Friday, March 21, 2014

Took the X-T1 Out All Day Yesterday; What a Pleasure! A Shooter's Camera but I have Mixed Feelings; Why?


I took the X-T1 out all day yesterday to photograph in a variety of situations to continue my "shakedown" of the camera and lenses.  I made about 100 images.  I wanted to primarily test the sensor on finely detailed  objects as well as green foliage.  I had read and seen, on the web, the issues with Adobe's raw conversion algorithm smearing detail, making foliage look as though it had been painted and the "white lines" or " white borders" that would appear on edges.  After looking at my RAW images in the new release candidate 8.4 for Adobe Camera Raw as well as the JPEGS in Lightroom 5, to be honest, I have mixed feelings about what I saw.  Read on.

I first went out and tested the camera's ability to track moving objects. I wanted to really test the X-T1's ability to successfully track fast moving objects and figured an extreme test would be to photograph oncoming Amtrak passenger trains coming almost directly at me at 79 mph!  Now THAT would be a test!  It turned out that the southbound train was in direct sun, but the northbound train was backlit.  An even harder task.  The camera performed admirably and successfully tracked the trains.

Here are the settings I used:

The first train I photographed was backlit and northbound.  I had set my advance for L, @ 3 fps.  The second train was southbound and for that one I set my advance for H, or 8 fps.

Autofocus was set for the camera to choose the area.  I tried the other setting and tracking a train the day before it did not do as well.  Setting the autofocus to letting the camera choose the spot to focus seemed to work well.  However, I normally wouldn't use that setting.

Shutter priority
1/1000th second
ISO 400
Auto White Balance
Raw + LF JPEG

These are from the JPEGS from the first set of images.  These are cropped somewhat to show just the train and signal towers.  Additionally, the shadow slider was increased to lighten the front of the backlit train.  I think the camera did a great job tracking this fast moving train and keeping it in focus.  I am pleased with the sharpness of the focus in these images.  Again, this was at 3 fps.






The next set are from the southbound train.  Again, from JPEGS.  These haven't been cropped or edited.  These were made at 8 fps.









The last two images are not quite as sharp as the others, but I believe that has to do with shutter speed more than focus accuracy.  At only 1/1000th second with an object that close, now moving "by" me instead of "at" me, the shutter speed, from my many years experience photographing trains, was inadequate.  Having written that, you don't see the slight blur unless you enlarge the image almost to 1:1.

The next series of tests were of the ability of the camera and lenses to record fine detail, especially of greens and foliage.  Here is where I found problems with the files.  All these files are from RAW files.  No editing except to add some sharpening and reduce the file size for web viewing.  But what I write is what I saw on my large monitor.

The first was made of the tops of southern yellow pine trees.  The focus was set for "S" and locked on.  In the full image, it looks good, but in the 1:1 crop that the image looks bad.  Detail is mushy.  That is the only way I could describe it.  Almost looks like the camera didn't focus properly.  But, I repeated it several times with the same results.  The focus in the viewfinder "locked on" in S mode.  I'm not happy with this at all.  I even pumped in more than normal sharpening for these images to try to improve them.




Let's look at the next test.  Another set of images from RAW files, again, with more than normal sharpening. Its a nice scene, backlit with lots of detail. The 1:1 image shows the rocks on the path as well as the small patch of green grass on the right as "painterly" and the detail doesn't look real.  Again, it appears smeared and doesn't render the individual rocks clearly.  Focus was on S and locked on to this area so it is in focus, according to the camera.  This not the result of JPEG compression as it also looked like this on my 24" calibrated monitor.  This is not good.



The next set shows similar effects.  Look at the branches next to the tree in the center of the crop.  The detail looks painted in and not crisp.  Same with the detail in the green grassy area.  Again, from RAW files in the release candidate 8.4 in ACR.  These are not JPEG artifacts as the RAW image looks like this also.  These were confirmed in focus at as I have also included a 1:1 crop of the branches against the sky, some of which are in front of the grassy area and some are behind that area.



