Saturday, November 24, 2012

Bookmark and Share

Fujifilm X-E1: Dynamic Range and More Comments



The above animated GIF I made, shows an example of what sort of range one can get from a single raw file. It is an image that was included in the night shot gallery linked to in my previous X-E1 post. The dark version is how the default conversion settings look (my defaults, that is, ones that already give slightly more shadow detail than normal), when processing through Adobe Lightroom v4.3RC. The bright version is the final result. Most of the adjustments were just done with the highlight and shadow sliders, but in this case, I wanted to darken the bright interior down more without darkening the outside lit walls too much, so I did some selective adjustments for the windows. I was a bit sloppy and didn't isolate the main yellow door frame enough in the "after" version, so it has darkened down a bit too much for my liking. In any case, it serves to illustrate that one can get a remarkably good dynamic range from a single raw X-E1 file. Hopefully people don't find the continuously running animation above too annoying!

If you click the image (or the link below it), you'll be taken to a gallery with 24 image pairs, all taken from my previous X-E1 galleries, each pair starting with the "after" version and the image to the right is the "before", the original unadjusted version. It is effective to use the right and left keyboard arrow keys to switch between image pairs, so you can see a smooth fade from before to after and back, similar to the above animation but in higher resolution. Some images have fairly minimal adjustments but I thought it would be helpful for people to see how much manipulation has gone into the final versions, especially the night shots. On a few you'll see some slight tilting caused by perspective corrections where I either didn't quite get the shot perfectly lined-up, or where I intentionally tilted up or down for compositional reasons.

It terms of further comments about the X-E1, to be complete, I should clarify a handful of things about it and the XF lenses I've tested, as well as add a few more comments on its operation.

In my previous review, when I said that the 35mm and 60mm lenses were "...as perfect optically as one could possibly expect, especially at their extremely reasonable price points!", I did mean just that... at their price points. They are not perfect in an absolute sense of course, but since both lenses are under $700, personally I find that to be very reasonably priced considering their overall superb quality. For example, both lenses do have slightly more flare that I would like to see, although since they are longer focal lengths and both come with good lens hoods, real problems would be quite rare. Flare is much more of a problem with wide-angle lenses generally.

Also, the 35mm does suffer from a bit of astigmatism (at least mine does), seen on bright lights in night shots near the edge of the frame when you are shooting at wider open f-stops. However when stopped down to at least f/8, even bright point source lights near the edge of the frame are rendered extremely well. Putting the performance of the $599 XF 35mm in context, note that even the rather expensive Nikkor AF-S 35mm f/1.4G ($1800) doesn't render difficult night shots perfectly unless it is stopped down a fair bit as well.

The 60mm lens, while superb optically, is indeed a bit slow to focus. One of the reasons for this, I'm guessing, is the extreme range it has to rack through, from closest focus to infinity. The other lenses are certainly snappier from a focus speed standpoint, that's for sure. The new 18-55 zoom is reputed to be a lot faster, although honestly, when I tested it, it didn't seem all that much faster than the 18mm or the 35mm. However, it was most definitely faster at 55mm than the 60mm macro is.

I also said the 18mm was "weak", and I should also clarify that a little. The lens is actually quite sharp across most of the frame, even at wider open f-stops. However when stopped down to even f/8, there is still a noticeable falloff in sharpness at the corners. If your shots tend to be more geared toward subject isolation at wider f-stops, then the 18mm will likely perform very well for you since in those cases, corner sharpness is usually not much of an issue. On my night shots, one of the reasons I shot many at f/13 when using the 18mm lens, was to try and equalize sharpness across the frame and ensure that point light sources near the edges were rendered well. On the other hand, the 35mm doesn't need to be stopped down as far for good corner sharpness, although I did shoot it at f/13 as well on a few occasions mainly for depth-of-field reasons, not for improving corner resolution.

One thing about the X-Trans sensor is that since it has no anti-aliasing filter and is inherently so sharp, one can start seeing the effects of diffraction beyond f/8 or so, which slightly reduces the absolute sharpness of an image, therefore I would rather see a wide-angle lens achieve harmonious sharpness from corner to corner by f/8 or thereabouts, or even earlier if possible of course. I do have high hopes for the the upcoming XF 14mm f/2.8 though... keeping my fingers crossed!

