Is the latest 24MP X-Trans III sensor ‘Good’?
….Nothing beats a proper comparison with a conventional ‘Bayer’ sensor, using a variety of challenging real-world subjects. So that’s that I’ve done, concentrating in this article on how well the X-Trans III records various types of details of a variety of sizes, relative to the sensor’s pixels.
Many of you will have read my Part One reviews on the Fujifilm X100F and the Fujifilm X-T2. Some will perhaps also have seen the lens performance and detail crops in my evaluation of the X100F’s 23mm f2 Fujinon lens. These will have given you a decent idea of final image quality at low ISOs. However, what it didn’t do was give you an idea of how a camera with a Bayer filter on top of its sensor would have fared in comparison.
A Little X-Trans Background
For those familiar with all this, skip ahead to methodology.
The original Fujifilm X-Trans and X-Trans II sensors were both 16MP. Some people complained about ‘the watercolour effect’, or were not happy about the resolution of green foliage or fine detail. While we know that some of these issues came down to the way Adobe software was rendering the RAW files, its important to note that the X-Trans array is fundamentally different to the Bayer array on the vast majority of sensors. The diagrams below make this clear (see how the reds and blues are in less of a checkerboard pattern, the two sizes of green blocks and larger overall proportion of green).
Fujifilm went to some trouble to produce the X-Trans design and the supposed primary benefits are reduced moiré, higher resolution (due to the lack of an Anti-Aliasing/Low Pass filter) and more ‘organic’ looking noise (due to the less uniform distribution of red and blue pixel filters). Whether these benefits are real or not is another matter, but for this article I am looking to poke and prod at the possible Achilles heel of the newest 24MP X-Trans III. I am looking at fine textures and fine green details, because these have been associated with criticisms of the X-Trans sensors of the past.
I will be honest and up front: I was in the camp that felt the X-Trans sensors of the past created at least as many problems as they solved. I could not care less how good Iridium developer is, or Photo Ninja, when I own numerous other cameras and I’m wedded to the day to day workflow provided by Lightroom. That’s of course a very personal perspective, but one I still feel is right for me. There are enough complications in life and work without having to run multiple workflows for different cameras as a matter of routine necessity (rather than occasional luxury).
So, has anything changed?
The fair test would be to see how the Fujifilm X-T2 files compare to those from a 24MP APS-C Bayer camera using comparable lenses. However, this seemed a little boring. After all, the Sony A7 II is comparably priced to the Fujifilm X-T2. While the Sony A7 II isn’t as responsive or fast (and the lenses are generally larger/heavier) it does of course have a larger sensor. It is also a Bayer sensor. This unfair test seemed a bit more relevant to me and perhaps will be of more interest to you. So, I pitted a Fujifilm X-T2 with Fujinon 56mm f1.2 APD against the Sony A7II with Zeiss Batis 85mm f1.8. Ouch.
Now to Discuss Methodology
The Lenses: The Zeiss 85mm f1.8 Batis is a phenomenal performer. It is exceedingly sharp, produces bags of contrast and really does get every drop out of the 24MP A7 II. However, the Fujifilm 56mm is every bit as sharp as the Zeiss optic in the centre of the frame, which is where all the below comparisons take place.
Apertures: I decided to shoot both lenses a few stops down from wide open, so that nobody would feel that one or other was being hobbled. Can we all agree that the Fujifilm 56mm f1.2 is rather good between f4 and f5.6 and that the Zeiss Batis 85mm f1.8 is also ‘on form’ between f4 and f8? I hope so! I have mostly shot the Fujifilm combo with a one stop wider aperture. This is because the smaller sensor and shorter lenses give roughly one stop more depth of field at the same aperture as Full Frame (FF). By using, say, f4 on the Fujifilm X-T2 and f5.6 on the Sony, we are using the sort of settings that a person may use when actually making real photos. Both lenses are absolutely singing here.
