Fuji’s X-Trans sensor sucks? (Part 2 – Details smearing, Noise and Lightroom)

“X-Trans sucks!  It always smears foliage!!! It’s so Noisy! Especially in Lightroom.  Jpeg is waaay better”*

*Trollz gonna trollz

So, let’s see if the trolls are right.

XtransSucksPart2Title

(Note: I know not everyone who has a problem with Fuji is a troll. I’m just having a bit of fun).

 

COMPARING TO OTHER CAMERAS

(X-Trans vs Bayer)

Lightroom settings:

Using sharpening set for Bayer (aka Lightroom default).

Settings for bayer (all set the same)

Settings for bayer (all set the same)

(Open in new tab for full size)

Open in new tab to see full size

Open in new tab to see full size

  1. 5Dsr – Plenty of details. Very sharp. Fine details evident. (highest resolution)
  2. D750 – Slightly soft, some smearing. No paint effects.
  3. A7rII – Sharp. Details evident. No smearing. (best per pixel performance)
  4. X-Pro2 – Sharp. Minor smearing in fine details. Details evident
  5. A6300 – Sharp. No real smearing. Details evident
  6. Pentax K-1 – Slightly soft. Smearing evident.  Some loss of details (worst one)

Funny enough the Pentax K-1 was the worst, but I’d say it would be one of the best if you used pixel shifting on it.

I liked the A7rII for foliage, seemed to be the best overall.

Xpro2 wasn’t too bad.  Certainly didn’t see any paint effects.

A6300 was better than the Xpro2 (by a small amount)

None stood out as terrible, or having paint effects or worms or whatever.

So, how about setting it up for X-Trans then?  Let’s compare the a6300 to the Xpro2 (since they are similar resolution and some say even the same sensor but with a different array).

So I added my favourite settings for high detail (Xpro2) into lightroom.  Let’s compare them to the bayer settings and the Sony a6300 with the bayer settings applied.

Lightroom X-Trans Settings

My settings for highly detailed LR sharpening.

My settings for highly detailed LR sharpening.

Comparison

bayervsxtrans settings

 

It’s actually really hard to see the subtle differences, but the new settings reduce the minor smearing and bring out the details.  I think the Xpro2 matches the A6300 or even betters it now.

Goes to prove good processing is the key.

 

“I have an X-T1 and it smears like crazy!!”*

*Troll with an older sensor.

So, I suspect at this moment perhaps the newer sensor might have something to do with the apparent lack of paint effects or something.  So it’s time to compare the older X-T1 to the Xpro2 (which will have the same performance as the X-T2).

Ah I see it now.

Ah I see it now. Xpro2 left, XT-1 right.

Oh wow.  I see it now.  Let’s see if I can get the X-T1 to catch worms and get a nice coat of paint.

 

X-T1 Pushed hard.

Usually when you push your sharpening hard (using a bayer filter) you do it with settings something like this:

Traditonal Push

Getting you results something like this:

XT1 pushed hard tranditionallya

Note the artifacts.

It’s a pretty well known fact that the X-Trans sensor can take pushing the Details slider harder than usual.  Also you can push the radius up a bit more too.  Here’s the same as above with more X-Trans centric settings.

XtransPush

 

XT1 pushed hard X-TransA

This setting seems to maintain details while not introducing weird artifacts.

Don’t push the sharpening slider, push details.

 

 

NOISE

(Is the X-Trans a noise machine?)

“X-Trans is so noisy, everyone else is waaaay better!!!”*

*Troll who loves everyone else?

 

Noise.  Let’s see how the X-Trans sensor holds up.  I’m going to compare the X-Pro2 to a few of it’s rivals.

Since the X-Pro2 has a isoless sensor, I can compare 12800ISO and it’ll be like pushing base ISO to insane levels.  So if you want to see how recovered blacks might look, this is the same level of noise.

Again, I didn’t just trust DPreview, I downloaded them all and did identical processing in Lightroom.  Which in this case meant turning OFF sharpening and NR.

Oh and yes I did counter for Fuji’s so called “cheating” with noise by pushing it approx 0.5 stops (like pushing the ISO higher).  Which is the max difference reputable testing sites have reported.

Per Pixel Noise Comparison

(NR and sharpening turned off)

For reference here’s a sample of 400asa film (that I found on the internet):

400asa film

400asa film

Notice with film the grain is not 100% even.  This is due to the fact that the crystals etc. in film isn’t even.  So more “film like” grain isn’t 100% even.

Don’t fear the grain.

 

 

 

 

 

They're pretty close.

They’re pretty close.

Grain.  It’s a SUBJECTIVE thing.  Also people’s noise tolerance differs.  I tried ranking them according to MY preferences for smooth yet film like grain (random patterns) without smearing/loss of details. Just because the grain appears smoother, if this is a result of softness it will lose marks in my mind, sharpening a soft image will bring out grain, so softness isn’t an advantage. I also looked at the amount of colour noise.

