By Hans Weichselbaum www.digital-image.co.nz
The progress we see in technology is breathtaking. It was only ten years ago when the serious photographer considered switching from film to digital. Then point-and-shoot cameras started to rival the heavy hardware, and today you can see everybody shooting with their mobile phone. Do we really need a large camera, especially now that we get 20 megapixel (MP) cameras built into our phones?
The sensor is the key, the heart of your camera. It dictates the size of the lens and the depth of field, but also determines the image size, resolution, low-light performance, dynamic range and, of course, the camera’s overall size. There has been lots of progress in technology and engineering, but you can’t bend the laws of physics. In the end, every pixel site on your camera’s sensor is collecting photons and the more it can pick up, the better. A small sensor can only give you a limited image quality, no matter how many MP it has got.
Many of my readers will remember the pixel race, starting at around 4 MP, about 10 years ago. Camera manufacturers were falling over each other to beat the competition with new MP records. Then camera makers realised that better means fewer pixels and reduced the pixel density again. Today the MP race is back on once again, this time with mobile phones and tablets with their teeny-weeny sensors.
Demystifying Digital Camera Sensors
The sensor size is often referred to with a ’type‘ designation using imperial fractions such as 2/3″ (inch) or 1/1.7″, which are larger than the actual sensor diameters. These units go back to the TV camera tubes from the 1950s. The sensor size is about two-thirds the size of the imaging circle. Image 1 gives you an example of a 2/3″ sensor. Note that most compact cameras have an aspect ratio of 4:3, whereas the full-frame (35mm film) has a 3:2 ratio.
Today, sensors range from about 1/3.6″ (5.0mm diagonal) to full-frame (43.3mm diagonal), and there are even larger medium-format sensors.
Sensor Size and Pixel Pitch
The more pixels we squash onto a given sensor size, the higher the resolution, but at the expense of smaller pixels. Smaller pixels collect fewer photons; the signal needs more amplification, and we get more noise.
The optimum pixel size (pixel pitch, measured in microns or µm = 1/1000 mm) is around 5–9 microns. Cameras with >9 micron pixels have problems with aliasing (low frequency artefacts showing up as moiré pattern). On the other end, pixels of less than four microns gather fewer photons, giving a weaker signal that needs more amplification, which in turn leads to more noise.
Point-and-shoot cameras go down to a pixel pitch of less than two microns and the problem only gets worse when you look at cell phone cameras. Image 2 shows you a few of the most common sensor sizes in comparison.
At every pixel site on the sensor photons are converted into electrons. You can think of a bucket collecting rain drops – the bigger the bucket, the more water (= light) we can collect. The electrons make up a charge which is fed to the A/D converter. The accuracy of the signal (electron charge from the individual pixels) is directly proportional to the size of the signal. The noise in the signal is equal to the square root of the number of photons.
For example, nine photons would give us a signal-to-noise (SN) ratio of three. For 100 photons the SN ratio improves to 10, etc. Why do we need to know this? Well, it turns out that the noise in modern cameras, from the simplest phone camera to the best DSLR, is dominated by photon counting statistics (for any signal above a few hundred photons). Read noise is only important if there are fewer than a few tens of photons, and only for very long exposures do we need to consider thermal noise. There is very little room for improvement by increasing the quantum efficiency. We have seen improved micro lenses over the pixels, gathering more light, but the fact is that we have almost reached the physical limits.
Other Consequences from Small Sensors
Image noise is one of the main problems that come with tiny sensors, but there are other factors you need to consider:
- Dynamic Range: This is defined by the maximum signal divided by the noise floor at each ISO setting. Cameras with larger pixels are mainly limited by noise of the A/D converter and can be improved by around two stops with 16-bit converters, whereas the small-pixel cameras do not collect enough photons to benefit in this area.
- The Crop Factor: some people dislike this term and call it ‘focal length conversion factor’, but it is a crop factor because the part of the image circle outside the sensor gets cropped. It is easy to calculate by dividing 43.3 by the sensor diameter. For example, Canon’s PowerShot series has a 1/1.7″ sensor with a diagonal of 9.4mm, giving you a 43.3/9.4 = 4.6 crop factor.
- Diffraction: this is the resolution limiting factor at small apertures. The aperture where the system becomes diffraction limited depends on the sensor size (eg. it is around f/5.6 for most point-and-shoot cameras). That’s why the smallest stop on a point-and-shoot camera is usually f/8.
- The depth-of-field: a smaller sensor will give you a larger depth of field (DOF). In theory, we do get the same DOF with a larger sensor by increasing the ISO setting and stopping down the lens. It all boils down to the number of photons trapped by the individual pixels. But yes, in practice the smaller sensor will give you an apparent larger DOF, making it more difficult to isolate that portrait shot from the background.
In the next issue we’ll be looking at three cameras, a full-frame DSLR, a point-and-shoot camera and a mobile phone. We’ll compare image resolutions and noise performance.