Stabilisation: The vast majority of shots were taken using a tripod. Those shot handheld were done without stabilisation on either camera and very high shutter speeds. I can assure you that all frames are as sharp as they get.
Shutter Speeds: In most cases, I tried to roughly match shutter speed on the two cameras. As the Sony was set to a smaller aperture and the Fujfilm uses a different ISO scale, settings might not seem comparable, however, trying to match shutter speeds seemed realistic to me, as this will often be the limiting factor for real world use. If the cameras are not at their base ISO (or either 100-200 on the Sony), it will be stated.
Other Imaging Details: You may also wish to consider the following:
- The Sony has a quite different colour balance in most shots. It is more ‘yellow-green’ and saturated in this part of the spectrum.
- The Zeiss Batis produces higher levels of overall contrast. In this regard, it could be seen as noticeably more ‘aggressive’ compared to the more ‘relaxed’ Fujifilm lens. While the Batis is likely to be used for mixed purposes, I suspect that most Fujifilm 56mm f1.2 (R or APD) owners point theirs at people. I have not adjusted for this in any of the files unless otherwise stated.
- All of the below files are at their default settings, as imported to Lightroom. They were imported as RAWs and exported at the exact same resolution and quality setting as JPEGs. I have only added basic ‘sharpen for screen’ in the Lightroom export dialog box. I will show some sharpened examples later, but thought starting with the defaults was sensible. It isn’t perfect, but it’s a start.
- I did not attempt to perfectly match exposures, as this is made all the more difficult by the different ISO calibration of the two cameras and different contrast levels of the lenses. Overall, the Fujifilm X-T2 tends to expose a bit further to the right.
Viewing: You will no doubt pixel peep and that is the idea. However, if you are viewing on a non-HD/Retina screen, you may find sitting back a bit more useful. 100 PPI will only give you a sense of what 60″/150cm prints might look like…
Example 1: I chose this scene because it contains a mixture of organic (green) detail as well as man-made elements. There’s a variety of textures and detail here, not to mention colours.
Example 1 Summary: I see a slight resolution advantage to the Fujifilm combo here, which is most obvious in the distant building in the top right of the crop, but also in the roof slates and bamboo fence on the left. This is mostly the A7 II’s Anti-Aliasing (or low pass) filter slightly softening the finest details. The fine green leaves take on a very slightly painterly effect on the Fujifilm file when inspected at 1:1, but when you reduce viewing to something more realistic (1:2/50%) this is not visible. You may also note slight moiré and false colour on the bamboo fence on the very left of the Sony file. This is absent with the Fujifilm.
Example 2: Fine green detail such a grass and leaves is most likely to give the Fujifilm X-Trans III sensor a headache if anything can. Let’s see what this medium distance example shows.
Example 2 Summary: The 24 MP Fujifilm X-Trans III sensor has recorded plenty of detail, it seems naturally rendered to my eye and the lack of an AA filter gives more bite to the Fuji file. Where the X-Trans III loses out is in the red/brown stems in the vegetation in the upper right of the frame. Both look a little different, but if we ignore colour balance and contrast differences (at default settings), the differences aren’t huge.
Example 3: What about broader green leaves, such as those in the centre of this frame?
Example 3 Summary: no real change. The X-Trans III sensor is doing a great job, with the same slight softening of the Sony file due to the AA filter visible in the brickwork and stone. The large green leaves seem a little more luscious in the Sony file, but it is hard to tell if this is just down to colour and contrast. I will look at this file again later (with sharpening) so lets move on.
Example 4: Now for fine monochromatic detail up against a blue sky. Here we have typical fine twigs and branches at a middle distance.
Example 4 Summary: In this case, note that I shot the Batis at f4, figuring that is where you get the very best possible resolution in the centre of the frame. I shot with f4 for the Fujifilm as well, to see what happened. The result, to my eye, is that the X-Trans III file is once again a bit sharper (looking at the finest details), but the Bayer file produces the most natural looking and pleasing detail when shown at 1:1. At lesser magnification, there isn’t really much difference in how they look.