These two above were made at 46mm, 1/140th second at f/11 with IS turned on.  If you are thinking that maybe there is movement blue in the camera, here is another 1:1 of branches against the sky from the same image.  There is no camera movement.


Not that all my images were not as I had expected.  Here is an image of a re-creation of the fortifications at the eastern front of the Petersburg National Battlefield Park.  The images taken of the fortifications look excellent!  Everything is crisp and detail is rendered as it should.  But note, there is not a lot of green foliage in the image.  Again, from RAW, a bit more sharpening than I normally would apply for viewing on the web.



Again, I'm not sure how well these issues appear on the web, but they are very apparent to me when looking critically at the images on my 24" high quality calibrated monitor.  One last set.  This is an old, abandoned house that I have photographed in the past.  Instead of the details of the old, weathered clapboards and tin roof being visible, it is almost as though the wood has been sanded smooth.  I can only describe the files as looking "processed" or "smoothed" rather than crisp and clear.  I pulled up images I made of this same house a year ago with my little Canon G11, shot on RAW, and they actually look better and more detailed than these images.  I'll have some final comments after you look at these two.



Again, these are from RAW files.  The only editing was adding some sharpening. 

I don't know what to think about all this.  I think Fujifilm has made a terrific camera.  The X-T1 is a real "shooter's camera" as I like to say.  I really like using it.  The size, weight, shape and controls are perfect.  The ergonomics and haptics are great.  Everything works as it should and it screams quality.  I think the problems with the files have to do with Adobe not rendering the RAW files to reflect the real quality available.  I had read one photographer's blog where he felt that Adobe had now solved the past issues with this version of ACR.  I disagree.  I see mush.  Additionally, in the LF JPEGS from the X-T1, I am seeing compression artifacts at 1:1.  I don't remember seeing compression artifacts in other files at best quality JPEG settings.

I don't want to pick on this camera or the resulting images.  I really like using it.  I believe Fujifilm has solved, for the most part, the problem with tracking focus.  Their color rendition is nice. They have the largest and fastest EVF.  The lenses are terrific. But, in my heart, I think my images from my Olympus E-M5 and E-M1 are sharper, crisper and render fine detail better.  That is probably because it has a conventional Bayer-type sensor and not the revolutionary X-Trans sensor. Image quality is the final arbiter of whether or not one is satisfied with a particular camera.  After all, it is about the final image, not the equipment used to make it.  Equipment/gear can make the experience more or less pleasurable, but image quality is what matters to me when choosing a specific camera for any specific assignment.

To say I am disappointed is an understatement.  I ended up purchasing three copies of this camera, the final one working properly.  So now I have a decision to make.  Should I wait for Adobe to develop demosaicing algorithms that will do this sensor justice?  Should I buy, what I am told is a better raw converter, Photo Ninja?  I don't mind spending the money, but I would hate to have to incorporate several additional steps into my import, conversion and editing process just to get files with which I will be happy.  As of this minute, I don't know.  I'll have to mull it over and weigh the pleasure of using this camera with the frustration of the mushy detail in the files.

Thanks for looking. Enjoy!

Dennis Mook

Many of my images can be found at www.dennismook.com.  Please pay it a visit.  I add new images regularly.  Thank you.


All content on this blog is © 2014 Dennis A. Mook.  All Rights Reserved.  Feel free to point to this blog from your website with full attribution.  Permission may be granted for commercial use.  Please contact Mr. Mook to discuss permission to reproduce the blog posts and/or images.

26 comments:

Gianlu said...

Hi, I discovered your blog recently and I like it very much. I own a olympus OMD em5 with some lens....I was very interested in Fuji when they come out, but after playing with some raf file I find the files to have the same problem that you stated, in addiction I find that they blow highlight much more easily then the OMD files.I don't like that use a some NR even in the raw files so the files look smoother and with less noise but with much less detail...sorry for my english! Ciao, Gianluca

HF said...

