X-Trans 2 vs 3 vs 4 in raw

I noticed recently I currently have X-Trans 2 (X-T1), 3 (X-H1) and 4 (X_T3) series cameras in my inventory and on a recent vacation shot all of them in raw. Looking through raw images in both LR and C1, I have to say that while I like using the newer cameras due to improved functionality, such as AF, my subjective assessment is that I find the raws most pleasing in the order 2 then 3 then 4, which is frustrating as the order is reversed in terms of my shooting enjoyment.

I know raws in theory can be pushed and pulled to look like whatever we want, but certainly in terms of colors I find I have to work harder to realise my vision with v4 of X-Trans than previous incarnations. It's certainly not, in my experience, easy to batch process in the same way a bunch of shots from say X-Trans 2 and 4 cameras from the same shoot using the same raw recipe, even if lighting was the same. Now I've seen some wonderful images from  X-Trans 4 cameras so I'm not saying it can't be done, it's just hit me how the images look different to me because I have been shooting simultaneously with three generations of X-Trans cameras. To me X-Trans 2 especially hit the mark in terms of having a kind of filmic nostalgic look even to the raw starting point in LR or C1, whereas for me X-Trans v4 in particular sometimes seems bland when first opened up in a raw editor by comparison to v2.

Would be interested to hear your thoughts - divergent or otherwise. I am sure that these sensors on paper have become better in terms of DR, high ISO, resolution etc, back side illuminated, on sensor phase detect pixels etc etc, but there's something I like so much about X-Trans v2 (and v1) that I am actually thinking about re-buying an X-E2 and even an X-E1 for X-Trans 1.  I don't think it's just nostalgia as I can easily look through my LR catalog and see that to my eyes I really think v1 and v2 had something that set Fuji images, even in raw, apart. Even when first opening up a raw I could say, oh yeah, that's a Fuji raw right there (I also shoot Sony).

Would be interested to know if you have tweaked your raw recipes and approach, those of you, who, like me, have  been with Fuji long term and moved through the X-Trans itertations.

I make customized profiles for each of my cameras (X-Trans III and IV) in ACR/Lightroom so that the color is the way I like it right from the get-go at import, and that the color always closely matches between cameras without me having to fiddle about every time.

Erik Baumgartner wrote:

I make customized profiles for each of my cameras (X-Trans III and IV) in ACR/Lightroom so that the color is the way I like it right from the get-go at import, and that the color always closely matches between cameras without me having to fiddle about every time.

I do have a datacolor target - is that the sort of thing you use?

I really wanted to find the difference for myself (RAW + SOOC).

A few months ago, I stumbled on a deal - I could buy a X-E1 for a “fair” price. I was then able to do a head to head comparison (same lens etc) on the subjects that were important to me.

I posted a few sample pics in this thread.

https://www.dpreview.com/forums/thread/4633121

Personally, there is a difference but it’s fairly subtle. I much prefer the AF speed improvements from newer models, but i do appreciate that shooting experience with the X-E1 as it’s “basic” (menus/AF-S only) once in a while.

Opportunity for you to now experiment with your gear - would be a bonus if you could share your findings (images)

eMarco Cinnirella wrote:

Erik Baumgartner wrote:

I make customized profiles for each of my cameras (X-Trans III and IV) in ACR/Lightroom so that the color is the way I like it right from the get-go at import, and that the color always closely matches between cameras without me having to fiddle about every time.

I do have a datacolor target - is that the sort of thing you use?

I have an x-rite color checker, but just downloading the appropriate test scene shot RAW files for your different cameras from DPReview or IR and tweaking the available color profiles (Fuji sims or otherwise) so they match and saving them as new custom profiles to use going forward will get well within the ballpark (you need to do this in ACR, but the new profiles will show up automatically in LR as well).

-- hide signature --

https://www.flickr.com/photos/marcoc/
"When words become unclear, I shall focus with photographs. When images become inadequate, I shall be content with silence." Ansel Adams.