Example 5: Fine green tracery that should upset the X-Trans III sensor…
Example 5 Summary: This sort of fine detail is very hard for any sensor to follow it seems! There are certainly differences in how the green details are rendered and overall contrast (which confuses matters), but the X-Trans III has fared better than I expected. The A7 II does resolve the criss-crossing of tiny shoots a little better, but suspect that neither would really be properly resolved in even a very large print. This is something I will investigate.
Example 6: Mmmm, a lovely scene in leafy Cheshire. Sensor testing is boring, but not as boring as not knowing the imaging characteristics of your camera system when presented with priceless, unrepeatable photographic opportunities. Let’s keep going.
Example 6 Summary: The same patterns are repeating themselves here. I’d say this file is where the Bayer does best of all, because this crop is almost entirely about tiny green leaves and stems far away at the edge of the sensors’ resolution. That said, I wonder if the X-Trans would show more ‘apparent’ texture in print, even if its not real? At this distance, both are really struggling to make sense of the finest details, but the A7 II shows the Bayer’s strengths, especially with the red flowers on the bush.
Example 7: Quite possibly one of the most boring shots in the history of photography, but it will show textural detail on the tree nicely.
Example 7 Summary: Its in the very finest textures that the lack of an AA filter gives the Fujifilm a predictable edge over the Sony A7 II. The Fujifilm X-T2 combo seems to pick out a bit more fine detail in the evergreen foliage in the top right, but it is rendered a bit differently. Both results have a somewhat different feel, but its a matter of opinion as to which you prefer.
Example 8: Fine texture, little green leaves and yellow flowers.
Example 8 Summary: The Sony A7 II produces more natural texture because of its ability to resolve finer colour details, most likely.
Example 9: This is another example that should favour the Bayer. Lots of distant fine detail in the middle of the frame.
Example 9 Summary: In my opinion the Fujifilm result looks better in every way. There seems to be more detail being resolved here and fine textures appear better rendered.
From now on I will show crops only.
Example 10: Now for some textural detail, because nothing says ‘photography’ like a brick wall. This one was especially built and weathered for this test, at great expense.
Example 10 Summary: Aside from the dramatic difference in white balance, the Fujifilm’s X-Trans III is recording a bit more refined detail in many resects, but not in others. This seems to be the reality when comparing these two cameras: Either can be stronger and weaker all at the same time. The AA filter induced fuzziness is absent on the X-T2 file. The white paint’s details are better, but the A7 II seems to do better with the red-orange rust and brick, as well as the finest detail in the mortar.
Example 11: What if we sharpen and adjust the above results to try to make them look more similar? This is what I have done. The white balances, saturation and contrast differences are not easily matched, but I did what I could to at least bring them closer together. There are 1000 ways you could skin this cat, but I applied sharpening that I felt produced the maximum visible sharpness without introducing obvious halos or nastiness. This could go back and forth, but with no dog in this fight, this is where I settled.
Example 11 Summary: You be the judge. To be honest, I cannot imagine there’d be a difference in a large print. The X-Trans III produces least diffusion due to the lack of an AA filter, but the A7 II produces the most organic look if you scrunch your eyes up and look very closely. Step back, however and the X-Trans file again seems sharper. I don’t know what to think, which says something.
Example 12: This water tower was some distance away and with both brick texture, straight edges and stone, it makes a good test. Again, both were sharpened and tweaked.
Example 12 Summary: There are pros and cons to each here. Once again, we’re splitting hairs. There seems a bit more sharpness to the X-T2 file, while the cables connecting to the antennae look a touch better on the Sony.
Example 13: More hair splitting, by returning to the leafy green bush.
Example 13 Summary: The Fujifilm X-Trans III has a bit more bite still in most places, but the lighter green veins in the big leaves are better resolved by the A7 II. At this point I am on the verge of narcoleptic coma and I am sure you are too, so let’s wrap this up!
