Monday, March 11, 2013

Nikon D7100 vs Fujifilm X-E1 and X-Pro1

Although the analogy is not completely congruent, the optical low pass filter on a modern imaging sensor is a bit like the brakes on your car. The engine of the camera, the sensor bed itself, is where the performance derives from, but the optical low-pass filter (also known as the anti-aliasing filter) is what slows down the camera to keep you out of trouble (moiré). Removal of the low-pass filter has been a favourite topic of people wanting to squeeze an extra bit of sharpness out of their cameras, but it's not without it's consequences.

The Nikon D7100 does away with it, theoretically giving it sharper-looking images than other 24mp APS-C cameras. On paper, with all else being equal, the images from a Nikon D7100 ought to have better acuity and contrast than those from a D5200, D3200 or Sony NEX-7.  However, in real life situations, I think most people would be hard pressed to tell the difference, unless the images were placed side by side. If you were given a 36mp image without any context and with the EXIF data removed, I think most people would be hard pressed to tell if the image came from a Nikon D800 or D800e (assuming that the moirĂ© had not been triggered). If you place images from both side by side it would then becomes obvious which image came from which camera, but real life viewing situations don't work like that.

There is another (more or less) mainstream camera system out there without the anti-aliasing filter, and that is the Fufjifilm X series. The pseudo-random colour filtration pattern of the X-Trans sensor was designed to reduce false colour detail patterns that can be triggered during de-moasicing of Bayer pattern sensors; because of this, it can do away with the optical low pass filter entirely. For the most part, X-Trans images look more contrasty than conventional sensors and hold onto detail better as the ISO settings increase. The trade-off is that the X-Trans sensor requires more computational horsepower to process its RAW sensor data, which has consequences as we will get into. However, now that the Nikon D7100 has done away with its low-pass filter, one wonders if this is the unique selling proposition that it once was for Fujifilm. It's almost like history repeating itself; Fuji's Super CCD SR sensor in the S3Pro an S5Pro cameras gave the masses unprecedented dynamic range, but the technology was outflanked by the migration of the pro market to full frame sensors and the increasing quantum efficiency of all sensors in general. In other words, it was an effective technical solution for its time, but it did not having an enduring market relevance as the landscape shifted.
If you compare out of camera images, 16mp to 16mp, the Fuji X-E1 gives more pleasingly sharp images than a similar Nikon D7000. Fujifilm claimed that a 16mp X-Trans image is closer in resolving power to a conventional 24mp sensor, but I've maintain that the claim is a bit of a stretch. In the old days, the optical low-pass filter would have robbed roughly 10% of the maximum resolving power of an imaging sensor, but as megapixel counts rose and photodiodes got smaller, the amount of anti-aliasing filtration grew accordingly less. The second factor is that traditionally (going back to classic SuperCCD), Fujifilm sensors tend to give higher numbers of synthetic test charts than Bayer sensors, but it is sometimes debatable whether or not the advantage is "true" information, and whether or not the advantage holds up in real life shooting.

However, with no optical low-pass filter on either the D7100 or the X-E1, then it ought to be that the Nikon will produce the better image quality. Even though the Nikon has smaller photo-sites, we've already seen that the base technology (first seen in the D5200) produces per pixel noise and dynamic range characteristics that are similar to the D7000's 16mp sensor with larger photodiodes. So all else being equal, more megapixels and more spatial samping will win. The X-Trans sensor is no slouch, and the current generation of 24mp Nikon sensors has so far been amazing... so really, this is a benefit to the consumer no matter which route you chose.

Fujifilm X-Trans CFA Pattern
However, it's not completely equal. The pseudo-random pattern of the X-Trans sensor has two important consequences. The first is that the number of red and blue photo-sites is proportionally less than in a Bayer sensor. With a conventional sensor, the colour filter array (CFA) is a rotating block of four colours, two green, one red and one blue. X-Trans uses a rotating block of nine pixels (5 green, 2 red, 2 blue) meaning that the X-E1, X-Pro1 and X100s produce images with less real image detail in the red and blue channels than a similar Bayer sensor. This has varying consequences, depending on the subject matter. The first instance that comes to mind is flower photography, where Bayer sensor cameras struggle to pick up fine low-contrast details in flower petals; theoretically, an X-Trans sensor will do worse, and for the most part, that is my impression.

