Featured Article

Guide to Sensors

Digital camera sensors work by converting light energy (photons) into electrical energy (electrons). This energy forms a signal that’s transferred to the camera’s processor.

The sensor is made up of millions of individual photosites, each one corresponding to a pixel in the final image. A six-megapixel digital SLR has six million pixels on its sensor.

Actually, there are a few more than this, but many pixels around the edges are used for calibration and other purposes. This is why you’ll usually see two figures quoted for sensor resolution: total pixels and effective pixels. The effective pixels are those used to form the image and are the most frequently used.

But there’s a complication. Each photosite, or pixel, is sensitive only to the quantity of light, not its colour. Film manufacturers got round this by incorporating several layers of film, each sensitive to a different colour. Most digital sensors, with the exception of the Foveon X3 sensor (see the boxout), have only one layer. What the makers do, then, is cover each pixel with a red, green or blue ‘micro-filter’.

In fact, it’s been found that the human eye is most sensitive to the green part of the spectrum and manufacturers arrange pixels in groups of four: two green (two pixels with green filters over them, in other words), one red and one blue. There are twice as may green pixels as blue or red pixels. This arrangement is known as the ‘Bayer pattern’.

The image data now has to be processed to produce full-colour data for each pixel. The camera looks at the surrounding pixels to estimate or interpolate this colour information. For example, a green pixel would have data from the red and blue pixels around it added, in order to produce a full-colour RGB (red, green and blue) value. The same applies to red and blue pixels. This colour interpolation is sometimes called ‘demosaicing’, since in it’s non-interpolated form, the arrangement of red, green and blue pixels, resembles a mosaic.

Like other interpolation processes, demosaicing is a technical bodge. Colours (or the outlines of coloured objects) are softened slightly and you may sometimes see slight colour bleed from one area to another when images are magnified on the screen.

There’s another source of blurring with digital images. DSLRs use anti-aliasing or low-pass filters in front of the sensor. These blur the image slightly before it’s recorded by the sensor. This is to prevent regular rectangular patterns in fabrics, distant fences or the corrugated cladding on industrial buildings interfering with the rectangular array of pixels on the sensor and producing moiré as a result. Some DSLRs can appear sharper than others even if the sensor resolution is the same. This may be due to differences in the strengths of the anti-aliasing filters.

This filter, incidentally, covers the sensor completely. If you ever need to clean dust off the sensor, it’s the anti-aliasing filter you’re cleaning, not the sensor surface itself. The filter surface is usually quite hard and resilient and will only be damaged with careless or vigorous cleaning. Nevertheless, you undertake any sensor cleaning at your own risk!

The combination of colour interpolation (demosaicing) and the anti-aliasing filter takes the edge off a digital SLR’s resolution. This is counteracted by the sharpening applied by most cameras as the image is processed.

The number of megapixels on the sensor is no longer a reliable guide as to how good  a camera’s pictures will be. In the early days of digital photography, sensor resolution was the limiting factor, but now we’ve gone past that point and the physical size of the sensor is paramount.

Return to features archive »