X-Trans III vs. Bayer: The Grand Conclusion
There are differences, yes. However, the differences are small in both directions. The photos are here so that you can form your own conclusions.
What I have personally concluded is that when a top quality 24MP Full-Frame sensor and the 24 MP APS-C X-Trans III are compared using vaguely comparable lenses, the results at low ISOs are very close indeed when using Adobe Lightroom CC.
The AA filter free Fujifilm X-Trans III produces noticeably more bite in the native RAW files, but the scope for greater sharpening of the A7 II’s files brings them closer. The X-Trans III retains a very slight advantage in edge contrast throughout. The end result is typical of what we see when comparing similar Bayer filters, with and without an AA filter. However….
The larger Sony A7II sensor does retain an advantage with fine green details when they are right on the edge of the sensor’s ability to resolve said details and we pixel peep. The same advantage is also true of fine red details. However, fine tracery of branches is not always clear cut. Sometimes the Bayer does better due to the layout of its filter array and sometimes the X-Trans does a touch better due to the lack of an AA filter.
There are of course a number of APS-C cameras out there without an AA filter, which should theoretically remain free of AA filter induced softness, while not suffering any of the negative aspects of the X-Trans filter array.
What this comparison has done is look at one single aspect to X-Trans rendition, but one I know is very important to many of you. In the larger comparison, we must also take note of some of the key claimed advantages of the X-Trans that are above and beyond the simple absence of the AA filter. These are 1) reduced moiré and 2) more organic looking noise at high ISO (not to mention a possible high ISO advantage). I cannot yet comment on the high ISO advantage, but in this short test if was already clear that moiré appeared in the A7 II file while being absent from the X-Trans file. Overall, I would say it is well recognised that the X-Trans generates much less moiré and this may matter to those photographing subjects that include fabrics.
It is my intention to print some of these files at various sizes to see what the results are here. Nobody looks at files online at full resolution and at 100% for pleasure. Online, we take in people’s portfolios in the form of much smaller resolutions where none of these differences would be visible. Therefore, in my view, making actual prints is the most important next step in understanding what ‘real image quality’ differences there are. From these tests, I feel that X-Trans cameras are good to go for critical work using Lightroom and I hope to explore other developers in due course, to see if the X-Trans III’s results get meaningfully better.
The dedicated landscape photographer may feel that a Bayer camera is still the best overall proposition, but lenses, size, weight and cost all impact such decisions. From where I am standing, I do not see any other mirrorless system that offers the lens quality and all round package of the X-T2 or X-Pro2, for example. Sony has several superb APS-C cameras, such as the A6000, A6300 and A6500; however, the lens line up is very different and so is the handling. Where the X-Trans III has weaknesses, it has strengths elsewhere and it’s for you to weigh up what’s right for you.
Would I use the Fujifilm X cameras for important work?
Yes, I would. I have rolled out these tests to methodically illustrate the impressions I formed quite quickly when first using the X100F and X-T2 quite a few weeks ago. However, those initial tests resulted in me purchasing two X-T2s and a X100F from the great team at London Camera Exchange, Chester (using my very own and very real money). It has been a complex decision for me and I will talk through this decision, and what equipment I will be waving goodbye to, in another article.
Suffice to say, I am very impressed by the latest Fujifilm X-Series bodies, but would not dream of trying to convince any of you that it is all a garden of Eden. In some of the equipment that I will let go, I will lose some qualities I like a great deal. However, in overall terms I strongly feel that Fujifilm has provided me with a liberating solution to many of my generalised photographic needs. I want to cut down on equipment, reduce financial liabilities, cut down on weight and maximise utility. Ensuring that I was content with the rendering of X-Trans files in Lightroom was the most critical step of all. It ain’t perfect, but what is?
Plenty more to come….. In addition to prints, I will look at noise/grain and high ISO performance, again using the A7 II for reference.