A detailed overview of noise in Astrophotography

In astrophotography we have to deal a lot with noise. But what is noise exactly? In order to be able to deal with noise and to improve your images in terms of Signal to Noise ratio (SNR), it is vital that we have a basic understanding of noise and the different types of noise we encounter when imaging our deep sky objects. For this overview I’ll use a very generic definition of noise; Noise is all the undesired signal. There are different sources of noise, and to identify them it is useful to have a look at how the imaging by your camera actually works.

The imaging sensor; an array of buckets in the rain

Whether you are using a CCD or a CMOS sensor (in most DSLRs) the analogy of buckets capturing rain drops is applicable. You can view each pixel on your sensor as a bucket and the photons as raindrops. Just like a bucket that is collecting water by catching the raindrops, the pixels on your sensors are detecting light by collecting the photons that come from your deep sky object you are imaging. This is the signal. The shutter of your camera opens, and the raindrops start falling in the buckets. When you close the shutter again you have lot’s of buckets that have a certain amount of water in them. This is the build up charge that is held in each pixel. Some buckets might be completely filled, which represent the full saturation of a pixel. This is the clipped signal. A bucket might even be overflowing some water into the bucket next to it, this is blooming.

www.microscopyu.com
image by www.microscopyu.com

When you have even rain, each bucket is not collecting the exact same amount of raindrops every time. There will be some random variation which we call shot noise. This is the nicely distributed statistical noise we talked about in the article on the benefits of adding more frames in stacking.
After closing the shutter, the bucket will be emptied to measure the amount of water that it caught, but this measurement won’t be 100% accurate all the time. This is your read noise. When the bucket is emptied, you might have some water that is left behind. This is your bias. Then you have the possibility that you have an inherent source of water within the array of buckets, maybe condensation, which is the dark current.
Some buckets will be broken and not capture and/or hold any water any more, the cold pixels, while others might never be emptied anymore and are always full; hot pixels.
And then you have of course all the other sources of water that may also be hitting your buckets. Maybe the neighbour is watering his plants and you catch some of those drops, which is analogous to stray light. But we have more sources of signal we don’t want to pick up; think of cosmic rays (yes you are actually catching those! How cool is that?!), airglow and of course satellites and airplanes.
All these unwanted sources will deposit some number of drops in the buckets, photons on your pixels.

A dark frame has multiple types of noise; read out noise, dark current, bias and hot-/cold pixels
A dark frame still has 3 types of noise; read out noise, dark current and bias

Summarising the types of noise

Let’s take a more detailed look on the types of noise we identified;

  • Shot noise or photon noise
  • Shot noise is the fluctuations of the number of photons that are detected due to their occurrence apart from each other. This noise is inherent to the nature of light and therefor can not be prevented. Due to it’s statistical occurrence however, we easily can get rid of them by combining multiple detections; stacking.

  • Read noise
  • The error that occurs when reading the charge of the electrons by the amplifier. The pixel will accumulate a number of electrons, which are measured in microvolts per electron. Before we can pass this signal on to the Analog/Digital Converter (ADC), we need to read and amplify the signal. This is where we encounter the read noise, as the amplifier can’t do a perfect job. This is also influenced by build quality. Well designed amplifiers exhibit high linearity and introduce minimal noise.

  • Dark current or thermal signal
  • Dark current is the build up of thermally generated electrons in the pixels. The rate of dark current accumulation depends on the temperature of the sensor. That’s why cooling is such a big thing in astrophotography. Since most DSLRs don’t have cooling, this is perhaps the most important part of unwanted signal. You can use darks to ‘subtract’ this thermal signal, but you will have to carefully match the temperature since the dark current accumulation is temperature dependant. An other way to get rid of most of the dark current signal is the use of dithering. Shot noise is also applicable on dark current. So next to the dark current signal, there is dark current noise. Please note that dithering doesn’t take care of this dark current noise.

  • Bias
  • This type of noise is very structured and inherent to the camera design. Because of the many electronics there will be interference with the very sensitive electronics of the sensor, resulting in noise. Luckily we can get rid of this noise quite easily because of it’s highly structured nature by using bias frames. Just create a superbias and calibrate your frames with that.

  • Hot- and cold pixels
  • The hot and cold pixels are basically two forms of ‘broken’ pixels. In both cases no signal is acquired anymore by those pixels. They are a problem in two ways; they present them selves as unwanted signal or a lack thereof and they prevent us from capturing the photons that fall on those pixels.
    We can get rid of them in our final image by using some form of pixel rejection. Furthermore we can use dithering to shift the image over the pixels so we do capture the photons that would otherwise be falling on the dead pixel.

  • Cosmic Rays
  • Even though this is of course unwanted signal, I can’t help but thinking it’s very cool we are actually capturing cosmic rays on our image sensors.
    Cosmic rays are high energy particles that continually bombard the Earth. Some cosmic rays are generated by the sun, while others originate from far outside of our solar system. Cosmic rays will appear as a bright cluster of pixels or as line of bright pixels somewhere on your sensor. We can get rid of them fairly easy by using pixel rejection in our integration process.

  • Air glow
  • Because of various processes in the earth’s upper atmosphere a faint but significant amount of glow (light/photons!) is visible at all times. Especially low on the horizon this becomes a real factor to take into account when imaging, because you look through more of the atmosphere than if you were to aim your telescope straight up. That’s (partly) why you normally don’t want to image objects when they are low on the horizon. This is not something to get easily rid of, so we will have to make sure our signal stands out strong enough above this airglow signal. This effects the faint(est) signal the most of course.

COMMENTS

  • […] that down to them to provide you with what your looking for. In the meantime you can read this A detailed overview of noise in Astrophotography | DSLR Astrophotography I hope it is of help to […]

    • Ruben

      Thank you Chris your sharing is much appreciated. I was in fact looking at consolidating all types of noise and unwanted signal that affect astrophotography and your paper is one that goes the extra mile.
      I found that Fixed Pattern Noise with its two components DSNU and PRNU are relevant to my DSLR also sensor glow.
      Would these be included within some of the items in your article? If not would dithering mitigate them?
      Thanks for your help.

  • Excellent article about noise , it makes all basic cleared !!! It would be very helpful if you provide some sample pics and explain more details.

  • Alen

    “Shot noise is also applicable on dark current. So next to the dark current signal, there is dark current noise. Please note that dithering doesn’t take care of this dark current noise.”

    Photon noise is shot noise, as you say. But the random part of the thermal signal is also shot noise, as you also say. So best to avoid using “shot noise” as an alternative (actually first) name for photon noise. Secondly, it is true that dithering (with pixel rejection/clipping) doesn’t take care of dark current noise. But that is also true for dark- frame subtraction, which you omit to say.

    • chrisvdberge

      Excellent remarks and you are right on both points. I chose to use ‘shot noise’ as this seems the more commonly used term, but I might have use Poisson noise to avoid confusion like you point out.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.