter subs increase your overall download time, so there is always some good compromise between that, guiding, light pollution and dynamic range.

We only analyze Mono cameras but allow users to change filters. Why? Because the Color cameras are actually 3 cameras with 3 different filters: R, G and B. The Sub exposure selection I described is equally valid, but you need to carry out the analysis for each filter separately, and then use the longest Sub exposure you arrive at. Remember, different chips have differently shaped QE curves – not all are proportionally equally sensitive in, say, Red. For this reason our App has Sony Chip R, G and B filters in the filter dropdown list. Please note that it is not only important to know your typical sky glow NELM or MPAS, but you also need to estimate your sky photo-chromacity. Sky with predominantly LED street light glow will have most of the photons coming in blue (450nm), thus heavily affecting the Blue Subs. If you have predominantly HPS lights (yellow) it will be very different. We provide users ability to specify the sky type in addition to the brightness.

Please read my answer to Astro Lux below, as it will have more information, so I do not repeat myself here.

[novalab]

Hi Astro Lux andd Tome,

While I understand the need for SharpCap to provide at least some guidance on the sub duration, this approach is very simplistic, inaccurate, and does not take into account many external factors. I do applaud them for doing it, and do not in any way degrade what they do (an excellent app), but Astrophotography CAD is specifically designed to perform these kinds of complex analysis. In particular, the analysis you quote regarding the F# does not agree with the comprehensive numerical simulations in the slightest (and will be a separate topic). I will talk about this as concisely as possible, but am not able to write a full white paper about it here:

There are many factors that affect the minimum optimal sub:

To start with, Filters used and the sensor QE curve shape heavily modulate the apparent photon flux, and will also vary for different sky photo chromatic content. Suburban sky, heavy with LED lights will be quite different to that from HPS yellow street lights (or a mixture of both. In color cameras you have 3 filters, and the blue in particular will be very dependent on many factors (also on elevation etc.). The optimal sub in color cameras should be the longest of the three colors to preserve the sensitivity and also the color balance. The 3X they recommend is a SWAG (silly wild-ass guess).

Below: LED Sky, Sony Blue Pixel


Below HPS Sky, Sony Blue Pixel


The camera quantum efficiency also has a significant impact. For example the Sony sensors have much better blue response, but very poor red/IR response. GSense and Kodak are quite different resulting in as much as 50% sub change for some cases. This does make quite a bit of difference. Again, the optimal sub in color cameras should be the longest of the three colors.

The F# does matter, but not in the way you expect. It was a surprise to us all to discover thru multiple simulations that each system has an optimal F# (or EFL), and took some time to understand why. In short, the camera read noise has a lot to do with it, and higher the noise shorter the optimum EFL. I will do a post on this separately as it is too complex to examine here. For now, look at the 2 examples below using KAF-8300 on 8" RC scope. In one case the camera has 0e read noise, and in the other it has 9.3e read noise. I chose a CCD, because CMOS sensors are even more complicated because they always bin Off-Chip (thus artificially raise the noise), introducing binning as yet another variable.

KAF-8300, No Read Noise


KAF-8300, 9.3e Read Noise


Continued in the next post... Sorry, I reached a pic upload limit!

[novalab]

Continued...

Another factor that has an impact is the length of the total exposure – and the ADC quantization level. Why? If your 25 mag star produces only 0.01e per sub exposure, and it takes 20e to flip one ADC bit, you will never see it, no matter how many subs you average. This is a simplistic explanation and probability also comes in play, but it will demonstrate a basic principle. So, SUB duration, ADC bits, as well as camera gain in e/ADU have a great effect on exposure LM (and image SNR). Let's look at an example of IMX428 sensor. This is a 12 bit cooled CMOS chip, which I chose to easily see things. The 16 bit cameras are less impacted, but are still affected for long exposures of > several hours). We'll use 1 hour total exposure here and 2 different subs, 30s and 90s:

IMX-428, 30 second sub


IMX-428, 90 second sub


As you see, with this system it makes no sense to expose longer than 15minutes with a 30s sub, and you will not see any more stars. Going to a 90s sub considerably improves the LM (almost a full magnitude for the same 1 hour total exposure), and allows you to extend the total exposure to up to 108 minutes to gain more LM.

There is so much more I could write on this topic, but my time is short today, and there is work to be done...

I hope this shines some light on the subject, and helps you with your imaging.

[Astro Lux]

I read this and I will stick to the simple solution, I like to know the science behind it but its pretty straight forward with the sharpcap formula and it does make sense. Ofcourse the F# doesnt really affect as it should be because aperture pixel size and well the pixel scale has more meaning than just a F# on its own. I would like to add that there should always be a compromise of something. And that even though you are using CCDs in the comparison and the 571 as the random cmos variable, there should for "amateurs" or astrophotographers always be a CMOS leaning metrics that not even that sharpcap provides because its already outdated with the new sensors. And yeah for OSC the logic is always like this for RGGB bayer pattern you are getting 3 colors but they are not placed the same neither is the QE value optimised for them especially on modern CMOS like IMX571 IMX533 IMX455. But the 3x multiplier makes sense if you dont account that people image in LRGB or HaRGB etc.

Im just speaking from an astrophotographer perspective because I dont want to get into the complexities of maximal reaching magnitude of stars that you are getting on those graphs, and more in the sense of
THE question which is usually what people ask, for my bortle/SQM value my telescope with this aperture and focal lenght and this camera how long i should expose for broadband and narrowband, what gain and offset to use and if doing mono how should i place the ratio on those images ?