Interesting review agreeing with what I saw from Fuji x-trans sensor images. In my opinion this is s.th. which is expectable when having to interpolate colours out of similar colours 1) farther away than on Bayer sensors and 2) with less regular pattern. You have blocks of 3x3 compared to 2x2 pixels and if detail has a given direction, it can happen that you need to interpolate over 4 pixels for green or 6 pixels for blue, for example. When doing the interpolation two-dimensionally you could get false detail if your detail is oriented in one direction only (e.g. hair).
Additionally, the error is proportional to the pixel pitch when interpolating (technically algorithms usually are "second order" meaning the error is some constant times the pixel pitch squared. I f your spacing between pixel is now 3 times that of a Bayer sensor, in this particular direction your error is 3*3= 9 times larger). Depending on the algorithm an other side effect is that a smearing occurs, as the interpolation looses accuracy at the finest scales (this is a mathematical fact which cannot be discussed away. It's related to the transfer function in Fourier space). So Fuji cameras have sometimes less detail in the very fines scales smearing noise and giving the impression of cleaner pictures.
This is a common thing I regularly encounter in my work (on numerics, including interpolation). And this is what shows up in your pictures: smearing and false detail. But it's only visible at 100% and something I wouldn't bother if it's not the goal of having fine art prints. I believe you can improve raw converters to achieve better results, but it's not an easy way of implementing this. But I'm sure they will find a solution.

Anonymous said...

Hi Dennis, I think in some images you are zooming in on areas that are naturally starting to fall out of focus. In many you can see the focus point is elsewhere and you zoom in on the background. If you zoom on on the focus point area and it is sharp then the issue is your aperture. On the other hand I would print these images out at 16x20 and worry less about what a monitor is showing you. If it prints really well then no issue. I do feel most of your problems here are because you shot at something like f4 im guessing.

Peter F. said...

Great insights and observations. I recently bought cheap an X-A1 with its bayer sensor, but I keep reading hoping the RAW conversion problems will go away so I can rationalize a used X-E2. In the meantime....

I've concluded it is hard to beat the E-M1 + 12-40. For travel I'm going to throw in the lightweight plastic 40-150R which can be bought new for $150 and is really sharp. Put a raynox or Canon 250D closeup lens on the 40-150 and you've got focus down to about 10" and a nice macro (I believe you wrote you had the equivalent in a Nikon closeup lens)

BTW, I found that because the 12-40 focuses at close range already, not much if anything is gained when adding a closeup lens to it.

I'm looking forward to your next report!!

Adrian said...

Dennis,
in my opinion LR is a convenient but not an excellent Raw converter. It has even issues with Olympus EM-1 Files. Try to shoot at high ISO indoors in dim light and look at the details- you will find color bleeding. Other raw converters (Iridient, Capture1) don´t show this. Another issue with X-Trans raws is moire (a joke-considering the fact, that according to fuji one of the advantages of x-trans is "no moire"). This is also highly dependent on the raw converter. With Iridient you will see very good detail rendering, but moire/artifacts in detail like eg.a tree; Capture one shows decent detail with little smearing and artifacts/moire only in architecture, LR has a perfect moire removal tool but shows no detail at all :) Photo Ninja is a nice raw converter (has the best detail) but has X-trans problems with posterization in gradients (e.g. in a blue sky) (btw: does not even support EM-1 directly after how long is the EM-1 on the market?)....
Have a nice weekend,
Adrian


Dennis Mook said...

Gianlu, I think you may have hit upon something with your comment. I'm now wondering if Fujifilm includes noise reduction in their RAW files by default. That would explain the "smoothing" I see and loss of crispness. That would also explain why they are able to show less digital noise at higher ISOs than other manufacturers. I sure hope they aren't doing that. That would be a similar corporate policy to Sony's forcing us to accept compression into their RAW files whether we want it or not. Let me decide how much noise reduction I want for any specific image file and don't make me accept your (the manufacturer's) idea of what is best for my images in the way of noise reduction. Then again, that may not be it, but it sure looks like noise reduction. Thanks for the comment and thought.

