Lenses: overall impressions
The X-Pro1's initial lens set consists of the XF 18mm F2 R, XF 35mm F1.4 R, and XF 60mm F2.4 R Macro, and we've used all three extensively while testing the camera. With angles of view equivalent to 28mm, 50mm and 90mm lenses on full frame they cover the classic wide/normal/short telephoto triad, with the added bonus of 1:2 macro from the 60mm. We're not going to formally test the lenses here - instead we're going to give a brief summary of our experiences with them. For a second opinion based on a full set of technical measurements we suggest you look at Klaus Schroiff's reviews at photozone.de, which match up pretty well with our real-world experience.
Fujifilm's XF lenses may look very traditional on the outside, with their large aperture and focus rings, but on the inside they're very modern indeed, and take full advantage of the additional design flexibility afforded by digital capture. Most notably, it's clear that the X-Pro system makes full use of software corrections, which are applied automatically for distortion, vignetting and chromatic aberration. This helps make the X-Pro1's images look very clean indeed.
The X-Pro1 follows modern practice by integrating software distortion correction into its image processing chain. This is essentially transparent to the user - the electronic viewfinder feed is corrected 'on the fly' (so you compose using a corrected image), the camera's JPEG output is automatically corrected, and RAW files converted through both SilkyPix and Adobe Camera RAW are too. You only need to worry about it if you use a RAW converter that doesn't understand how to apply corrections.
It's easy enough to see what the camera is really doing, though; if you enable the 'Shoot Without Lens' menu option and partially unmount the lens (so it's still firmly held by the mount but not locked in place), then the camera shows the uncorrected image. This shows that the 18mm F2 and 60mm F2.4 both use software correction, for barrel and pincushion distortion respectively, while the 35mm F1.4 is fully optically corrected. We've illustrated this below using a traditional distortion chart (note that the camera moved slightly between shots with the 35mm).
|35mm, uncorrected||60mm, corrected||60mm, uncorrected|
One oddity is that distortion correction for the XF 18mm F2 lens is applied somewhat differently by the camera's JPEG processing, the supplied Raw File Converter, and Adobe Camera Raw. This is shown in the rollover below; the extreme edges of the frame are slightly compressed in the ACR conversion relative to the camera's JPEG. Raw File Converter, in comparison, does something very odd indeed, and manages to move objects in the centre of the frame. There's no such problem with the other lenses.
|Camera JPEG||Adobe Camera Raw||Raw File Converter|
It's important to appreciate that while only one of these versions can strictly be 'correct', none of them looks obviously wrong in isolation. You'll only really notice the difference in distortion correction if you overlay them like this, or look closely at the positions of objects at the edge of the frame.
Distortion correction-induced softness
Software distortion correction inevitably results in a slight softening of areas the image that get 'stretched' during the process. Barrel distortion correction stretches the corners of the image projected by the lens, and therefore tends not to be too problematic in normal use, as the corners of the image rarely contain critical image content. (Let's not forget that conventionally-corrected lenses often show soft corners too.)
The pincushion correction used by the XF 60mm F2.4 Macro is potentially a different story, because it stretches the centre of the image and is therefore likely to affect the main subject. The extent of this correction is small though (by our reckoning, about 2.5%). If the distortion correction is well-implemented, then the drop in sharpness should be pretty well proportional to the rescaling, and such a small decrease isn't likely to be a serious problem in the grand scheme of things. The crops below show the exact degree of rescaling involved here.
|XF 60mm F2.4 Macro @ F4, normally-corrected camera JPEG, 100% crop from centre|
|XF 60mm F2.4 Macro @ F4, uncorrected JPEG, 100% crop from centre|
The big question is, of course, how much this might affect your images in the real world? To illustrate this, we've taken an uncorrected image with the 60mm F2.4, converted the raw file with no sharpening, and manually applied the same amount of distortion correction as the camera using Photoshop's 'Lens Correction' module. We've then sharpened the uncorrected and corrected images by the same amount (Unsharp mask, Amount 200, Radius 0.3).
|XF 60mm F2.8 Macro, F4, uncorrected||Distortion correction applied|
|100% crops, centre of frame|
|100% crops, lower centre|
If you stare at these crops really closely, then you can see that the image has been slightly softened by the resampling process used for distortion correction, and in these side-by-side examples the corrected version just fractionally lacks 'bite'. But the effect is small; you can see it at 100% on-screen, but the impact on a print would be minimal. (In practice the camera also appears to compensate by applying a fraction more sharpening after correction.)
