Astrofriend's homepage


Advertisement /

Project News

Advertisement /

Camera time efficiency simulation

Camera simulation

  1. Introduction
  2. Why I like these cameras
  3. The Excel sheet with camera simulations: Low Dynamic, example 1
  4. High Dynamic, example 2
  5. Thermal Noise, example 3
  6. High Dynamic and very weak signal, example 4
  7. Readout Noise from calibration, example 5
  8. Download of Excel sheet

1: Introduction

I must be a strange person, sometimes when I have boring times I can spend hours to calculate things. Today I got to think about cameras and how efficient they use the time when doing astrophotographing. It's not just the QE that's important.

My four favorite cameras today are (if I can afford them):

  • Atik 16200 mono
  • Canon EOS 6D
  • ZWO, ASI1600M Cool mono
  • QHY, QHY367C
  • I just have the Canon 6D, very satisfied with it, but still curious on other cameras. The Excel sheet I have setup simulates the above cameras.

      It calculate:
    • What focal length you need to have a pixel scale of 1" / pixel.
    • How long it takes to saturate the pixel for a max level you choose.
    • How many sub images it takes to reach a given S/N (Signal / Noise relation) for a weak object, it also include light pollution and thermal noise (late version).
    • The total time it takes included the dead time between the sub images.

    This is very simplified and don't use this to decide of what camera to buy. But it give an interesting overview what happens under different circumstances. I will say all these cameras are very good, what you get depends on how you use them and your environments, heavy light pollution or not, out door temperature etc and the object of course.

    I don't take any responsibility for it, all risk on you!

    Back to contents

    Advertisement / Annons:

    2: Why I like these cameras

    Here is a short information why I like these cameras.

    Atik 16200 mono:

    The Atik 16200 has the newly developed APS-H CCD sensor, it's size is between a full frame and a APS-C crop sensor. It has a QE of about 0.56 and 16 bit ADC. CCD sensor technique are a bit old compare to CMOS sensors, but still going strong.

    It cost a lot of money.

    More information:

  • Atik 16200 mono
  • Canon EOS 6D:

    Canon's ESO 6D are an affordable camera with full frame sensor, it has together with the QHY camera the biggest sensor of these four cameras and are the cheapest. Its QE of 0.5 is high, and after modification with all filters removed it can be even higher. Very low readout noise, 1.7 e- at high ISO setting, but high ISO gives very low dynamics. Color Bayer pattern sensor.

    Not a cooled camera.
    It's a color camera.

    More information:

  • Canon 6D
  • ZWO ASI1600:

    ZWO has lately developed a lot of interesting cameras. ASI1600 has a 4/3", a bit smaller than an APS-C. The sensor are built with CMOS technology. At high gain it has an impressive low readout noise of just 1.2 e-.

    People reporting amp glow but that is something that can be handled.
    It has the smallest sensor of these four cameras.

    More information:

  • ZWO, ASI1600M Cool mono
  • QHY367C:

    QHY has a lot of interesting cameras. Lately they have offered a full frame CMOS camera with Sony sensor. It's a color camera, but there are ways to handle the image as a normal rgb mono chrome process. See my AIJ macros I use to Canon DSLR cameras: ../tutorial-astroimagej/tutorial-aij-01-introduction.html. This camera has a cooler and then we don't get the problem with high dark current in hot environment. Readout noise 2.4 e-.

    Color camera.

    Perfect Camera:

    To make it a bit more exiting, why not put in a almost perfect camera to compare with? Zero Noise, high resolution ADC, no dead time etc. And even a real color camera like the Foveon Foveon technology or even better, Red, Green, Blue and IR at the same pixel, a four layer construction. It's optimized for a 10" RC telescope with a 3" x1 field flattener.

    It doesn't exist! -Yet.

    More information:

  • QHY367C
  • Table of important data:

    Namn Atik 16200 mono Canon EOS 6D ZWO ASI1600 QHY367C Perfect Camera

    Pixel width x high

    4499 x 3599

    5472 x 3648

    4656 x 3520

    7376 x 4938

    4096 x 4096

    Pixel pitch

    6.0 my

    6.58 my

    3.8 my

    4.88 my

    10 my

    Sensor size, width x hight

    27 x 21.6 mm

    36 x 24 mm

    17.9 x 13.4 mm

    36 x 24.1 mm

    41 x 41 mm



    0.5 (0.55 mod)

    0.65 ?

    0.55 ?


    ADC resolution

    16 bit

    14 bit

    12 bit

    14 bit

    18 bit

    Full Well capacity

    40 ke--

    80 ke--

    20 ke--

    56 ke-

    100 ke-




    Electronic rolling shutter

    Electronic rolling shutter

    Electronic shutter

    Image download time, seconds

    8 ?

    1 or 8

    0 to 2

    5 sec ?


    Back focus distance

    19.5 mm

    44 mm

    6.5 mm

    18 mm ?