Here’s my ranking (I tried to do this comparison without knowing what each one is, so it’s as honest as I can be):

Noise Ranking

(remember this is really subjective, if you disagree that’s OK)

  1. D500 (smooth grain, not too much luminance or colour noise.  Fair detail in the grain).
  2. D750 (similar to D500, only let down by softness)
  3. A7rII (Has the details in there, but grain is a bit clumpy. Has FILM like grain)
  4. D810 (Not too bad but a bit more clumpy than the A7rII.  Has FILM like grain)
  5. 6D (grain a bit more clumpy than D750, bit more softness than A7rII, grain is fine and smooth.  However it is ranked low because it is one of the softest)
  6. 7D II (similar to the Xpro2 below it but it isn’t as soft, has more color noise than the Xpro2. Grain is slightly finer than the Xpro2. Has FILM like grain.)
  7. Xpro2 (Less colour noise than some that are ranked above it, has slight softness thank brings it down the ranks. Has FILM like grain)
  8. A6300 (Very close to Xpro2 and 7DII but is very soft)
  9. 5Dsr (Largest MP, but the pixel pitch should be similar to Xpro2 and A6300.  Noise that is  quite apparent, and copious colour noise lets it down.  It does hold it’s details though, slightly better than the K-1 because the K-1 is a bit soft.  Has quite FILM like grain)
  10. K-1 (Quite uneven grain, softness brings its rating down, bit too much colour noise)

I feel the A7rII, D810, Xpro2, 7DII and even the 5Dsr have the most film like grain (in my opinion that is).  This does NOT mean the grain is smooth, but it has the randomness of film.

Others are smoother, but that’s because of softness in the case of the 6D, A6300, D750 and K-1.

To me the D500 has the best balance of nice fine grain, lack of softness and only minimal color noise.

ALL OF THEM HAVE ACCEPTABLE NOISE FOR 12800ISO.

The modern X-Trans sensor (X-Pro 2 and X-T2) is right in the middle according to my ratings.  That’s fine by me.  Doesn’t mean the X-Trans is bad or good, just it is what it is.

I would be happy with any of these when it comes to high ISO performance.

 

 

REAL TREES

(Who needs a test chart?!)

“RAW is stupid, everyone knows Fuji Jpegs are better!!!!!!!!”*

*Troll who also loves Ken Rockwell.

 

I have found since getting the X-Pro2 that Lightroom matches Fuji colour VERY well now.  However, it’s time to see if Lightroom are as good as Fuji jpegs.  Fuji Jpegs are excellent and probably the best you can get.

For this test I’ve used Pro Negative Standard and I shot RAW+Jpeg in camera.  Here’s the comparison.

(Shots taken with 90mm f/2 lens.  F/8. ISO200. On a tripod)

LRvsJpeg_StandardImport

Raw Left. Jpeg Right.

Opps, seems Lr doesn’t match the Pro Neg Standard profile very well… until…

RAW with -85 contrast

RAW with -85 contrast

Ahh that’s pretty close.  LR seems to add contrast to the Pro Neg Standard profile.  Taking away -85 on the contrast gets us pretty close.  You can always adjust contrast to tastes so this is OK.

Now let’s see some 100% details.

STANDARD ADOBE SHARPENING

(Open pictures in new tab to see full size)

 

Standard Adobe Settings

Standard Adobe Settings

 

With Standard Sharpening -85 contrast (to match jpeg)

With Standard Sharpening -85 contrast (to match jpeg)

 

I know the Pro Neg Standard Jpeg isn’t supposed to be very sharp, so just take this as you will, but to me it looks like the Adobe processing is just as good as the Fuji Jpeg, I don’t see any paint effects or worms and it’s slightly sharper.  So far so good.

 

So how about pushing it a bit more.

“Super Sharp Cleaner setting:”

LRvsJpeg_Strong_smooth_sharp

Super Sharp Smooth setting

Super Sharp Smooth setting

 

Still looks ok to me.  Slightly more details.  No worms or paint effect to my eyes.

 

How about the same thing but with distant trees?

Here’s a picture taken with some distant trees (if the paint effect is going to show up, here’s the pic):

Prime location for smeared foliage

Prime location for smeared foliage

 

All with standard LR settings.

(open in new tab to see full size)

DistantTrees_Standard1

DistantTrees_Standard2  DistantTrees_Standard3

 

The only place maybe there’s any paint effects is in the last sample. It’s not too bad.

 

So what are people doing to make this happen?

 

I suspect overdoing the noise reduction.  So let’s add some sharpening and NR and see what happens.

Noise reduction set to 37

Noise reduction set to 37

 

AHHHHH there it is. I didn’t even push NR too much.  Be very careful with Noise Reduction if you’re using Lightroom, and DON’T FEAR GRAIN, as I said before.

 

How about those dreaded worms?

So let me just turn off NR and add 100 to the Sharpening slider:

Worms.

Worms.

Don’t push the damned sharpening slider too much and watch your NR.

 

Conclusion for part 2

I didn’t see much evidence that X-Trans lacks details (see part one) or smears foliage more than Bayer sensors.  Foliage was also fine even when processed with Lightroom. While the X-Pro2 (X-Trans) sensor doesn’t have the least noise, it’s certainly not bad, lacks color noise and has a very film like grain.

