|
Foveon and Sigma claim that the SD14 has a color resolution advantage over the Nikon D80 and many other 10MP DSLRs. But there can be a big difference between color resolution and black and white resolution. Traditionally, black and white targets such as the standard IT-10 (now called ISO12233 test target) have been used to test a digital camera's resolution. Before digital cameras, the 1951 USAF test target was used in the Pop Photo Lab to determine film resolution and to confirm the results of our lens bench analysis when testing 35mm and other interchangeable lenses.
Black and white test targets are a natural choice for testing resolution, since the human eye is more sensitive to monochrome details, and most camera systems are optimized for recording b&w details. Foveon and Sigma basically agree that the Foveon X3 sensor doesn't outperform other digital cameras with similar pixel counts (as determined by Foveon, which claims its X3 has 14.1 million pixels) when measuring resolution using either the IT-10 target or 1951 USAF target (or any number of other b&w targets). However, both claim that the unique design of the X3 sensor gives it the ability to outperform cameras such as the 10MP Nikon D80 when measuring color resolution with a chart (or set of charts and filters) that compare pairs of primary and secondary colors (RGB and CMY).
Click photo to see full sensor test comparisons of the Sigma SD14 and Nikon D80.
Their argument is that CCD and CMOS sensor-based cameras that use a Bayer pattern color array often sacrifice extremely fine details in color images. The Bayer pattern consists of two green pixels for every red and blue pixel, emphasizing the detail gathered in the green channel over that gathered in the blue or red. The Bayer Pattern was designed to take advantage of the human visual system's increased sensitivity to the green portion of the visible spectrum (see Part II of this test on the next page). In post processing, Bayer-pattern cameras derive the most b&w resolution data from green pixels with additional data from the red and blue pixels. Foveon's X3 sensor overlays R, G, and B pixels so that there are even numbers of all three colors. Black and white resolution can be determined by combining the exposure data from all three sensor layers in post processing.
One way to test color resolution is by placing specific color filters in front of the lens, but as you know, any filter in front of a lens can affect the system's resolution and may cause some color crosstalk in post processing. The better method is to use a color resolution test target similar to the Foveon Color Zone Chart sent to us by Foveon.
Click photo to see full sensor test comparisons of the Sigma SD14 and Nikon D80.
For our tests, we equipped both DSLRs with highly rated lenses (Nikkor 50mm f/1.8 and Sigma 18-50mm f/2.8 EX ) and manually focused each lens on the target (a series of through-focus tests helped determine the sharpest image). The distance to the target was determined by the Nikon D80's ability to resolve detail to the outer edge of the black and white section of the chart (Section C). Because of its 1.7X 35mm lens factor (vs 1.6X on the Nikon), the Sigma SD14's distance to the target was then lengthened slightly to reproduce the exact same field of view as the Nikon (when set to 50mm focal length position). The result was a matching field of view and the same perspective.
Test photos were shot in RAW format and converted to 16-bit TIFF files using the manufacturer's RAW conversion software (Sigma Pro 3.0 and Nikon Capture NX). In the case of the Nikon, we also converted the RAW-NEF files using Adobe's Capture Raw tool in Photoshop CS2. The Sigma software lets you convert its RAW files to the largest pixel size of 5280 x 3520 pixels (18.6MP). For Nikon images we converted RAW-NEF files in Nikon's Capture NX to normal size and then up-sampled them in Photoshop so that the targets reproduced at nearly identical pixel dimensions and made it possible for better side-by-side comparisons. For the images converted in Adobe's Capture Raw, we set the pixel dimensions to 5120 x 3427 (17.5MP). In all three cases, up-sampling didn't increase resolution but made jagged edges appear slightly smoother.
We were surprised in this step to discover how well Adobe's Capture RAW converter performed in comparison to Nikon's Capture NX, and preferred the results from Adobe in nearly every section of the chart.
|
