This article is more than 1 year old

Snapper's decisions: Whatever happened to real photography?

Automatic for the People

Zoom trends

Digital cameras initially copied the viewfinder or SLR model, and later so-called bridge cameras were available in which the optical viewfinder was replaced with a live view on a screen and mirror-less cameras having electronic viewfinders. However, due to the economics of silicon manufacture and yield, digital cameras mostly have sensors that are smaller, sometimes a great deal smaller, than the 24 x 36 frame of 35mm. Thus for the same field of view the focal length of the lens would have to be reduced.

Unfortunately, optical systems do not scale because the wavelength of light has to remain within the range visible to the human eye and so the energy of photons doesn’t change. If a lens with no tolerances or manufacturing defects could be made, it still could not focus to a point because of diffraction limiting that causes point spread that gets worse as the aperture is reduced.

There is no point in engineering a lens that is fantastic at full aperture and deteriorates as it is stopped down, so a good lens will have approximately balanced point spread due to aberrations at full aperture and diffraction at minimum aperture. The best performance will be somewhere in the mid range.

Bayer sensor filter pattern

Bayer sensor filter pattern – source: Wikimedia, Cburnett, Creative Commons

One immediate consequence is that the number of pixels output by a camera has no direct bearing on how sharp the pictures will be, even though pixel count features heavily in the specification. The first point to bear in mind is that digital camera sensors don’t have pixels; they have photosites. Most single chip cameras capture colour images by having a Bayer pattern of three different colour filters distributed over the photosites, so that some capture red, some green and some blue.

As the eye is most sensitive to green there will usually be as many green photosites as red and blue together. The raw data file that is typically recorded in the camera memory is the photosite data and needs, inter alia, de-mosaicing to compute pixels that produce a digital image. Typically four photosites are needed to create one pixel.

There’s no point in having enormous pixel counts on a small sensor, because once the resolution limit of the lens is reached, increasing the pixel count is futile. A four-thirds format digital camera is unlikely to deliver more than four megapixels of information per frame, irrespective of how much data it outputs. Pixel count is to cameras what sampling rate is to Pono: both allow marketing to quote a bigger number to the gullible.

Bayer pattern on sensor profile

Bayer pattern on sensor profile. Photo via Wikimedia, Cburnett, Creative Commons

Another great leap backwards for photography was the proliferation of zoom lenses on consumer digital cameras. A glance at the optics of a good wide angle lens and a good telephoto lens will reveal that the bits of glass are somewhat different. Trying to be both at once is going to be a compromise that will result in loss of aperture and loss of resolution. But that’s not the worst problem.

The problem is that the unskilled user simply stands where he is and zooms in to fill the frame with the object of interest and ends up with an image that is as flat as a cow pat and completely lacking in perspective. The photographer gets off his butt, trudges closer with a fixed wider lens, spends an age selecting a viewpoint and comes back with a photograph. As Ansel Adams said: “A good photograph is knowing where to stand”.

As photons don’t scale, smaller photosites just get noisier, which is why cameras in iPhones croak in the dark. There is such a thing as an ideal size for a photosite, which is where four of them more or less fit in the point spread function of the lens so sensor and lens match. The number of pixels then depends on how big the sensor is. As sensors go, big is beautiful and Moore’s Law is irrelevant. Big sensors offer the best resolution and the lowest noise which translates into high dynamic range.

Next page: In the frame

More about

TIP US OFF

Send us news


Other stories you might like