I remembered a few more items of feedback for Fujifilm, listed here...
  1. On pretty much every (non Fujifilm X) camera I have ever shot with, when you half-press the shutter to lock focus, then squeeze off a shot but keep the shutter half-pressed after the shot, the focus will stay locked on until you fully release the shutter. On the X-E1 (and my X100 and the X-Pro1), keeping the shutter half pressed does not maintain focus lock. This is actually quite annoying, especially when photographing people, since I will often focus-lock, recompose slightly, then take a few shots in succession, prompting the person (people) that I am photographing to change their expressions, to smile etc. Having the camera refocus each time is truly annoying when quickly trying to shoot that way! This is one of the reasons I set my AE-L/AF-L button to toggle a focus lock. However after the focus lock is toggled on, as I mentioned before, one cannot change the f-stop or shutter speed! Grrr... I hope both those items get corrected in a firmware update soon!
  2. An enhancement I would love to see is the ability to create a customized menu that can be accessed by double-clicking the 'Q' button in rapid succession. While the existing 'Q' menu is useful, it is cluttered with many settings that I will never change on a regular basis. Also, the 'Q' menu grid, while space efficient, perhaps doesn't have quite the same degree of clarity as the regular text-based menu for some settings. For example, I would love double-click access to a custom menu that has white-balance, IS-mode, self-timer, dynamic range, Autorotate PB and format on it, all in a regular text-based menu. Further to using this custom menu quickly, see the next item below...
  3. When in the text based menus, why not have the rear thumb-scroll-wheel also change settings just like in the iconified 'Q' menu? This would make changing settings in the main menu much more efficient!
Anyway, point number one is very annoying and I hope it gets fixed sooner rather than later. Point number two probably would require a fair bit of user interface work, so I don't expect we'll get that soon, if ever. Point number three ought to be quite easy to incorporate one would think...

Lastly, with regards to raw conversions, keep in mind that the 2400 pixel wide images in my previous galleries will rarely show any minor image quality problems. The Adobe / X-Trans conversion issues will often only show up on much closer inspection of the images. Someone asked me the other day whether or not I thought the X-Pro1 or X-E1 would be good for product photography. When I asked what sort of products, the photographer said bottles, cans, tubes of toothpaste... those sort of grocery store items. With all the fine black letters on white backgrounds that one can find in products like that, well that could turn out to be a post-processing nightmare! Once fine detail gets to be at a certain pitch, suddenly you might find demosaic errors caused by Adobe's non-optimized raw conversions of X-Trans raws, such as colour bleed between black lettering. On the other hand, in-camera JPEGs and SilkyPix raw conversions generally don't show any of this...



Click on the above photo for a closer look! On the left of this 100% zoom comparison, is an out-of-camera JPEG, on the right is the Lightroom rendering of the RAF file. Some of the black text has severe amounts of green bleed in between the lettering, the red text has some red bleed, and the black text towards the top right even has an overall magenta cast. This is the worst Adobe / X-Trans conversion example of this type that I've come across, but it shows well the type of problematic demosaic errors one can potentially run into. By the way, Fujifilm's supplied SilkyPix software also handles this particular image just fine, with no colour bleed, looking essentially like the camera's JPEG rendering.

A similar sort of colour bleed can be seen in the railing in front of the blue taxi, near the top right of the very first image in my "Thirty" gallery. The railing itself has taken on the taxi's colour. In a few other shots in the same gallery, you might be able to notice a similar sort of colour bleed in the corners of the white lettering on stop signs (image #23), or other smaller letters on more distant signage (image #5).

I have now shown several large 24x36 inch prints to other photographers, several of which do exhibit a few of the Adobe / X-Trans conversion issues (colour bleed in signage, some "watercolour" smoothing in foliage) and everyone simply commented on how lifelike, clear, sharp and three-dimensional the prints looked. Not one person singled out any issues and made any comments, at least not until after I showed them exactly where to look for problems. I am certainly not trying to belittle the severity of the issue, and Adobe will hopefully address this soon, but I would say that for many types of shooting, unless you are really "pixel peeping", the raw conversion problems may not really be that big of a deal. Well, except maybe if you are shooting a lot of photos filled with fine dark details against light backgrounds! In any case, to be on the safe side, it wouldn't hurt to shoot raw+jpeg for now, if you are relying on Adobe for your X-Trans raw processing.