Dennis Mook said...

Peter F., Thank you. What a quandary? I really like everything about the camera, but I can't seem to understand what is wrong with the images. Why aren't others seeing what I am seeing? It is very apparent to me. I'm not really a "pixel peeper," but one has to look at the detail in file to ensure it is rendered properly and these files just don't render the detail for green foliage and some select other subjects as it should. In-camera noise reduction in the RAW files may be one answer besides poor RAW conversion software. I'd like to see it worked out.

I still have my E-M1. I had used the Panasonic 12-35mm f/2.8 lens extensively and I can categorically state that the resulting images were technically fantastic. Since I have now sold that lens, I will purchase the Olympus 12-40 for several reasons, one of them being what you stated–the close focusing ability. I have a Nikon 6T filter with dual elements that I have used for over 20 years, but I don't think I will need it for the 12-40. I'm preparing for a month long road trip following the Lewis & Clark Trail from St. Louis to the Pacific Coast. I am planning on leaving the Nikon D800E and lenses at home and only taking the E-M1 and a couple of lenses. I trust it that much. The idea is lighter, smaller, less expensive with no sacrifice in image quality. That's my mantra these days.

Anonymous said...

Finer detail from the OMD with which lens? If I where planning a pixel peeping party, 'testing fine detail', the 18-55 kit lens, presumably what you shot with, might not be the last one I'd choose, but it would be close. So far, though I've been retaining all my raw files, I've avoided DNG conversion. After reading your article, I walked out onto my deck, took a shot of the XT/23mm and did a side by side comparison after running the DNG converter. Bottom line is Abode has a serious flaw, one seemingly peculiar to this rev of the DNG converter.

Here's a link to a test shot... http://www.flickr.com/photos/15286515@N00/13327868885/

It then occurred to me to do the same thing, but using with an X-E1 jpg/raw straight into LR 5.3, using the current integrated raw converter.

http://www.flickr.com/photos/15286515@N00/13329291974/

Sadly the focus want quite right, but the difference is obvious.
You are using release candidate software. Its clear there is a problem with it. Why the results are so obviously flawed when one would have expected all they needed to do is recognize a new camera name and do whatever they've been doing for the XEs is certainly a mystery. But I think your disappointment in this case should be either be directed at Abode or yourself for thinking that beta software is the equal of production.

Finn said...

That's what keeps me from going all in into the Fuji X system .... I would love to sell my Nikon gear but the odd rendering of fine details just doesn't look crisp and clear to my eyes. So many people praise the picture quality of the Fuji cams but I just don't see it ... :\

Wish: X-T1 with Bayer filter .. :P

Gianlu said...

yes, you are right!...but from a marketing point of view it's a different story...they can say: we have the best apsc sensor on the market regarding high iso noise, we can comete with FF..and so on...
Really don't understand why so many people can't see these things like a real problem...
I think Photo Ninja can do much better then LR even for Oly files. I use the program with lightroom and I'm very happy, but you have to export and reimport tiff files...so it's not the best for a fast develop of raw files.

HF said...

First of all. I stumbled over your blog recently by chance and really like it. I don't know whether my comment didn't get through or was deleted this morning. I tried to explain that the x-trans sensor pattern inevitably requires interpolation over larger distances (6 pixel widths for blue, for example, in one direction if details are oriented accordingly). Therefore interpolation errors are much larger and if 2D interpolation is used, false details may occur. If biased interpolation is used, technically speaking this further leads to real and imaginary part of the transfer function causing dampening (smearing of detail). I think this is exactly what you are seen. This behaviour is very well known in numerics applied to engineering or physics (where I work in). More sophisticated algorithms are therefore necessary. My bet is, that raw converters within the next 1-2 years get this under control. As this is usually seen at 100% only, I wouldn't care if wall-sized fine art prints are desired. The new Fuji seems otherwise to be a great camera.

