How the Test Data is Shot
Test images for are shot in RAW at the camera's base ISO, and processed using a common converter (Adobe Camera Raw) with all sharpening disabled – this eliminates any differences from in-camera processing. Using RAW is crucial, as many current cameras can apply lens aberration correction to JPEGs - for example many Nikon DSLRS feature automatic correction of lateral chromatic aberration, and recent Canon EOS models feature automatic peripheral illumination (falloff) correction. Using JPEGs from these cameras for testing would therefore not provide a true description of the lens itself.
Sharpness is calculated from the slanted-edge patterns arranged across the frame; the test chart has four ‘arms’ extending from the centre to each corner of the frame, and the sharpness data we display is obtained by averaging the results from each, to give the most accurate representation of the lens’s performance.
For wideangle to normal focal lengths (up to 50mm equivalent), we use a test chart 3m x 2m in size, which gives a far more realistic working distance than the A0 size charts more generally used (for example, using a 24mm equivalent lens the subject distance is 2m as opposed to 0.72m). Due to space constraints, we revert to a smaller chart (approximately 1.05m x 0.7m) for longer focal lengths. This gives a shooting distance of about 30x the 35mm-equivalent focal length of the lens.
Correct focusing is critical, so where available we use magnified manual focus in live view, which is the most reliable and consistent method available. On cameras that don't feature live view, correct focus is established by determining the maximum sharpness obtainable from the lens using our in-house analysis software. In all cases, at least two replicate data sets are shot to confirm reproducibility of the results.
The test chart is a simple grid pattern, and aligned extremely precisely with the camera. Images shot in JPEG are translated directly to the data displayed in the widget. In addition, for systems which are based on the use of automatic distortion correction (most notably Micro Four Thirds), we also convert the RAW versions of the test images using a program that does not correct distortion, and present these results later in the review. As distortion is essentially independent of aperture, we only shoot at F8. The chart is approximately 1.05m x 0.7m in size.
Results are obtained by shooting an evenly-illuminated white wall through a highly diffusing ExpoDisc filter. We shoot in JPEG with any in-camera vignetting correction turned off; the displayed results use greyscale values derived directly from the original image files. The falloff value in stops is calculated using the measured tone curve of the camera body used for testing.