Marco Cinnirella wrote:

I noticed recently I currently have X-Trans 2 (X-T1), 3 (X-H1) and 4 (X_T3) series cameras in my inventory and on a recent vacation shot all of them in raw. Looking through raw images in both LR and C1, I have to say that while I like using the newer cameras due to improved functionality, such as AF, my subjective assessment is that I find the raws most pleasing in the order 2 then 3 then 4, which is frustrating as the order is reversed in terms of my shooting enjoyment.

I have an X-Trans, X-Trans3 and X-Trans4 and I agree with you.

The original X-Trans is my favorite and I'm not sure how different the X-Trans2 files are - I know Fuji added PDAF but I'm not sure how much they changed the original RAW/color formula other than this.

I know raws in theory can be pushed and pulled to look like whatever we want, but certainly in terms of colors I find I have to work harder to realise my vision with v4 of X-Trans than previous incarnations. It's certainly not, in my experience, easy to batch process in the same way a bunch of shots from say X-Trans 2 and 4 cameras from the same shoot using the same raw recipe, even if lighting was the same.

RAW files can, in theory, be pushed but there is an inherent character to them, and it's my opinion that the relatively few calibration points offered by a color checker or similar don't QUITE make up for the character of the sensor such that you can simply calibrate to a color checker and expect the same results.

There are also the shadow / highlight characteristics of each that can't quite be simulated.

Also if you prefer the camera colors - why would you want to use a color checker to reduce them?

Similarly, Adobe attempts to reduce the differences between cameras and likely uses way more datapoints than we do, but does not truly eliminate differences between sensors.

Now I've seen some wonderful images from X-Trans 4 cameras so I'm not saying it can't be done, it's just hit me how the images look different to me because I have been shooting simultaneously with three generations of X-Trans cameras. To me X-Trans 2 especially hit the mark in terms of having a kind of filmic nostalgic look even to the raw starting point in LR or C1, whereas for me X-Trans v4 in particular sometimes seems bland when first opened up in a raw editor by comparison to v2.

Again I agree.

The older X-Trans sensors seemed to have a different goal from the newer ones.

As far as colors - the older sensors seemed to aim for more of a balance between warm and cool. Newer sensors seemed more warm and the newer film sims cooled off that warmth.

The "same" film sim on newer cameras is warmer and more saturated than older cameras with the "same" film sim. And each new introduced film sim is cooler and less saturated. It's like Fuji know that each new generation sensor / film sim warms up the photos, so they purposefully introduce new film sims that desaturate and cool down the colors.

And each new film sim is celebrated as looking good - but there's a difference between a well balanced photo and a photo that was first overly saturated and then desaturated. IMO.

And newer cameras seem to have a different demosiacing and sharpening algorithm that makes things seem less filmic to my eyes. I don't know if it's just the jump from 16 to 24, but newer cameras have a different "dimensional" look to them that I attribute to demosaicing and then sharpneing.

Would be interested to hear your thoughts - divergent or otherwise. I am sure that these sensors on paper have become better in terms of DR, high ISO, resolution etc, back side illuminated, on sensor phase detect pixels etc etc, but there's something I like so much about X-Trans v2 (and v1) that I am actually thinking about re-buying an X-E2 and even an X-E1 for X-Trans 1.

There's more to cameras than the specs. Or rather, there are specs we don't bother to measure.

For example - if you weaken the color filter array, you get may better high ISO performance (less noise at high ISOs). This is because more photons reach the sensor / a stronger CFA by definition reduces the number of photons that reach the sensor.

So you can optimize for better color separation or better ISO performance (at the CFA) but not necessarily both. Though Phase1 has shown that a stronger CFA can result in lower high ISO noise (at least on the Chroma scale).

This is of course a vast over simplification of the actual tech involved.

I don't think it's just nostalgia as I can easily look through my LR catalog and see that to my eyes I really think v1 and v2 had something that set Fuji images, even in raw, apart. Even when first opening up a raw I could say, oh yeah, that's a Fuji raw right there (I also shoot Sony).

