Fujifilm X-Trans CMOS II sensor
Fujifilm has a long history in designing its own unique sensors which don't use conventional Bayer-pattern colour filter arrays. The X-Trans CMOS is the latest design to emerge from its out-of-the-box thinking, and while it uses a conventional square-grid pixel layout (unlike the company's EXR compacts), the colour filter array over the pixels has been completely redesigned. The result, according to Fujifilm, is minimal susceptibility to colour moiré, which in turn allows the company to dispense with the anti-aliasing filter that's used by almost all other cameras. In principle, this means the X100S should be able to resolve more detail than Bayer-array cameras with a similar pixel count. The 'II' version of the sensor adds another trick; on-chip phase detection autofocus.
The color filter array
Almost all digital cameras use what it called a Bayer color filter array, named after the Kodak engineer who developed the pattern. Over the years it's proved to be an excellent way of capturing both color and detail in a scene. Essentially, it consists of a simple repeating pattern of four pixels, two of which are sensitive to green light, one to red and one to blue, in a square RGGB layout.
However, one problem with the Bayer array is its susceptibility to false colour artefacts when faced with an image that contains finely-repeating patterns (such as textiles), caused by interference between these patterns and the regular grid of photosites. The result is unsightly bands of color, which in most digital cameras is suppressed by the addition of an optical low pass (or 'anti-aliasing') filter in front of the sensor that blurs away the finest image detail. This reduces any moiré patterns, but with an inevitable loss of resolution.
Film never showed an analogous effect due to its random grain structure, and Fujifilm's engineers reasoned that modifying the sensors' colour filter array to make it look more irregular would have a similar effect. The result is the X-Trans CMOS's 6x6 colour filter array, with red, green and blue photosites on each row and column (diagrams courtesy of Fujifilm):
 |
 |
| The common 2x2 'Bayer' pattern used in most digital cameras | The 6x6 color filter array pattern of Fujifilm's X-Trans CMOS sensor |
|---|
Use of an unconventional CFA is not without its complications, though; most obviously, it demands a completely different demosaicing algorithm for RAW conversion. This has proven to be a problem for third-party raw conversion software, as we documented in our review of the X-Pro1. For many users this is offset by Fujifilm's superb out-of-camera JPEGs and flexible in-camera raw conversion, but committed RAW shooters should bear in mind that mainstream support is still limited.
On-chip phase detection autofocus
Almost all compact and mirrorless cameras use contrast-detection autofocus - in its simplest form, this racks the lens through its range of focus distances and picks the one which delivers the highest contrast at the selected AF point as being in-focus. Once upon a time this was painfully slow, but focusing algorithms and lens designs have improved substantially over the past few years, and it can now be extremely fast and accurate. But it struggles in some situations - most notably with moving subjects.
One solution to this is to use phase-detection autofocus. In short, this is able to tell from a single measurement exactly how to adjust the lens to achieve an in-focus image. All SLRs by necessity use this approach, with a dedicated autofocus sensor. But it's becoming increasingly common on mirrorless cameras too, using a system that's integrated onto the main imaging sensor. It's employed most successfully by Nikon's 1 System cameras, but other manufacturers such as Sony and Canon use similar technology, with varying degrees of success.
Fujifilm was in fact the first manufacturer to bring on-chip PDAF to market, with its F300EXR in 2010. But it's appeared curiously reluctant to exploit the technology in higher-end models until now. Both the X100S and the smaller-sensored X20 use the system, and promise much-improved focus speeds over their predecessors (and rivals). As tends to be the case, it appears that the PDAF sensors are localized towards the centre of the frame, and use pixels that are masked to receive light from the left- and right-sides of the lens's exit pupil. The phase difference between the paths allows the focus distance to be determined.
Summary
As we've said elsewhere in this preview, we haven't seen a fully working X100S yet, so can't draw many practical conclusions right now. But on paper it looks like it should offer a cross between the physical design and basic handling of the X100, and the image quality of the X-E1. But that's not all; most notably you also get phase detection AF, a higher resolution electronic viewfinder, improved manual focus, and Fujifilm's latest on-screen interface (including its excellent Q-menu).
Despite its manifest quirks we like the X100 a lot, especially with the latest firmware that irons out most of the bugs; indeed it's one of our favourite cameras of the past few years. We've also been impressed by the image quality from the X-Pro1 and X-E1, with the superb out-of-camera JPEGs going a long way to offset slightly flakey RAW support. So the X100S looks like a mouth-watering prospect indeed, and we're looking forward to getting our hands on a working model to see if it lives up to its undoubted promise.
The CES show gave us a chance to get our hands on a closer-to-production X100S, to allowing use to offer some hands-on first impressions of aspects of the camera's operation we haven't been able to cover in this article: Click here for some more hands-on impressions, including a quick look at autofocus and manual focus behavior. We've also published a technical explanation of how its Digital Split Image focusing works, including a video by Fujifilm illustrating how it appears in practice. Click here to read more about on-chip phase detection and the X100S's Digital Split Image focus aid.
Comments