The second consequence of the X-Trans sensor is how far away the red and blue photo-sites are from each other compared to a conventional Bayer array. This increases the amount of computer power required to interpret the missing colour information in-between the red and blue photosites. This leads to two different types of artifacts; colour smearing and fractal patterns. Colour smearing shows up in fine details where the green photosites can accurately pick out the outline and edges of a subject matter, but where the sparsity of the red and blue photosites means that the colour tends to bleed outside of the edges and boundaries. The effect can be subtle, but it can look like a water colour painting where the paint has been over-applied. The second type of artifacting that can be produced is false fractal patterns. This shows up in areas with repeating (perfectly or otherwise) fine detail where the X-Trans sensor has a difficult time reproducing the structure of the real-life pattern. What happens is that the a weird alien texture can be produced instead.

That said, a bright spot about the Fujifilm X-Trans sensor is that third-party RAW support has been constantly improving since it's introduction, with the latest versions of Adobe Camera Raw greatly reducing the amount of artifacting produced by early RAW converters. It's no longer the deal-breaker that it used to be for serious RAW shooters. The upshot of this is that for most usage, both the D7100 and the Fujifilm X-series cameras will produce high quality images. However, as far as I can tell, most professionals would rather stick with a conventional Bayer sensor because of the question of fractal artifacting. This probably isn't an accident. The type of shooter that likes the punchy default image quality of filterless sensors tends to be more consumer-oriented than professional, and from what I can tell at least, the X-series tends to appeal to enthusiasts rather than working pros. If you watch carefully, most non-professional enthusiasts actually don't pay close attention to fine low-contrast detail and detail within colour channels... the excitement is usually reserved for "razor sharp" looking hard edges and high contrast detail. X-Trans and cameras without a low-pass filter tend to cater to this crowd. There's nothing wrong with crowd pleasing images, but sharpness can always be adjusted later during post processing, whereas artifacts are a bigger nuisance from the get go. Actually, this is history repeating itself. The consumer oriented D70s had a fairly light filter and was capable of inducing false patterns in all of the usual suspects. The professionally oriented D200 that came after it was much more heavily filtered (like the D100 before it), and required sharpening, otherwise the default output looked rather soft.

However, there is more to life than just image quality, and for that, the D7100 is by far a more versatile choice. Faster frames per second, faster, autofocus, better motion tracking, better video implementation, image stabilized lenses,overall lens variety... you get all of this in a conventional dSLR over the X-series. For example, if you want to shoot your kids (photographically) playing soccer at 300mm, then there's no native option for the Fuji. There is the new Fujinon 55-200mm F3.5-4.8 R LM OIS lens, which comes up 100mm short, but if you want to move to f/4 or f/2.8, then it becomes obvious that this type of shooting this isn't the X-series strong suit. As good as the Fujifilm cameras are, they are still dedicated to being compact rangefinder-esque systems, whereas the strength of the SLR format is that it can be many things to many people. If you are shooting mostly still photos within the classic 24-90mm FX equivalent focal length range, the comparison between the two systems is more than fair and reasonable. The majority of casual photography can be described as such, but if you shoot long or at fast moving objects, then the X-Series stops being in contention. So in the end, the deciding factor should be the style of your shooting and not the absolute image quality of the camera. There will always be differences in image quality between brands, but these are small differences compared to all of the different photographic needs and preferences out there.

Lastly, I would be remiss if I didn't point out that 2012 was not a good year for quality control with Nikon. That in itself should not be a deciding factor between an X-E1 or a D7100, since both are quite different cameras, but I would honestly say that I would be hesitant to pick up an early run of the D7100 in light of the way that the D800 left-side focus issue and D600 debris issue were handled. With every new generation there tends to be something... D7000 and D4 mirror oil splatter, D200 highlight banding, etc. etc. It remains to be seen how things will turn out.

Update: March 24, 2013

Having now handled the Nikon D7100, I can't think of an objective reason to pick the X-E1 or X-Pro1 over the D7100 on still image quality alone. There is indeed a noticeable difference in sharpness and contrast with the D7100 over the D5200 and D7000. The X-Trans layout allows Fujifilm to remove the AA filter at 16mp, but the higher pixel density of the D7100 lets it shed the filter at 24mp, but most importantly, with no significant penalty in per pixel noise and dynamic range characteristics... the Fuji cameras might be a bit cleaner once you climb into the higher ISO range, but that starts becoming an academic exercise, since you've already lost a chunk of image quality once you pass ISO 3200. So in other words, the D7100 has the same benefits of the X-Trans, but with higher resolution. I would say that the advantage that Fujifilm enjoyed with it's crisp image output at this price level is gone. Of course, outright image quality isn't the sole consideration in choosing a camera, and for that, the advantages of the X-system still stand.

