Is Fuji X-Trans really better ? 30 April 2020

X100V

Since the early days of digital photography most cameras have used a Bayer filter array over the imaging sensor to allow the device to record color information.

But Fujifilm has tried various alternative approaches over the years, presumably in an attempt to gain some benefit over the standard Bayer filter array.

There was Super CCD then EXR Super CCD then EXR CMOS and from 2012  X-Trans in evolving iterations. I note that X-Trans sensors do not have an AA filter.

Imaging Resource recently posted a piece on Fuji sensor development by Dave Etchells:

I have read numerous claims by Fuji’s public relations people  and by enthusiast individuals, claiming a range of benefits for X-Trans over Bayer assuming the same 28mm diagonal APSC sensor size and the same or similar pixel count.

In no particular order these have included:

* Reduced likelihood of moiree

* Better sharpness

* Higher resolution

* Better color fidelity

* Lower color noise

* More pleasing film-like grain (whatever that means) at high sensitivity (ISO) settings

A related issue in the X-Trans debate is the ability or inability  of Adobe Camera Raw [ACR] and/or Lightroom to convert the X-Trans RAW files in optimal fashion without artefacts.

I first encountered Fuji X-Trans in the form of the X-T1 in 2014.

I found that when processed in Adobe Camera Raw the X-T1 files had some unwelcome characteristics including loss of detail in foliage and grass and strange artefacts in some kinds of foliage, often referred to in discussion forums as “worms”.

It was suggested to me that a different Raw converter might alleviate some of these problems but I found it much easier to avoid Fuji X-Trans.

So I had no dealings with Fuji or X-Trans until this year.

The X100V arrived on the market with great fanfare. Several reviewers gave it high praise.

Some said they had fallen in love with it. Really.  For instance

This author presents a romantic view, proclaiming the X100V to have  “character” and a “soul”.  

Most said that problems with the previous model had been rectified.

One proclaimed it to be the best prime lens compact ever made or something like that.

So I bought one when it became available in Australia and have been using it since although opportunities for social and documentary style photography have been severely limited in the Covid19 lockdown.

As it happens I also have on hand a Nikon Z50 which uses the same size 28mm APSC sensor, with 20Mpx, no AA filter and a standard Bayer color filter array.

So I have been able to compare the two APSC cameras with respect to their imaging characteristics.

If X-Trans really is better the X100V should demonstrate a clear advantage whether it has a soul or not. It uses the latest version of the 26Mpx X-Trans sensor while the Z50 uses an older sensor with a lower pixel count.

The short take: I have been unable to find any advantage to the X-Trans sensor at all but I have found two disadvantages which are basically a more subtle version of those I saw in 2014.

I continue to use Adobe Camera Raw  (Lightroom uses the same or very closely similar processor) and Photoshop. I have no interest in other Raw converters.

I have read several reports on user forums that in 2020 Adobe Camera Raw now delivers results from Fuji X-Trans which are just as good as those from other converters such as Iridient.

I have also read that invoking the [Enhance Details] function in ACR  solves some of the texture/detail problems seen with X-Trans files.

I have tested this on numerous files of different subject types and found that while [Enhance Details] gives a slightly different rendition to fine subject elements in some cases I did not find any actual improvement in detail or texture rendition using Bayer or X-Trans files.

Going through the issues:

These are related but not identical issues.

My test chart consists of printed pages of classified newspaper advertisements. The subject elements are printed letters giving a fairly high brightness difference across an abrupt transition from figure to ground.

On this test the Z50 and X100V deliver identical resolution with both lenses at f4 and 23mm.

The next test was to photograph a person in a variety of lighting conditions  then look at the rendition of the texture of the skin of the face. This is a more challenging test for the lens/sensor/processor/converter system because it has to resolve more subtle gradations of fine detail.

I experimented with a range of sharpening and texture enhancing strategies for this test but the result always came out the same:

The Z50 is able to render fine skin texture details more clearly than the X100V.

The X100V rendition is smoother but the fine wrinkles are missing.

The difference is present in both Raw and JPG files.

The next test was grass and foliage.

The two cameras are generally equal on this task. The X-Trans sensor in Adobe is now giving much better rendition of grass and foliage than was the case in 2014. So things are improving.

But on some kinds of foliage and somewhat unpredictably the X-Trans files can show “worms”. So this issue has not entirely been eliminated.

Color noise is no longer a significant issue as it has been virtually eliminated in ACR at least up to the kinds of ISO settings which will be used on most photos.

Luminance noise is effectively the same at ISO 6400 when both files have the same settings in ACR and are viewed at the same output size.

Actually I found the Fuji files to be just slightly smoother with slightly less sharpness and slightly less noise. But when I adjusted sharpness to equal that of the Nikon then noise levels looked the same. The character of the noise is slightly different. The grain pattern from the Nikon is small and sharp, that from the Fuji larger and softer. I cannot say one is more appealing than the other.

I do not have a laboratory to test this in scientific fashion. But on numerous tests of subjects having high subject brightness range both cameras were able to deliver the same amount of highlight and shadow detail.

Bill Claff at Photons to photos has not posted results for the X100V yet but the X-T3 is said to use the same sensor and the levels for this are very similar to the Z50.

I have not seen any evidence of more accurate color from either camera.

I have yet to see this in any photo from either camera.

On the basis of  the tests which I have done I cannot find any evidence that the X-Trans sensor in the Fuji X100V delivers any photographically meaningful real world benefit over the standard Bayer sensor in the Nikon Z50.

There are disadvantages though, in the form of reduced fine texture detail and the occasional appearance of worms in foliage.

If X-Trans is really better than Bayer  I think one might reasonably expect  that  all the camera companies would be using it or some version thereof.

So why does Fuji persevere with X-Trans on some of its models (but not all of them, the low cost models and the high end medium format models use a standard Bayer filter array) ?

Apparently sensor cost and processing speed are factors. I have read that sensors for X-Trans are more expensive than standard ones and also that X-Trans requires greater processor capacity.

These factors could explain why Fuji’s budget APS-C models use a Bayer sensor and might be relevant to the use of  Bayer for the medium format models.

Which still does not explain why Fuji uses X-Trans for its premium APSC models. 

Maybe someone at Fujifilm believes X-Trans really is better, or maybe not, in the case of medium format.

Maybe they just want to continue the Fujifilm tradition of offering their customers something a bit different and maybe a bit special.