Would be interested to know if you have tweaked your raw recipes and approach, those of you, who, like me, have been with Fuji long term and moved through the X-Trans itertations.

Personally - I'm looking to buy another X-Trans1 sensor camera and maybe an X-Trans2 (PDAF for faster autofocus) camera to see if it's got better AF performance and at least comparable colors.

PS - check the Camera JPG Portrait Shootout link in my sig - I've compared numerous cameras and the original X-Trans wins hands down for skin tones, at least in SOOC camera JPGs. Both in studio lighting and natural lighting tests.

PPS - I still shoot with all of the above cameras and for turning over video or finished files to clients where autofocus is a factor - the newer Fuji cameras win out. But for my personal shooting - it's the original X-Trans all the way.

FujiShooterCY wrote:

saltydogstudios wrote:

RAW files can, in theory, be pushed but there is an inherent character to them, and it's my opinion that the relatively few calibration points offered by a color checker or similar don't QUITE make up for the character of the sensor such that you can simply calibrate to a color checker and expect the same results.

As I mentioned in some earlier comment, you need an IT 8.7/1, IT 8.7/2 (ISO 12641) target to get a trusted result, and a profile-cooking software, too.

Now THAT looks like a decent color profiling piece of kit.

There are also the shadow / highlight characteristics of each that can't quite be simulated.

Also if you prefer the camera colors - why would you want to use a color checker to reduce them?

Excellent point, in fact.

The older X-Trans sensors seemed to have a different goal from the newer ones.

As far as colors - the older sensors seemed to aim for more of a balance between warm and cool. Newer sensors seemed more warm

I disagree. The sensor itself (and the RAF file, consequently) are color agnostic.

RAF file don't contain any mention of any "image" at all, until developed.

So what you are saying is neither about the sensor nor about the RAF file. You are speaking about the default settings of particular RAW development software.

I agree that RAW files don't contain an image per-se and that the RAW converter is what produces an RGB image. You could take it one step further and say that color doesn't exist in nature - frequencies of what we humans call visible light - exists in nature. Color is created in the brain.

To that end - a camera sensor's job is to capture and detect different frequencies of light in a way that can be turned into an RGB image that the human brain interprets as a color image. (Phew that was pedantic).

The way most cameras detect different frequencies of light is through the color filter array (CFA).

Camera makers specify the dyes in the CFA and those dyes have different characteristics.

For example Nikon cameras pick up  more "Red" in frequency ranges that also trigger the blue sensels than Canon camears. With each sensel getting a different input from any given frequency of light, translating those into full color images means ... some complex math that's beyond my understanding, but to do it in a way that produces either consistent results (Adobe Color attempts to reduce the difference between say Nikon and Canon) or results in line with the camera manufacturer's in-camera JPG engine, requires profiling the characteristics of each camera (and the in-camera JPG output).

Which is to say, the camera manufacturer made decisions about the CFA that influence the final RGB image. In which case I would argue that this is evidence that the RAW file is not "color agnostic" - though fancy profiling techniques can attempt to reduce these (purposefully introduced) differences.

https://www.mdpi.com/2313-433X/6/8/79/htm

And newer cameras seem to have a different demosiacing

Actually we have no clue about which exact algorithm and and at what exact settings does Fuji use in their in-camera firmware.

and sharpening algorithm

Fuji doesn't sharpen RAFs (unlike SONY).

Good to know. I meant the sharpening that happens after demosaicing in the RAW converter (in camera, Fuji software, Adobe, Darktable, Raw Therapee etc.). To my eye X-Trans files have a different look from Bayer files that is akin to surface blur which I attribute to the sharpening algorithms that are used on RAF files.

PS - check the Camera JPG Portrait Shootout link in my sig - I've compared numerous cameras and the original X-Trans wins hands down for skin tones, at least in SOOC camera JPGs. Both in studio lighting and natural lighting tests.

Wow. This is quite interesting observation. Thanks for pointing this out. Maybe someday I'll chase some X-Trans 1 camera for the price of a beer and look at it deeper.

