DXO Analyzer test results
Dpnow has invested in a DXO
Analyzer digital camera test system. We have set up a permanent dedicated test studio that accommodates the 1.2m by 0.9m DXO Analyzer test target and our other optical test equipment. The DXO test target is mounted on 1cm thick plate glass, weighing 25kg and, in turn mounted on a custom-manufactured stainless steel stand. As the studio does not need to be set up again each time it is used, lighting and other variables remain consistent from day to day.
This is the first review we have published using DXO Analyzer test data. We believe the use of DXO Analyzer provides valuable test information to supplement, but most definitely not replace, our more conventional use of standard ISO resolution and Gretag Macbeth colour charts and, of course, analysis of actual images taken using the camera being tested – both on-screen an in prints using an Epson Stylus Photo 2100 printer.
In brief, the DXO Analyzer system provides computer analysis of test target images taken by cameras on test to provide data on curvilinear distortion (barrel and pincushion bowing), chromatic aberration (where colours separate due to insufficient focus correction in the lens optics), vignetting (darkening of the corners of the frame), image noise and, finally, image sharpness, measured in a DXO-proprietary Blur eXperience Unit (BXU).
To help put the DXO Analyzer test results for the Casio Exilim EX-Z40 into perspective, we present a comparison with selected results from a similar specification competitor camera, in this instance the Fujifilm FinePix F440. Both cameras are compact, slim and have a similar specification, price and target market.
All camera lenses exhibit a degree of chromatic aberration (CA) because the component colours of white visible light cannot be focussed by a single lens in one spot. You can see this easily for yourself by using a magnifying glass to project an image onto sheet of paper. Especially towards the edges of the scene, you start to see rainbow colours where you'd expect to see solid hues – this is CA. Modern camera lenses use a combination of lens elements to optically correct for CA but it's impossible to eliminate completely across the entire focussing and zoom range. Digital camera purple fringing, found at bright, high contrast edge boundaries in an image, is not optical CA in the conventional sense and this test is not designed to measure purple fringing. We look for purple fringing, in real world test images.
The graphics below show measured CA plotted across the image frame, as indicated by the colour scale under each plot. The colour scale does not relate the colour of any aberration but to the extent, in pixels of the problem Blue (starting at zero) is good and red (to three pixels) is bad. The sharpness and shape of boundaries in the colour maps also indicate whether CA gradually appears towards the corners, which is good, or if there is a sharp or non-symmetrical degradation.
Casio EX-Z40 (left) vs Fujifilm F440 (right)
At the wide angle zoom setting and at full aperture, the EX-Z40 has a fairly typical compact camera CA plot. The F440 shows a more gradual trend towards the corners – both are pretty much normal for the type of camera.
At a mid-zoom range setting the EX-Z40's average CA level shows an increase but the corner trend is less steep and the same can be said of the F440, though the latter has the edge in the centre of the frame.
At full telephoto zoom, the EX-Z40's CA map has become asymmetrical and particularly poor in the left hand corners. We repeated the test target shot on several occasions to eliminate any possible errors, but the results were consistent. The F440's lens extends its advantage clearly over the EX-Z40 at this zoom setting.
Here we show results, once again, for each camera at a wide, mid and telephoto zoom setting to evaluate any geometric distortion of the image. Two plots for each camera at each zoom setting are shown, one representing the actual distortion as if it had been photographed and reproduced normally and the second showing a 5x amplification of the distortion to magnify the error.
Both cameras perform well given the wide angle zoom setting tested, with slight barrel distortion being the norm.
At the mid-zoom setting the curves have almost disappeared, with the Pentax-sourced lens of the EX-Z40 marginally superior.
At full telephoto setting there is hardly any visible distortion. Some cameras will produce visible inward bowing of the margins (pincushion distortion) but not here – both cameras have passed this test well.
At first site this might seem overly technical, but it's easy to understand really. A Blur eXperience Unit or BXU is a measurement of perceived sharpness in an image. It was devised by DXO to approximate the way the human eye sees the sharpness – or lack of sharpness characterised as blur – in an image. This test was designed to differentiate resolution from sharpness as an image with good resolution can also look blurry. The test doesn't really differentiate between 'good' sharpness and 'bad' over-sharpening that is found in a lot of cheaper cameras, but we will let you know if over-sharpening is detected in our visual appraisal of real world test images and our ISO resolution test target shots.
The main thing to remember with BXU, the smaller the unit value, the better!
The BXU test has immediately highlighted a problem with our EX-Z40 we already noticed in our test shots – very poor sharpness in the diagonally opposite corners top left to bottom right, although sharpness in the centre is good. Remember, the smaller the BXU number the better the sharpness. This would suggest that the EX-Z40's image sensor is out of alignment with the lens along the top-right to bottom-left axis.
The F440's results are much more even. The EX-Z40 has superior sharpness at the centre, though looking at our ISO resolution test results for both cameras, it does appear that the EX-Z40 applies more in-camera sharpening than the F440.
At the mid-zoom setting the EX-Z40 has recovered some composure, but the F440 remains more even and consistent.
The F440 remains consistent, though with an average BXU rating of nearly 3, sharpness is not especially good and this was certainly evident in the test photos we took.
At full telephoto zoom setting the poor sharpness in the EX-Z40's result has migrated to the opposite diagonally facing corners. This would suggest that the lens mechanics are not aligned consistently throughout the zoom range. This could be a design defect, a manufacturing problem or even the result of a knock to the camera, but if it's the latter, there is no external sign of damage. We're expecting to test another Casio camera which uses the same lens soon and we will be looking out for similar inconsistencies then.
Vignetting is the darkening of the corners of an image visible in skies and other scenes where there is an expanse of light area. Vignetting can be noticeable in digital cameras for a number of reasons including physically small diameter lenses in relation to the sensor size and lens optics not optimised for digital sensors.
As with the CA test earlier, blue is good and red is bad. When you see a cross in the image map, this denotes the central point of the vignette when it is not dead centre in the frame.
In both cameras, as you would expect, vignetting is minimal at the telephoto end of the zoom range. The F440 once again proves superior overall especially at the wide end of the zoom.
We're currently working with DXO Analyzer version 1.25 and although it does measure noise, for various technical reasons the way it works means we can't usefully compare results with different cameras. We're informed that version 2.0 of the software, expected to be available some time in November, in conjunction with an additional test target, should solve this problem. In the mean time, below we can show the increase in noise the higher the ISO setting of the EX-Z40 and, in particular, the much worse noise detected in the blue channel.
Although blue channel noise is a common problem with digital cameras, causing graining of blue skies, for example, the unusually large difference between the overall luminance noise level and the blue channel is a poor result.