    Cooling delta T

    50 o Celsius


    45 o Celsius

    35 o Celsius



    1.3 kg

    0.76 kg

    0.41 kg

    0.79 kg

    1 kg

    Back to contents

    Advertisement / Annons:

    3: The Excel sheet with camera simulations: Low Dynamic, example 1

    Excel sheet:

    The Excel sheet is very simple, but I have done some refinement lately. The goal with it is to get the time it take to have a final image with a given Signal to Noise ratio. These calculation are thought to have adapted telescopes that give equal photon flow per pixel for the different cameras because of different pixel sizes.

    Here is an overview of the Excel sheet and first example:

    Camera time efficiency, example 1

    Data fields:

    The green fields are the camera data, some examples:

  • ADC, Analog Digital Converter's bit depth
  • QE, cameras photon efficiency
  • Full Well capacity, how many electrons a pixel can store, small pixels store less
  • Dead time between photos, could be very important when doing Lucky Imaging
  • Pixel pitch
  • It's not so easy to find the needed data, in some cases I just have estimated a value. You can replace them if you have a better idea of what it could be. Beware, behind the blue and green fields there are formulas, don't overwrite them!

    Input fields:

    Your input data are the red fields. Take a look at your old photos to have an idea what your telescope and camera delivers. Fitswork is easy to use for this purpose, use the CFA (before transform to color) if it's a color camera. Fitswork subtract the bias by automatic. Here I work only with the green pixels. If it's hard to recognize them do a linear interpolation deBayer with gain = 1 for all color to get a color image, examine interesting (saturate level, weak signals, background) green pixels in the center of images to not get any vignetting effects.

  • Number of green wave length photons coming to one pixel per second
  • The max flow of photons that you don't want to saturate, high levels keep the dynamic
  • The weak flow of photons of the object you taken photos of
  • The back ground photon flow from light pollution
  • Your goal to what S/N ratio you want to have in the stacked image
  • The figures I have put in here are not to uncommon, but 100 incoming photons as max level before saturate is low and 5 photons per pixel and second is maybe high. It depends very much on the situation, object and telescope. Out of focus images give much lower peaks in level.

    Output fields:

    The output fields after calculation are blue.

    It calculate which focal length that give a 1" / pixel scale for each camera and pixel pitch. It don't use this information but it could be interesting to know.

    At lower half follow a list of steps with different calculated values:

    • Step 1: Exposure time to reach saturation
    • Step 2: Noise from different sources
    • Step 3: S/N single sub image
    • Step 4: Number of images to reach S/N goal
    • Step 5: S/N in stacked image
    • Step 6: Time needed for one stacked image
    • Step 7: Time needed to rgb or retrive resolution for cfa
    • Step 8: Time needed to build mosaic based on no of pixels

    Step 1:
    This is thought to be the highest level that you don't want to saturate. Too long exposure and you get low dynamic but less noise in the subs. You can take many short subs and stack them and then get better dynamics. To succeed with short exposures the camera must have low readout noise.

    In this first example I have set highest signal that I don't want to saturate to 100 photons / pixel / seconds. Not very high but of course it depends on telescope or camera lens used. In most cases I will say this create a low contrast image with bright stars clipped.

    Step 2:
    As it's now it calculate noise from read out, dark current, object and light pollution. You can test with different signal flow from a weak object and different light pollution background. Analyze your old images, here is one method you can use: ../tutorial-find-photon-flow/tutorial-find-photon-flow.html. Light pollution noise can be high, but if you do narrow band imaging, like a H-Alpha filtered it's normally very low and then readout noise is more important to have low.

    Step 3:
    Here is the total Signal to Noise relation, S/N, calculated for one sub image. As you can see, a camera with small pixels has lower Full Well depth and get faster saturated, even worse with high gain. As a result you have to have shorter exposure time to not over saturate it. On the other hand with shorter exposure time you can take more of them.

    Step 4:
    In step 4 I do a calculation how many sub images it's needed to reach some S/N that you can set at the beginning. In this example S/N = 50. Heavy stretching of the image needs better quality. 50 could be a good value to start with. Number of images are rounded upwards. Compare the different cameras and the different settings of gain.

    Step 5:
    Here is the S/N ratio in the final stacked image, it differ a bit when number of images are low because of rounding error. They shall be close to the S/N value you set earlier.

    Step 6:
    Final we get how many minutes that is needed to make this image with different settings, it include the dead time between exposures.

    Step 7:
    This is an extra step. It gives information how much longer time it takes to make a color image from a mono chrome sensor, or how to recreate the lost resolution from a color camera with dithering technic ../tutorial-dithering/tutorial-dithering.html and Drizzle or similar method. See my other tutorials about Super Resolution with AIJ: ../tutorial-astroimagej-align-drizzle/tutorial-aij-align-introduction-matrix.html.

    Step 8:
    And last, it could be interesting to know how much longer it takes to built a mosaic from different cameras. If your interested in high resolution wide angle images this is important. In this case I refer to number of data points (pixels). It can be sensor area too. It calculates for seamless joints. Normally you have a overlap, but then you also get more exposure on the pixels, then less need of sub images. It gives about the same result when having S/N in mind.

