The fact that sensors are kept small, but get sectioned into more and therefore smaller pixels, leads to problems with sensitivity. The explanation for this is easy to understand. The lens still passes the same amount of light onto the sensor. But if there are more and smaller pixels, there is less light left for each one.
Every pixel then converts the light into electrons. Let’s pretend these electrons are water and the pixels are buckets. Depending on how much light hits the pixel, the bucket gets filled with water. Whenever the bucket is full and more light hits it, the water spills over or even floods the surrounding area. The result is white areas in a picture.
When a pixel gets divided into 4 smaller pixels, it means the bucket becomes smaller and receives just a fourth of the water. However a certain amount of water is needed because some water could get spilled during transport and thus the final amount is falsified.
This minimum amount needed doesn’t change through having smaller pixels. This means, with smaller pixels the minimum amount of light has to be 4 times greater in order to receive the necessary amount for every pixel. At the same time, the buckets become smaller which leads to faster flooding. This results in the reduction of the difference between the brightest and the darkest area in the picture, which both show details (dynamic range).
In order to compensate the lesser degree of sensitivity, the signal from the sensor needs to be amplified so that smaller “amounts of water” become visible. However, drops that were spilled or flooded appear as noise in the picture.
Please enlarge a picture for details.
The following sections are extracted from pictures taken with cameras with a different number of megapixels. On every picture you’ll see the sections in the following order:
upper left: Fuji F31 fd w/6.3 megapixels, upper right: Olympus 725 SW w/7.1 megapixels
lower left: Canon 950 IS w/8 megapixels, lower right: Casio Exilim Z1200 w/12.1 megapixels
Please note that the cameras were randomly chosen. According to our experience they show a typical behaviour for the represented camera class. To make the displayed images believable we decided to mention the cameras we used but we do not want to criticize a single model or manufacturer but the pixel race in general.
In "Picture center, ISO 100", you can see that Olympus and Canon smooth out parts of the tree trunks, the path and the bushes to the left. The difference of the detail reproduction between the 6.3 and the 12.1 megapixel camera is not very significant. It is obvious that Casio smoothes out as well and sharpens contours that are easy for it to recognize. Furthermore, additional errors become visible because of JPEG compression, e.g. in the form of white dots (visible in the bushes).
In "Picture center. ISO 400", it is obvious that the Fuji F31 fd is the camera that features the highest detail reproduction and has the smallest amount of megapixels in our comparison. The others seem blurry.
In the crops of "Picture edge, ISO 100", you can’t see many details in the Olympus image and just moderate details in the 8 and 12.1 megapixel cameras. In the examples of the last ones, the grass stalks seem to melt into each other. Single stalks and little stones are still recognizable with the Fuji F31fd.
It is obvious in "Picture edge, ISO 400": the 6.3 megapixel camera shows the most details.