Dennis Mook said...

Thanks for the comment. Most of the images were made at f/8 and some at f/11. I checked other parts of the image, both in front of the area in question as well as behind, such as small branches, and both are well within the depth of field. I first thought as you did, but I ruled that out with further inspection.

Dennis Mook said...

I had been using the Panasonic 12-35mm f/2.8 as well as the Panasonic 35-100 f/2.8. Both superb lenses, in my opinion. If you look at some of my precious blog posts, say from mid-November on, most of the images were taken with the E-M1 and one of those two lenses.

Dennis Mook said...

HF, my apologies as I have been tied up all day and just pushed the "publish" button about 30 minutes ago. You have raised some very interesting scientific points. I would like to better understand the physics of the X-Trans sensor as compared to a CMOS/Bayer sensor. I'm always trying to better understand and improve, even after 44 years as a photographer. It is about the journey!

Dennis Mook said...

You make a good point. I looked at both sets. Easy to tell the difference.

Peter F. said...

Dennis, You might be interested. I posted something about the closeup capability of the 12-40 here:
http://peterfraileyphoto.blogspot.com/2013/12/olympus-e-m1-journal-entry-7-12-40mm-v.html

Mark said...

Isn't it true that Bayer versus xtrans is really about colour interpolation and the spatial intensity resolution should be comparable. Maybe shooting in black and white (for a test) will remove xtrans interpolation as a variable?

Anonymous said...

Without question, Adobe seems to have had more problems with XTrans than one would have hoped. And buying into XTrans has so far meant having a bit more patience and fortitude than had one stuck with a Bayer sensor. Some of us feel its worth it, others do not. Regardless, the latest version of LR if not perfect is quite acceptable for XTrans raw. The results you and I both saw, look as if they took the original version of the XTrans converter instead of the latest one. Were anything this bad to make it into production, there'd be hell to pay.

If you'll permit me, as someone who's left DSLRs and has shot Fuji since the initial release of the XE, I'll offer a different view on the question of Olympus versus Fuji. I've yet to touch a 1, but I have spent a little time with the 5 which I found to be a brilliant little camera. In the end, the IQ differences between these two lines of cameras is, from where I'm coming from, largely moot.

With the XT appearing and resolving so many of the criticisms related to speed from the previous models, now more than ever the decision to shoot Fuji comes down to whether you favor a shooting interface from the 20th versus the 21st century. Fuji has consistently tried to create a UI that preserves tradition, Oly has preferred to offer a new take on how things should be done. Those of us already in one camp or the other have already made that decision and unless you're exceedingly unhappy with it, IMO there's little practical reasons to switch systems at this point.

Just as I was tempted by the release of the M1, I'd suspect its adherents find aspects of the XT quite compelling. But buying into a camera system, given the many thousands spent on glass, is a far larger and longer commitment than should be driven by the release of the latest body. As great a camera as I believe the XT to be, if your considering dumping your Oly and 5Gs worth of 4/3 glass for one, its likely you need a GAS intervention. OTOH, Adobe's annoying software problems not withstanding, if you're coming from DSLR land the non-traditional choices are getting more and more interesting and if you like a more old school interface the XT should be on your list.

c0ldc0ne said...

Thanks for your ongoing thoughts on the X-T1 which largely reflects my findings with the X-E2. I have found myself switching between several RAW converters, hoping to find the image quality that everyone and their brother rave about in reviews and blog posts. And like you, I found that Light Room is not the best option by far in this regard.

An observation about your focus tracking test if I may. It may be that it's just hard to gauge critical focus from web-sized images, but I found that most of these tests I've seen so far (including yours) mask any and all issues with focus tracking by the massive DOF resulting from the chosen test setup.

