Noise is a problem for any system, but digital systems are more immune to it than analog systems. This is because margins are built into the digital portions. However, when noise does get into a digital system, it doesn't degrade nicely. It generally has no correction mechanisms built in, and when it does, such as with a high-definition TV signal, it can be very intrusive.
High-speed communications is an interesting middle ground where digital data is being transmitted — but it relies on an analog connection. So jitter, supply noise, and other factors have to be taken into account to ensure a clean digital signal at the other end.
Noise in an analog system is often subjective — at least for consumers of such devices. We look at specs for an amplifier to see how much distortion it creates, but it is difficult for us to measure it. My girlfriend is an audiophile and prefers the sound of valve amplifiers — a subjective assessment. At the end of the day we say, “I like the way that one sounds.”
I am a photographer, and I am very happy for the digital revolution that has made taking photographs quicker and cheaper in terms of seeing results. We can now often do real-time learning and correct for mistakes, and have cameras that are powerful enough to help us overcome a number of difficulties.
But at the heart of this digital camera is an analog component — the sensor. The quality of the sensor is one of the things that sets one camera apart from another. At the same time, we push those sensors to the extreme where noise knowingly becomes a problem. Let me take a step back for those who may not know too much about cameras and define some terms.
Taking a picture requires a balance of three primary things in order to get a correct exposure. The lens aperture determines the amount of light that reaches the camera sensor. The wider the aperture, the more the amount of light that is let through. At the same time, the wider the aperture, the shallower the depth of field. Depth of field means what range of the objects in the picture (at different distances from the camera) will be in focus.
The second factor is the shutter speed; how long the light is allowed to hit the sensor. Too slow and things in the picture may have moved and become blurred.
The third is what is called ISO (International Standards Organization). This is a term that dates back to film cameras and tells us the film speed — how sensitive the film's chemical emulsion is to light. While it explicitly indicates light sensitivity, it implicitly tells us the grain size — hence image resolution. In a digital camera, the sensor only has one setting, but there are amplifiers that boost the signal from the sensor and these provide a way to adjust the sensitivity, so we continue to use the ISO terminology.
Now, both the sensor and the amplifiers are noisy. Here is a picture that was taken at a high ISO setting — meaning the gain on the amplifier was turned up.
At this point, you probably can't see the noise, but let me zoom into it a little more.
I can assure you that nobody's skin looks like this. This is noise from the amplifiers, but it is not a problem until you want to do a large print, or blow it up as I did.
There is a second kind of noise that comes from the sensor and is a little more difficult to see because it only shows up with very long exposures. This is caused by heating in the sensor because it is basically designed to only be on for short amounts of time, but when you want to take pictures at night, exposure times of many seconds become required.
I wanted to do some experiments to actually measure the noise in my camera, and in part two of this blog I will share those results with you.
I wonder if Pointillism should be considered as noisy art, and what those paintings would look like if passed through a noise reduction filter?
Next week, some additional thoughts on noise as it manifests in my digital camera's light sensor.