Overall, this means that images from the 60mm F2.4 Macro may not always quite match the pixel-level detail and microcontrast that typifies the 35mm, but they won't be obviously 'bad' either. Indeed if you're using the lens to shoot portraits wide open, then this normally won't matter at all - most subjects feel more flattered by a softer image. Likewise at the apertures typically used for macro work, any resampling effects will likely be overwhelmed by diffraction softening.
The X-Pro1 also applies vignetting correction in-camera, with the 18mm F2 using it most. This is illustrated in the rollover below, with all lenses set to their maximum aperture. A comparison between corrected out-of-camera JPEGs and the corresponding RAW files converted using ACR shows identical results, suggesting that the correction is applied before the RAW file is written. To be honest we'd prefer to have the option to turn vignetting correction off, although on aesthetic rather than technical grounds (vignetting quite often visually enhances, rather than degrades certain types of image, and we frequently end up adding it back in using Photoshop).
|35mm, F1.4, uncorrected||60mm, F2.4, corrected||60mm, F2.4, uncorrected|
The X-Pro1 applies chromatic aberration correction to its JPEG processing, and Raw File Converter does the same. However Adobe Camera Raw doesn't do so by default, and this allows us to gauge how the lenses look prior to correction. The conclusion from this comparison is that, even before any correction, the 35mm F1.4 shows no obvious lateral CA, while the 60mm F2.4 shows just an insignificant hint of the blue/yellow variety. Typically for a wideangle, though, the 18mm F2 shows visible green/magenta fringing towards the corner of the frame.
|18mm F2 @ F8, JPEG||100% crop from camera JPEG, top left|
|100% crop, RAW + RFC||100% crop, RAW + ACR|
The 100% crops here tell the story clearly enough. Adobe Camera Raw isn't applying any correction in this conversion, and cyan/magenta fringing from lateral chromatic aberration is clearly visible. In contrast, both the camera JPEG and the RAW converted through Raw File Converter are applying partial correction (technically they appear to be scaling the image's red channel to match the green, while leaving the blue alone). As a result the files show only blue/yellow fringing, which is generally less objectionable.
While Adobe doesn't correct CA by default, with ACR 7 the process is a simple 'one-click' option in the 'Lens Corrections' tab. When this is enabled, ACR removes the worst of the CA to give results very similar to Raw File Converter.
|ACR 7 offers a 'one-click' box option for CA removal in the Lens Corrections tab.||100% crop, RAW + ACR, CA correction enabled|
Longitudinal Chromatic Aberration
Unlike the lateral variety, longitudinal chromatic aberration (colour fringing in front and behind the plane of focus) doesn't appear to be corrected at all. However while it can be visible if you go looking for it, it's rarely any kind of problem in practice. The example below was shot using the 35mm wide open, which we'd expect to be the worst-case scenario with the current lenses. Longitudinal CA is visible as fringing that's green behind the plane of focus and magenta in front, but it's not hugely objectionable and certainly no worse than we'd expect from a fast prime.
|XF 35mm F1.4 @ F1.4, JPEG||XF 35mm F1.4 @ F1.4, RAW + ACR|
|100% crop, behind plane of focus|
|100% crop, in front of plane of focus|
Does software correction mean bad lens design?
There is a school of thought that using software corrections counts as 'bad' design, and that all lenses should be fully optically-corrected for distortion in the traditional fashion. We simply don't agree. This approach was essential when shooting film simply because there was no other option; there's no sensible way of correcting distortion on slide film or when printing a negative. Likewise, SLRs ideally need fully-corrected lenses so that the viewfinder image allows accurate compostion (distortion correction inevitably discards parts of the recorded image).
However, with digital capture software corrections are easy to apply and work very effectively. They allow manufacturers more freedom in designing lenses, not just in terms of optics but also physical size. Wideangles in particular can be made smaller and lighter, as the front element can be made smaller, for example. And because with mirrorless cameras you never view directly though the lens, the (electronic) viewfinder image can be corrected 'on-the-fly' to allow accurate composition (and invariably is). Obviously when using the optical finder, it simply doesn't matter how the lens is corrected.
Because of this, it makes perfect sense for lens designers to leave a little distortion behind to be corrected in software, and concentrate on minimizing other aberrations instead. Likewise, if any residual chromatic aberration can be cleaned-up in processing, then why not do so? Surely it's the final image that counts, not how how you get there.