Studio Tests - Lumix G HD 14-140mm F4-5.8 OIS lens
The GH-1 is only sold in a kit with the video-optimised Lumix G HD 14-140mm F4-5.8 OIS lens, so we've taken the unusual step of evaluating the lens as part of the camera review process. As usual our results are obtained by shooting RAW and converting the files using Adobe Camera Raw at standardized settings. However, analysis of this lens's performance is complicated by the fact that some fairly extensive software correction of lens aberrations is clearly going on 'under the hood' - with the camera removing geometric distortion and lateral chromatic aberration from the final image, in a fashion which will be completely transparent to the vast majority of users.
The 14-140mm puts in a very good performance for a superzoom in our studio tests - it's at its best in the 18-50mm range, but gets soft towards the long end of the zoom. It's more or less on a par with the cream of the current crop of APS-C image-stabilized superzooms such as the Tamron 18-270mm F3.5-5.6 VC, and therefore marginally more consistent than the 18-200mm zooms from Canon and Nikon. Thanks to the use of software corrections, it registers minimal distortion or lateral chromatic aberration.
|Sharpness||Sharpness is pretty good for a superzoom - it's not at all bad at 14mm, and very good indeed between 18mm and 50mm, but then declines progressively towards the long end. Typically for the Four Thirds format, sharpness tends to be consistent across the frame into the corners, with optimum results obtained at apertures around F5.6-F8.|
|Chromatic Aberration||Lateral chromatic aberration is being corrected in software, and therefore to all intents and purposes is negligible.|
|Falloff||We consider falloff to become perceptible when the corner illumination falls to more than 1 stop less than the center. There's about a stop of falloff at the extremes of the range wide open, which disappears on stopping down - really nothing to worry about.|
|Distortion||Geometric distortion is being corrected in software, and is therefore negligible across most of the image range. There's some residual barrel distortion visible at wideangle (1.6% at 14mm), but this is a notable improvement on the extremes normally seen with superzooms.|
The 14-140mm's specified closest focusing distance is 0.5m at all focal lengths, which will give a working distance of about 0.34m (13.5") between the camera and the lens. Maximum magnification is 0.2x, which equates to an image area of approximately 87 x 65 mm - about average for a superzoom.
Software correction of lens aberrations
One of the more controversial aspects of the Micro Four Thirds system is Panasonic's decision to integrate software lens aberration correction as a fundamental component of the imaging chain (which the company has unfortunately chosen not to publicly document). For most users this is completely transparent - the camera corrects both the viewfinder image 'on the fly', and the JPEG files it records. Also, both the supplied SilkyPix software and industry-standard alternatives such as Adobe Camera Raw convert raw files correctly too (the relevant correction parameters are encoded directly in the raw file, and Adobe has recently revised its DNG open-standard raw format to accommodate them properly too). However users who like to experiment with more obscure raw developers which are unable to apply the requisite corrections will find themselves with highly distorted images, mainly when shooting at wideangle.
To see exactly what is going on, we converted the raw shots from our standard distortion test using dcraw, which does not apply any corrections. This reveals that the 14-140mm has extremely high barrel distortion at 14mm, which becomes more modest at 18mm, and this can be seen clearly below (the grids are scaled to give an accurate representation of the transformation that is actually occurring). At longer focal lengths, distortion turns out to be extremely low.
|14mm uncorrected||14mm corrected||18mm uncorrected||18mm corrected|
It's interesting to consider this behavior in comparison to current 18-200mm lenses for APS-C/DX sized sensors. The graphs below compare the uncorrected and corrected distortion results for the 14-140mm with a typical superzoom, the Nikon AF-S 18-200mm F3.5-5.6G DX VR (we've used 35mm-equivalent focal lengths to compare the lenses at roughly the same diagonal angles of view). Positive values on the vertical (y) axis indicate barrel distortion, and negative values pincushion.
The Nikon has fairly severe barrel distortion at wideangle, but also shows significant pincushion distortion across the rest of its range. Panasonic's approach starts with highly distorted images at wideangle, but gives much better corrected final images, while (in principle at least) allowing the lens to be made smaller and less complex.