Anyway, that's it for now. Again, I hope nobody gets the impression that I don't like the X-E1 with all this seeming negative feedback. On the contrary, it is because I enjoy shooting with the X-E1 so much, because I appreciate the tactile controls and handling, and because I find the lenses and image quality to be so outstanding generally, that I want to see Fujifilm update this body to make it an even better camera for serious work. And the same goes for the X-Pro1 too of course...

7 comments:

rogerb said...

Thanks Mike, for your excellent post. Your illustrations of dynamic range recovery are very interesting and most useful. Many complain about how difficult SilkyPix is to use but they've probably forgotten how brutal the learning curve was for Photoshop;-)
Do you think a workflow involving raw batch processing with SilkyPix (contrast reduced across the board to eliminate blown highlights) generating 16-bit TIFFs then imported into LR would give you a higher quality product than just using LR by itself?

Mike Mander said...

Hi rogerb: You're welcome. All these samples were from Lightroom and despite its occasional issues with the X-Trans sensor files, it is still *way* easier to make dramatic adjustments, such as the animated example, with it than with SilkyPix. I suppose if you create a preset to generate a really flat 16-bit TIFF, then yes, maybe that workflow would be better in some instances, but I still find that for most images, staying with Adobe software for the entire raw conversion is better. However if a particular image has an issue (colour artifacts or too much "watercolour" smoothing), then processing a really flat (wide dynamic range) 16-bit TIFF is certainly feasible and would be a way of salvaging an image. I have now seen a small handful of really bad examples of smoothing or artifacting, but it is nearly impossible to predict when this will happen. Luckily, I'd say that well over 95% of the images I've shot would reproduce well, even on a large print, with virtually no noticeable X-Trans raw conversion issues... especially with careful processing in Lightroom or Photoshop. In other words, not the default settings...

Justin Sangani said...

Great post, thanks for so many wide ranging examples. Your processing is very well done. I am glad to see the colors in the shadows hold up when pushed like that, similar to the new sony sensors. It's hard to tell at these sizes, but do you find the details in the shadows good as well? With my 5D it's a lost cause ;).

Thanks again-

Mike Mander said...

Hi Justin: Yes, shadow details are good, clean with good colour tonality and no banding or cross-hatching noise. Many of the processed example shots can be seen much bigger in the other X-E1 galleries I've posted previously and I don't think you'll see any issues. I would still say that the best Sony sensors (for example Nikon D600 and D800/E) outperform Fuji's X-Trans as far as pushing crazy amounts of clean shadow detail, but my X-E1 isn't all that far off my D800 for dynamic range, at least for the vast majority of shooting situations. Usually the X-E1 dynamic range is more than sufficient but if you really have to reign in the highlights and dramatically open the shadows, well then the D600 or D800 will likely do a better job. Note that my statement is coming from comparing raw images through Lightroom. On JPEGs, I'd say the Fuji might actually be leading the pack!

Fuji X-Files Admin said...

Hey Mike,

Thank you very much for your effort to show the DR of the Fuji X-Trans sensor. I too see quite a bit of DR increase compared to my Canon 5D MK II. And the better color rendering and WB are also superior to that of my Canon cameras.

To underscore your statement about the Fuji XF lenses, I have just done an extensive test comparing the XF 35mm to the highly regarded Canon EF 35mm f1.4 L - both cameras attached to the Fuji X-Pro 1.

Have a look for yourself to see how the Fuji lens did in comarison:

http://fujixfiles.blogspot.com/2013/01/how-good-is-fuji-xf-35mm-f14-r-lens.html

krist said...

Hello Mike,

Thank you for all the information provided! One question: saying that "On JPEGs, the Fuji might actually be leading the pack", could one assume that the X-Pro1 could be better in dynamic range than D600 for files developed with SilkyPix vs. RAW D600 files developed with Adobe products?

Mike Mander said...

Hi krist,

No, I wouldn't say that. I just mean that out-of-camera JPEGs from the X-E1 and X-Pro1 are better than the out-of-camera JPEGs from most other cameras. I would definitely say that the dynamic range of a Nikon D600 raw, processed through Adobe raw converters will be better than the Fuji raws processed through any converter I've tested so far. The D600 seems to be as good, maybe even a bit better, than my D800 as far as raw dynamic range and the Fuji sensors are not quite there. Still, considering the Fuji cameras are APS-C, versus the D600 being full-frame, they are putting in a darn good performance...

Post a Comment