Words and scans by David Millier, images shot by and © Erik Muehlberger
The Foveon X3 sensor based DSLR range from Sigma provides the only current alternative to the ubiquitous Bayer Colour Filter Array sensor type that is used in just about every digital camera. The Foveon sensor has its photosites arranged in 3 colour layers rather than the more usual single-layer 2-dimensional mosaic pattern.
There are advantages and disadvantages to both approaches; one advantage of the Foveon approach is that fewer pixels are required in the final image file to render a sharp image, giving smaller file sizes for the same enlargement potential. Unfortunately for Foveon and Sigma, this technical advantage has backfired from a marketing perspective. In a megapixel obsessed world, trying to persuade customers to buy cameras equipped with fewer but higher-quality pixels is a tough sell.
Sigma have adopted a controversial solution. They appear to have abandoned their earlier attempts to re-educate the buying public as to the usefulness of MP counts and now market their cameras by their photosite count instead of the traditional pixel count. A photosite is a single light sensing element in a sensor. Foveon sensors use vertical triplets of photosites to build each pixel so, using the traditional method, Foveon sensors appear to have fewer pixels than you might expect for the price point – giving them a bit of a marketing headache.
This change of tack provides them with the marketing opportunity to triple the claimed MP count of their cameras in brochures and advertising. While some might consider this misleading (because the standard pixel counting method is so well established), I have no doubt Sigma and their fans would argue the conventional method undersells the quality of the Foveon sensor. This move is understandable but it has certainly caused some fractious squabbles on online forums!
This test sets out to compare the resolution of the Foveon sensor against the traditional Bayer CFA competition and to see whether Sigma are justified in their MP count approach or whether it is merely a marketing gimmick.
We decided to pit the latest Sigma camera against a conventional camera based on megapixel claims. The Sigma SD14 is described as using a 14MP sensor (more accurately, 14 million photosites), so we decided to put it to the test against the 14MP Bayer sensor Kodak 14nx.
We don’t have fancy lab test equipment, so this is more of a field test of real world shooting. One of us (Erik) shot 12 pairs of test images using equivalent high quality prime lenses – a 50mm Sigma macro on the SD14 and a Nikon 85mm F1.4 on the Kodak. The test shots were shot at medium apertures for a reasonable compromise between maximising depth of field and minimising lens aberrations (F8 on the Sigma and F11 on the Kodak).
The cameras were mounted on a tripod and a number of shots taken using focus bracketing. The sharpest files were selected.
Displaying the results
One of the problems with internet-based test publishing is the difficulty of presenting the results in a way that doesn’t favour one camera or the other. A typical problem is how to deal with demonstrating differences where file sizes are not the same. We can upsample or downsample to match file sizes – but this can cause dispute over the algorithm used. Our solution is simple: we printed the results at 19 x 13 and 24 x 16 inches and scanned the prints on a good flatbed. Small crops were taken from the scans and these are paired for comparison.