Lateral chromatic aberration
The studio test data also clearly shows that lateral chromatic aberration is being corrected in the ACR-converted raw files (out-of-camera JPEGs are the same). The practical upshot of this is that the camera produces images which are essentially free of the color fringing which is normally visible with this type of lens, most obviously towards the corners of the frame. Note that this isn't anything new with JPEGs - Panasonic's compact cameras and Four Thirds DSLRs have clearly been doing it for a few years - but this is probably the first time that it has been applied in a mandatory fashion to third party raw conversions.
Again we've used dcraw to process our studio test raw shots in order to reveal the lens's true characteristics, and illustrated the fringing visually using our the top left checkerboard pattern from our lens test chart (this is a 100% crop direct from the test image). For the sake of clarity, the 14mm and 18mm samples were also corrected for distortion using Photoshop's 'Lens Correction' filter (with values of +12 at 14mm and +5 at 18mm), to give reasonably square checkerboards. There's pretty strong fringing at the extremes of the range; at wideangle it's of the most visually objectionable green/magenta type, and at telephoto mainly blue/yellow in color. However it's not really any worse than a typical APS-C 18-200mm superzoom.
One curiosity is that, while Panasonic is correcting lateral chromatic aberration in software, its partner in Micro Four Thirds, Olympus, doesn't seem to be following suit. Consequentially, the 14-140mm will reveal all of that CA to anyone who chooses to use it on an Olympus body (such as the EP-1).
So is it cheating?
With all this software correction of lens aberrations going on, it's perhaps inevitable that some people will look upon Panasonic's approach as 'cheating'. After all, the argument runs, film camera lenses were always properly corrected optically, so surely the use of software to achieve the same effect is simply cost-cutting, and therefore somehow 'cheating'.
We think this is fundamentally the wrong way to look at it. In photography, what ultimately counts is the final image - the means to get there is relatively unimportant. As we've shown, Panasonic's software correction results in pictures which are technically superior to those obtained using more conventional, purely optical correction methods. This is no doubt helped by the fact that none of the software manipulations used are particularly extreme (aside perhaps from those for distortion at wideangle) - we'd be much more worried if software was being used as a crutch to support the use of significantly sub-par optics. It's also worth appreciating that this new approach is only possible because of the opportunities offered by a system that has no optical viewfinder, and therefore no requirement for the lens to be corrected for SLR-type viewing.
Now some users have a more rational concern that systems which incorporate software correction won't be 'future-proof', and lenses which require such assistance won't be usable in future. This seems highly unlikely - the trend in software development is clearly to embrace such opportunities, as shown by the recent revision of the DNG specification to include lens correction instructions. It's not so long ago, after all, that raw converters offered no form of manually-adjustable noise reduction, and now no self-respecting piece of software would be seen out in public without the requisite chroma and luminance noise sliders. It seems likely that in a few years time, automatic lens aberration correction will become equally universal.
Some doubts have also been raised about whether such 'optically imperfect' lenses will be fully usable on future cameras, but this seems unlikely to be a real concern - at least as long as the Micro Four Thirds system itself continues in existence. We're told that distortion correction (at least) is a fundamental aspect of how MFT operates - the correction parameters are stored on a ROM inside the lens itself for communication to the camera, and written into the raw file for converters to use. So these lenses seem to be no more likely to become unusable in future than those for any other system.
So our considered opinion is that no, it's not cheating, and indeed this kind of processing is clearly the future. All of the major manufacturers are attempting to address lens flaws in software in one way or another, and one advantage of mirrorless system with purely electronic viewing is greater flexibility in what is possible. So the best approach for the photographer is to accept what technology can offer, and enjoy the images it can produce.
- 1 Introduction
- 2 Specifications
- 3 Body & Design
- 4 Body & Design
- 5 Body & Design
- 6 Operation & Controls
- 7 Operation & Controls
- 8 Operation (Live View)
- 9 Displays
- 10 Menus
- 11 Menus
- 12 Performance & IS
- 13 Photographic tests (RAW)
- 14 Photographic tests (Noise)
- 15 Photographic tests (Noise)
- 16 Photographic tests (DR)
- 17 Photographic tests (Kit Lens)
- 18 Photographic tests
- 19 Movie Mode
- 20 Compared to
- 21 Compared to (JPEG)
- 22 Compared to (JPEG)
- 23 Compared to (JPEG)
- 24 Compared to (RAW)
- 25 Compared to (RAW)
- 26 Compared to (RAW)
- 27 Compared to (Higher ISO)
- 28 Compared to (Resolution)
- 29 Compared to (Resolution)
- 30 Conclusion
- 31 Samples