Pre-PMA 2008: Dpreview is pleased to announce the launch of a new section dedicated to testing and reviewing digital SLR lenses. Designed to offer the same class-leading testing standards as its legendary camera reviews, dpreview’s new lens reviews are the result of months of intense research, development and testing, including the development of new test charts and proprietary analysis software designed to overcome the limitations of existing systems.
Dpreview launches lens reviews
At the heart of the reviews is a unique interactive ‘widget’ has been designed to present the test results in a way that allows visitors to experiment visually with variables such as focal length and aperture and see immediately the effect such changes have on overall image quality, and to compare the performance of different lenses. The user-friendly presentation ensures that even those with limited technical knowledge of optics or digital imaging will be able to use the reviews.
The first four reviews are available to read now at www.dpreview.com/lensreviews/. Many more lenses are currently being tested and we expect this section to grow in a short time into a valuable and comprehensive resource for anyone looking to buy a lens for their digital SLR.
About the tests: in detail
The DPReview lens tests are the culmination of months of research and development by our imaging and programming team. Rather than use an off-the-shelf image analysis software package as a ‘black box’, we decided to develop our own custom test suite in-house, in accordance with our long-running philosophy of offering the best possible data to our readership. As always, we wanted the numbers to be backed up by real images, so you can view standardised samples from each lens across the full spectrum of focal lengths and apertures, and make your own assessment.
Our tests therefore have been designed from day one to describe and quantify the most important optical characteristics of any lens, specifically sharpness, chromatic aberration, distortion, and light falloff (also known as vignetting). In addition we test other key parameters, including macro performance and the effectiveness of optical image stabilization units. And last but by no means least, this is all placed in the context of the extensive amount of time we spend actually using the lens for real photography, backed up by galleries of sample photographs to help you judge its real-world performance for yourself.
How the tests are done
Our studio testing is carried out by photographing a set of sample images of our own in-house lens test charts under standardised conditions. Crucial factors such as lighting, exposure, focusing, and camera alignment are all strictly controlled, in order to ensure that our results are accurate and reproducible.
1) Sharpness/chromatic aberration
This is the most complex of the tests. Our chart is uniquely designed so it can be easily interpreted by computer, to generate numerical data, and also by our readers: for each measured data point, we also provide a checkerboard pattern so you can see how the numbers translate to real imaging performance. The MTF-50 sharpness data on the graph correlates directly with the perceived sharpness of the pattern, and the chromatic aberration data with both the colour and width of fringing. So rather than asking you to simply take our word for it that our data is correct, we allow you to check for yourself.
Shooting is very strictly controlled – the camera is aligned by laser precisely with the centre of the test chart, and the camera-chart distance kept between 2 and 4 meters to minimise any possible alignment errors and avoid introducing other subject distance related issues. To achieve this we use test charts of a range of sizes, the largest being 3 x 2 metres permanently fixed in our studio which is used for wideangle lenses. Focusing is critical, and we use live view with magnification wherever possible to ensure maximum accuracy.
We use black/white slant edge test targets across the entire frame to generate MTF and CA data, which is processed automatically by DPR-exclusive software routines. A checkerboard pattern is placed next to each test target, and 100% crops of these are automatically extracted and made available for you to view.
The data from the entire area of the frame is averaged to give the MTF and CA curves displayed on the graphs; we’ve found the lenses we have tested to date give generally symmetric data, with most variations in the extreme corners of the frame. We collect MTF data at each marked focal length on the lens’s zoom ring and each full aperture stop, plus at the maximum and minimum apertures when these do not coincide with full stops.
Results from this test are dependant upon the camera body and image processing used, which is why we explicitly identify the camera used for each review (note that real-life shooting with each lens, as reflected in the gallery pictures, use several different bodies to identify any specific issues). To provide a as level a playing field as possible between different camera platforms, all of the data is shot in RAW and processed using our standard converter, Adobe Camera Raw. Our processing workflow has been developed after consultation with Adobe, and the effects of software enhancement (sharpening etc) are as low - and as equal across brands - as they can possibly be.
Distortion data is generated by photographing our grid-array test chart, and analysed automatically using our own routines. Because distortion is effectively independant of aperture, we use F8 for each test. We report the type of distortion (barrel or pincushion) with percentages for each dimension of the frame; this is because distortion is often complex and uneven across the frame, for example from manufacturers re-correcting barrel distortion on wideangle zooms by ‘pinching’ towards the edge of the frame.
As for the MTF chart, the camera is precisely laser-aligned to minimise errors. It’s important to realise that distortion tends to be highly distance-dependant, and for some lens designs increases dramatically at closer focusing distances. Our chart is approximately 1.05 x 0.7 metres, and gives distortion figures perhaps slightly greater than when the lens is focused at infinity.
We measure light falloff by photographing an evenly-illuminated surface through a highly diffusing filter with the lens focused to infinity. We shoot at each marked focal focal length on the lens, and at 1/3 stops down to F11. The falloff patterns are then analysed, and compared to a pre-determined exposure curve for each camera body. This allows us to determine the falloff pattern in stops. For easier visualization we posterize the image in 1/3 stop increments for display; we also calculate the maximum falloff values at the extreme corners of the frame and report these.
We consider falloff below one stop to be essentially imperceptible, and certainly nothing to worry about. We therefore don’t show the first 1/3 stop on our charts at all, to simplify the display.