In all of the images from your test I looked at, I found that pretty much everything between the front of the train and the area closest to the camera is in focus. That would mean that even if you had put your camera in manual focus and not touched the focus ring at all, the front of the train would still be in focus throughout the series you shot.

IMO, to test focus tracking, you need a big aperture, long focal length and/or close distance to narrow down the DOF and really put the tracking performance to the test.

Bob Jensen Photography said...

I have a complete 5D MK3 system and a complete D800/800E system and am making the switch to Fuji X for various reasons. I did find that the digital raw files look much better when I open them with Photo Ninja - vastly superior, at least for the moment, to Adobe.

Dennis Mook said...

Thanks for the comment. I agree entirely with your point. However, instead of testing the theoretical limit of gear, lenses and processes, I try to use them and see how they work as I would in the real world. I have been photographing trains, among other things, for over 40 years. The SB images of Amtrak were made at F/5.6 at 1/1000th sec, typical of what I would normally use when photographing a moving train. So, that is how I based test as well as many of my tests–how do things work when using them how I would normally use them?

Another example is when I compare image files from different cameras or lenses. Some reviewers compare files directly out of the camera. I don't as I would never use a file, either JPEG nor RAW as it directly comes out of the camera. I compare files after making the best edits and producing the best images I can with the files in question. So, the end result is more important than strict empirical testing, in my case. But, again, I agree with your premise for strict testing of tracking focus is important.

c0ldc0ne said...

Far be it from me to question your technique for shooting trains. I only ever shot slow moving or stationary steam trains myself, so I guess it's a safe bet that in terms of skill, you're better equipped in this arena.

What I was suggesting though is that this approach, while perfectly sound in and of itself, may not necessarily lend itself to testing AF tracking. I recently came across another review where the X-T1's tracking performance was tested on someone walking/running towards the camera (also a pretty typical use case) using a much smaller DOF, and the results were not quite as impressive (about a 30-40% keeper rate).

As for your second point: I see what you mean and I agree up to a point in the sense that I too am not a big fan of scientific tests and prefer a more hands-on, real life review. I also agree that it is ultimately the end result that should be the deciding factor. But the problem in this approach is that it also embodies you post processing skills. So while I could walk away from your tests under the impression that I can get great results from a particular camera, my personal experience may be quite disappointing if I fail at reproducing your PP technique.

Dennis Mook said...

I certainly didn't infer that you disagreed with my methods, only commented and giving another opinion. Diverse opinions are great and necessary. I also agree with your understanding a final image may not look like another photographer's due to editing skill level. Absolutely valid. I feel the same way and always seek out several reviews, tests, etc., from a number of individuals, then weigh all their opinions to help formulate my own. Thank you again for your comments.

Unknown said...

I'm not sure I understand the value of inspecting small areas of an image at 1:1 (a.k.a. "pixel peeping) if the differences you see would not be visible at normal screen resolution or in anything other than a gargantuan print. This would puzzle me regardless of what camera you were testing. That said, even if we disagree on the ultimate importance of the "mushiness" you refer to, I'm not implying I'm right and you're wrong. You have every right to suit yourself. If you can find the image quality you seek in some other camera and its drawbacks (all cameras have them) are more acceptable to you, then I'm happy for you. Thanks for taking the time and trouble to publish your test results and to be so honest about what you did and didn't like.

Dennis Mook said...

For my personal work, I usually don't "pixel peep" as I believe the ultimate judgment of a photograph is content, not technical quality. However, I submit much of my work to an agency for stock purposes. Their editors do pixel peep and I have to ensure that what I submit is very detailed and as sharp as it can be. Otherwise, the images don't make the cut. Less than detailed and sharp images don't suit their clients' needs. So, I'm caught between content and technical quality.

As for the "mushiness" I saw in many of my test images, it simply shouldn't be there, whether or not anyone looks at the image 1:1. The detail is clear and apparent from all my other cameras, which I believe is the industry standard.

Thanks for the comment.

Unknown said...

Thank you for your prompt and enlightening response.