I sometimes eye a used X-Pro1 but they're still around $600 USD and can't justify the cost.

PPS - I still shoot with all of the above cameras and for turning over video or finished files to clients where autofocus is a factor - the newer Fuji cameras win out. But for my personal shooting - it's the original X-Trans all the way.

FujiShooterCY wrote:

saltydogstudios wrote:

FujiShooterCY wrote:

saltydogstudios wrote:

RAW files can, in theory, be pushed but there is an inherent character to them, and it's my opinion that the relatively few calibration points offered by a color checker or similar don't QUITE make up for the character of the sensor such that you can simply calibrate to a color checker and expect the same results.

As I mentioned in some earlier comment, you need an IT 8.7/1, IT 8.7/2 (ISO 12641) target to get a trusted result, and a profile-cooking software, too.

Now THAT looks like a decent color profiling piece of kit.

This tech is known for many years though it's too seriously professional to be widely known, not to mention widely adopted. But it is here, at your arm distance, to your pleasure.

So what you are saying is neither about the sensor nor about the RAF file. You are speaking about the default settings of particular RAW development software.

I agree that RAW files don't contain an image per-se and that the RAW converter is what produces an RGB image. You could take it one step further and say that color doesn't exist in nature - frequencies of what we humans call visible light - exists in nature. Color is created in the brain.

It seems like you are inviting us to dive into a pretty deep rabbit hole now...

I'm continuing the conversation with you.... So... Not sure who invited whom?

To that end - a camera sensor's job is to capture and detect different frequencies of light in a way that can be turned into an RGB image that the human brain interprets as a color image. (Phew that was pedantic).

The way most cameras detect different frequencies of light is through the color filter array (CFA).

Camera makers specify the dyes in the CFA and those dyes have different characteristics.

For example Nikon cameras pick up more "Red" in frequency ranges that also trigger the blue sensels than Canon camears. With each sensel getting a different input from any given frequency of light, translating those into full color images means ... some complex math that's beyond my understanding, but to do it in a way that produces either consistent results (Adobe Color attempts to reduce the difference between say Nikon and Canon) or results in line with the camera manufacturer's in-camera JPG engine, requires profiling the characteristics of each camera (and the in-camera JPG output).

Which is to say, the camera manufacturer made decisions about the CFA that influence the final RGB image. In which case I would argue that this is evidence that the RAW file is not "color agnostic" - though fancy profiling techniques can attempt to reduce these (purposefully introduced) differences.

Yes, this is perfectly correct but I think it's a tad deeper dive inside this rabbit hole than OP was intended to do initially

My answer will be short. If you want to dive THAT deep, use linear input profile and linear tone function (not a curve but a straight line) and apply these to all your RAW breeds, than (given identical demosaicing) you will compare RAWs directly.

But I doubt it would be interesting to someone who just want to make a visual comparison with his own eyes, because the resulting images will be too bland and dull to look at them, so no meaningful subjective, qualitative (not measurable and quantitative) comparison could be done on these.

Fuji doesn't sharpen RAFs (unlike SONY).

Good to know. I meant the sharpening that happens after demosaicing in the RAW converter (in camera, Fuji software, Adobe, Darktable, Raw Therapee etc.).

At least for darktable and RawTherapee I am pretty sure they won't apply any sharpening if you don't want it to be applied. Same with denoising.

Yes correct, I was referring more to Fuji and Adobe than Darktable or Raw Therapee.

To my eye X-Trans files have a different look from Bayer files that is akin to surface blur which I attribute to the sharpening algorithms that are used on RAF files.

I can't parse what are you speaking about; what "sharpening after demosaic" in darktable did you ever observe, given you didn't used it explicitly with your own hands?

https://medium.com/ice-cream-geometry/x-trans-vs-foveon-a-mostly-monochrome-photowalk-1a931f8fb277

I sometimes eye a used X-Pro1 but they're still around $600 USD and can't justify the cost.

I'll wait until someone will give me X-T1 or X-T10 just for a few beers or a dinner with wine