    As you see, the time difference differ a lot between cameras. But you should put in your own values that depends on your situation.

    Test of different input values:

    It can be very interesting to test with different input values. In this case the saturation level was adapted to a very low max signal and then low dynamic.

    Even if a star is a point object in a normal telescope it spreads its photon not on only to one pixel, strong stars at least cover a radius of 3 pixels.

    In next part I increased maximum level ten times to get better dynamics.

    Back to contents

    Advertisement / Annons:

    4: High Dynamic, example 2

    Here I have changed one parameter, the maximum incoming photon flow that I don't want to saturate, now 1000 photons per pixel and second:

    Camera time efficiency, example 2

    Compare these values with Example 1 above, see Step 3: the S/N ratio of sub images and step 6: how many minutes it took to do the sequence.

    If you have a camera with mechanical shutter you shouldn't do many short exposure photos, it has a limited lifetime and will broke. You must have a camera with electronic shutter. The ZWO has USB3 ports, that can give very high speeds and can reduce the dead time a lot but it also heat the camera.

    Which camera is best? I will say all of them are very good, but it depends very much of what you will do and what telescope and environment you have.

    Back to contents

    Advertisement / Annons:

    5: Thermal Noise, example 3

    I was curious to see how the thermal noise from an uncooled DSLR camera influence the total S/N. With the latest version of my Excel sheet I can simulate and calculate dark current and thermal noise. If you are lucky and live in a cold climate country as I do with temperatures around zero degrees or lower it doesn't increase the total S/N very much, but if you have +10 degrees or more it's bad. Here is a screen dump that show how it differ from Example 1 with 0 o C sensor temperature on the Canon 6D DSLR camera:

    Camera time efficiency, example 3

    Test with different input parameters, you will see interesting things happen. Look at the Canon 6D at ISO800, sensor temperature 0 o C degrees (example 1) and +15 o C degrees (example 3):

    Compare above Step 2 with example 1: Thermal Noise, 6.8 vs 16.1.

    Compare above Step 3 with example 1: S/N, 17 vs 15.

    In this example I have set the temperature for the cold cameras to -20 o C (red field), I think that is good a value in normal climate countries. The DSLR camera sensor is set to +15 o C, to have that on my Canon 6D sensor the out door temperature has to be +9 o C or lower. The camera built-in temperature sensor that measures the cameras inside temperature is 6 o C above the surrounding temperature. If I use the live view the camera increase the inside camera temperature with a whopping 20 to 25 o C degrees relative the surrounding temperature, that take a long time to cool down again.

    Back to contents

    Advertisement / Annons:

    6: High Dynamic and very weak signal, example 4

    Lets play with the numbers. Say that we want high dynamic and a very weak signal. In this case the noise from background is 2.5 stronger then the weak object signal. Can the cameras handle that in theory? Note: This is just a simple simulation, lot of important things are not included, one example, static pattern.

    Camera time efficiency, example 4

    This simulation show that the perfect camera needs a little more then one hour for this weak object to reach S/N = 50. The real cameras from 2 to 5 hours. Could this be done in reality? Maybe I can do a test with my Canon camera how far I can come.

    Looks good, but in reality very short exposures on weak object will be difficult to align, good tracking solve that, but then no dithering.

    Back to contents

    Advertisement / Annons:

    7: Readout Noise from calibration, example 5

    Do you calibrate your images with dark images? Or even also bias images? Every calibration increase the readout noise. If taken many sub darks and sub bias it will reduce the noise, but zero images will give zero readout noise from them, but how? You can try dithering instead../tutorial-dithering/tutorial-dithering.html , it works well on my Canon 6D camera under some circumstances. Low static pattern is crucial from camera. It still give noise, I will calculate that later.

    Test with 0, 10, 25 sub images and look at Step 3: S/N single image. How much did the total S/N change?

    Back to contents

    Advertisement / Annons:

    8: Download of Excel sheet

    Here you can download my Excel sheet. But remember this is very simple made, don't trust it and use it for decision to which camera to buy.

    You download this and use it on your own risk! I don't take any responsible for it:

  • Excel sheet: Camera time efficiency simulation
  • History:

    version 20171023 (latest)
    Added readout noise from calibration, example 5
    Perfect camera optimized for 10" RC telescope

    version 20171022
    Added QHY367C camera
    Added more precise calculation
    Added more info text
    Added perfect camera
    Added high dynamic and very weak signal, example 4

    version 20171020
    Added thermal noise from dark current

    version 20171006
    Integer of number for sub images and cosmetic changes

    version 20171005
    First release

    Test with different input parameters, you will see interesting things happen.

    If you found this exiting and want to dig deeper in the details my friend Håkan has provided two links of information:
  • Fundamentals of Image Sensor Performance
  • Sony PDF: CCD and CMOS Image Sensors
  • Note:
    It could of course be some mistake in my calculation, I correct it when I found something.

    Back to contents

    Advertisement / Annons: