Following on from our look at what a CCD sensor really is, we’re taking a look at the types of noise associated with them. Noise is an unavoidable feature of image sensors, but there are plenty of ways in which we can control it and keep it to an absolute minimum. This is particularly important when trying to detect very small signals, such as those we’re trying to capture when we image the night sky. This why all our CCD cameras are designed to be as low-noise as possible, but what do we really mean when we say this?
The kind of noise we’re talking about comes in three main flavours – read noise, shot noise and dark current. We’re going to take a look at all three over the coming months, starting with read noise. In this video Steve Chambers explains what read noise is in relation to a CCD camera, where it comes from and what we do to keep it as low as we can.
Transcript – Read Noise with Steve Chambers
Hello. What I’d like to talk about in this video is read noise and how it applies to CCD cameras.
It’s worthwhile starting off just by mentioning that when we talk about noise, we’re not talking about sound noise. This is electronic noise – uncertainty – that relates to the measuring of voltage, or a digitised voltage. So if we’re expecting a value from a pixel of 50, maybe one time we measure it, we might get a value of 55, the next time we might get a value of 48. The reason why we don’t measure the same value each time really relates to the amount of noise that’s accumulating within that signal.
Most sources of noise within a CCD camera have what’s called a normal distribution so we can deal with it in a very simple, statistical way.
Normal distributions have this kind of bell shaped curve. We describe the standard deviation of this normal distribution as being, well 68% of the values will be contained within plus or minus one standard deviation of the mean.
So when we describe one of our cameras as having 3 electrons read noise, what we’re saying is that if we just look at the read noise element of the uncertainty, then 68% of the values would be contained within a region of 6 electrons worth of ADU counts.
So where’s this noise coming from? Well, the most important source is the CCD itself and the amplifier that’s doing the conversion of electrons into voltage.
These amplifiers work phenomenally well, so you can amplify just a handful of electrons and produce a voltage from them. But there is a noise associated with that and one of my jobs as a camera designer is to make sure that the rest of the electronics involved in handling that signal are adding less noise than the actual CCD itself is adding, and then we get to the point that the limiting factor within the camera is the CCD itself, and that’s not something that we can change.
So we have the CCD amplifier, the main source of noise, then we have some level shifters, and circuitry associated with correlated double sampling if it was present. These are things that we are adding, adding voltage to these levels, or maybe we’re subtracting voltage, and we’re using things like operational amplifiers to do that.
Finally we then digitise that value, and we then we have the analogue to digital converter which in all our cameras’ cases is 16 bit ADC.
So the read noise can be defined as the noise added between the charge conversion, the conversion of electrons into a voltage, and the digitisation phase.
If we look at some of the values from our cameras, they really are class leading. We’ve got values for say, the 420 of just 3 electrons, the 414 with just 4 electrons, up to the Infinity where we’re looking at a camera that’s uncooled and has faster readout rates and that’s only 6 electrons worth of read noise. And for CCDs these are very, very good values indeed, and will mean that what signal you’ve got – you’ve got a very small signal, you’re really interested in signal to noise ratios – the noise element becomes as small as possible, and your signal, the smallest, smallest signals become detectable.
So that’s read noise, and how we control it.
Thank you very much.