Advantages Nikon D7100

  • Much more developed lens system
  • Able to save money with used/legacy lenses
  • More versatile camera
  • Conventional Bayer sensor, conventional post processing
  • Removal of low-pass filter puts it on equal better footing with Fuji

Advantages Fufjifilm X-E1 and X-Pro1

  • Great image quality in small body size
  • Does 80-90% of what dSLR shooters do, but in a smaller package 
  • "Fuji" colours
  • "Retro" design, more effort put into aesthetics. More joy of use for some.
  • On-chip PDAF, better focusing during live view
  • A poor man's Leica, better in every practical regard except for the lenses.


  1. Regarding the "weird allien texture" of X-trans sensor mentioned above, I would just like to comment that in my experience this is due to over-sharpening. More importantly, such texture also occurs on Bayer-type sensor when image is over-sharpened. My comment is based on images displayed here: , for Pentax K10D (grass) and Fuji X-E1 (trees).

  2. To my opinion things are clear and simple. Let me be more specific.

    1) Both sensors have the same area, albeit with different pixel density. Assumming not very dissimilar image processing capabilities in both cameras, the Nikon should resolve more detail at lower ISOs (up to 800), and suffer more from noise from ISO 1600 and above. That's natural, as Nikon's pixels are smaller. Nevertheless, the X-trans sensor is charismatic and at the moment of writing, outperforms any APS-C sensor in the market from ISO 1600 and higher.

    2) Anyone who has invested seriously in Nikkor optics and does not mind carrying the weight of the Nikon DSLR system, has no reason to convert to Fuji. Faster NIkkor lenses can in any case keep the D7100 working at ISOs below about 1000, where the sensor performs at its best.

    3) Anyone who wants to travel super light and needs to capture any available natural light without loss of quality, should get the Fuji. XF lenses have great build quality, they are super sharp and internally corrected by the camera, when shooting JPEGS.

  3. I agree that X-Trans gets the benefit of larger pixels, per-pixel they seem to be about 1/2 a stop(?) more than the D7100 between ISO 1600 and 6400, but if you downsize the 24mp Nikon to 16mp, I think the Nikon is better, especially with anything with fine soft detail in the red and blue channels. Or put in another way, a 24mp X-Trans sensor would be at a disadvantage to the D7100, both filterless, but the Fuji's red and blue pixels would be less dense across the sensor. However, I completely agree that Fuji is doing a good job so far of building up the X system from scratch... not a lot of lens offers, but compelling ones. Sort of what you wish Sony could have been with their NEX system.

  4. Just a bit too subjective and Nikon fan boy-ish to be credible.This sort of blathering, apart from being ill-informed, is just plain wrong:

    "The type of shooter that likes the punchy default image quality of filterless sensors tends to be more consumer-oriented than professional, and from what I can tell at least, the X-series tends to appeal to enthusiasts rather than working pros."

    Sorry, mate, but "working pros" like Zack Arias and David Hobby saw the Fuji sensor differently. No doubts about their photography chops. Yours? Not so sure.

  5. Well, that's opinion and its obviously a generalization, as you are entitled to yours. Note in the same paragraph also that I mentioned the same trend with he Nikons before it. It's not a knock on Fuji or their users, it's an observation about their design choices and the marketing implications. Cameras - of all brands - once you get into the serious-enthusiast and pro-level, the majority of buyers are still non-paid professionals, so it's fair to comment on how each appeals to the bulk of the consumers that purchase them.

  6. Users interested on image quality alone, may now view a side-to-side comparison between Nikon D7100 and Fujifilm X-E1 in the acclaimed online reviewer dpreview at

    1. It's actually a very interesting deep dive if you go back and forth between JPEG and RAW. On the whole, there's not much practical difference for the day to day shooter, but in some situations like low contrast detail in the blues (especially the feathers), the Nikon does better at resolving detail, but at higher ISO's (3200ish), the larger pixels in the X-E1 helps preserve the detail better, at least in JPEG. It goes without saying that the Fuji does well with the high contrast detail, especially anything that would be heavy in the green channel and/or black/white.