I DID find evidence of the paint effect and worms in Lightroom!  Though, this is likely from pushing the files too hard (with the X-Pro2 and an up to date Lightroom at least). I’ve seen it claimed that Lightroom has an “inability to process X-Trans files correctly”.  I don’t believe this to be true.  I believe it has the ability to be easily pushed too far however. I do wish that the X-Pro2 has less noise at low ISO levels.  Because people push the NR hard to try and eliminate it, this causes the famous paint effect. Also, people need to remember not to push the sharpening slider too hard, this will introduce the dreaded WORMS, but instead push the details slider.  Even up to 100.

I’m not saying people don’t have issues with it (especially in the past) I’m saying be careful if you’re using Lightroom.  It could be that Lightroom’s sharpening and NR sliders are quite powerful and easy to abuse without knowing it.  It could be Lightroom has improved a lot (I believe this is very true).  Certainly Lightroom matches Fuji COLOURS really well now (it didn’t when I used the xpro1 in the past), which is a major bonus.

Still, I do wonder if other RAW processing apps can push the files harder without getting the worms/paint effect.  I will explore that in the next part.

 

Recommended reading:

SHARPENING X-TRANS FILES IN ADOBE LIGHTROOM

 

Thanks for reading,

signature

Part 3 Will compare Lightroom to other RAW conversion programs.  Let’s see if LR can cut it.

 

 

LINK FOR RAW FILES (used at the end):

https://drive.google.com/folderview?id=0BzRrvTehHxr9Tjlzd3VCSnlXc0E&usp=sharing

 

  • Sabin Kolarov

    “Since the X-Pro2 has a sensorless sensor” you mean isoless?

    • Ha ha. Did I say that? Yes I meant isoless. Will edit.

  • Akira Yamaguchi

    Thanks for the analysis. It’s very useful.

  • Lee Smith

    The reality is that when printed none of the cameras you reviewed look bad. It’s sad when you feel the need to defend a billion dollar company like Fuji just because some people complain about how their images look at 100% on a computer monitor. The reason I sold my Fuji X100T was not due to the image quality, but due to the stupid ergonomics….that is what really makes a camera good or bad.

    • Thanks for commenting. I’m sorry you feel sad.

    • GCHQ

      X100T is a beautiful camera

  • GCHQ

    What about this article ? He says that a jpeg processed in the camera has a plastic look and the characteristic worm look. What is he doing wrong ? or Who is lying ?

    http://joelwolfson.com/on1-photo-raw-2017-2-now-great-fuji-raw-processor/

    • No one is wrong I agree with him. I’ll point you to my conclusion particularly the part here:

      “I DID find evidence of the paint effect and worms in Lightroom! Though, this is likely from pushing the files too hard (with the X-Pro2 and an up to date Lightroom at least). I’ve seen it claimed that Lightroom has an “inability to process X-Trans files correctly”. I don’t believe this to be true. I believe it has the ability to be easily pushed too far however. I do wish that the X-Pro2 has less noise at low ISO levels. Because people push the NR hard to try and eliminate it, this causes the famous paint effect. Also, people need to remember not to push the sharpening slider too hard, this will introduce the dreaded WORMS, but instead push the details slider. Even up to 100.”

      You can read “paint effect” as “plastic effect”. You can get it from pushing processing hard. Using Irdient X Transformer as a plugin for landscapes (or if you see this effect) will help a lot.

      Just an FYI, people can see if you upvote your own comment. 🙂

    • BTW I’m not paid by anyone to do my tests. I don’t advertise or do affiliate links. I like to remain as unbiased as possible. 🙂

    • Dee Mac

      Or who is lying? Do your own tests, cock sucker.

  • Klaus Bo

    Great work Adrian, but I am sorry to say that there is still no software out there, that can process X-Trans RAF files from X-T2 correctly. They are still working on it. And comparing my 4 years old Leica M240 with the X-T2 at i.e. 1600 iso falls out in favour of the M240.

    X-T1 and X-T2 might have the same sensor type on paper. But when it comes to real life and the way the 24 megapixel X-Trans sensor works vs. the 16 megapixel, there is big differences. X-T1 definitely looks way better on 1600 iso that the X-T2. The best way to work around it is to expose your high iso shots at +1/3 to 1/2 stop. Let me know if you want examples.

  • Thanks for this article. I just bought a Fuji X100F, and I was getting worried about this supposed issue. But…something just didn’t add up. Why would an alternative CFA pattern cause these kinds of problems? It just didn’t make sense. Thanks to this series, it all makes sense. *Of course* it’s just trolls who don’t know how to make use of RAW files. What else could it be?

    It’s weird, because I keep see articles like this every few years that say explicitly, “Dudes. Cut it out with the sharpening and noise reduction.” But no matter how many time people like you keep saying it, the trolls will keep doing it. I suppose the average person goes, “Sharpness? I like sharpness! _Crank it up!_” Or they think, “Noise? God, I hate noise! Noise reduction, go!”