I almost hate to write this post, as Exposure To The Right (ETTR) is a near religious topic on the inter webs and I don't want to wade into the fan-boy vs. haters mudfest. But I do want to know how ETTR plays out with the unique Fuji X-Trans sensor in the FujiFilm XE-1, and all other Fuji X Series cameras that share this sensor such as the X-Pro 1 and the X-100s.
The stated advantage of ETTR is you can overexpose a scene in order to give shadow detail a fighting advantage over the noise that lurks in those same shadows. One caveat of ETTR is that the scene must be a low-contrast scene - the highlights and shadows must be well within the exposure range of the sensor, enough so that the scene can be overexposed by a chosen amount and then later "pulled" down in post-production without losing highlights. Another premise is that you MUST be working with the RAW file to correctly pull the exposure down without losing detail in the highlights. There is plenty of ETTR theory, go to Google for that, but I'm much more interested in actual photo tests. Boots on the ground, if you will.
"Boots On The Ground" Test
I created a dark low-contrast scene for this test. The boots and floor are dark and rich with color, with plenty of shadow areas for the noise to live in. In this sequence I worked my way up through the ISO range of the XE-1, one stop at a time, compensating with an equivalent increase in shutter speed. I then repeated the same sequence one stop over exposed. Using Fuji's in-camera RAW conversion I then applied a -1 stop "pull" to each of those shots. What I am showing in this are the JPEG outputs of each shot, with three sample areas shown at 100% zoom.
Why In-Camera RAW Conversion? I used the in-camera RAW converter because this is available to all Fuji Xr's who want to replicate this test. The in-camera RAW converter is excellent - it applies noise reduction, color correction, film mode grading, moire correction and generates the beautiful JPEGs Fuji is known for. The RAW converters in Capture One, Lightroom or Aperture will all provide somewhat different conversion, so for consistency and simplicity I chose to use the in-camera converter for this test. I believe the end results will be similar whichever tool is used.
Histograms are important with ETTR, so I'll show my work. I've made sure the shadows (the black boot soles) aren't clipped to black. There is a small highlight spike, but that is a specular highlight so I'm not worried about losing that. I'm also including the histogram from Apple's Aperture as it shows each of the color channels at work.
Here's the test sequence shot with a Fuji 35mm f/1.4 lens. The teal boxes show where the 100% zooms are taken from:
(The sequence may be scaled on your screen, click the image for a full scale "lightbox" view, or download the full quality version by clicking here.)
So I would agree that in each case, the "pulled" ETTR version looks better than the properly exposed version - there is less noise. But there are also two very interesting discoveries here.
- There is noise throughout all tones of the image, not just the shadows. There is noise in the highlights and mid-tones, but especially in the contrasty transitions and details. The concept that noise lurks primarily in the shadows doesn't seem to hold true.
- A "pulled" image has nearly the identical noise to that of a correctly exposed image one ISO stop lower. For instance, the "pulled" 400 ISO image has the same noise qualities as a properly exposed 200 ISO image.
Disovery #2 is quite interesting, let me demonstrate by stacking properly exposed images side by side with their the "pulled" ETTR version from one stop higher. I'm rotating the images on the right so they seam up together, hopefully to make the comparison easier.
It seems to me that the only time ETTR gains an advantage is at the base 200 ISO, where we simply can't go further down in ISO. At any higher ISO the ETTR technique adds complication, and possible highlight loss and tone shift, without an advantage.
But even at ISO 200 this wouldn't really be ETTR at all, but rather the long-standing, pre-digital concept of underexposing a high-key scene or over-exposing a low-key scene to get the important information recorded at the sweet spot of the imaging system. More easily stated as this: Place a low contrast scene at the middle of the histogram.
Please keep in mind, I'm not making a claim for any other cameras, just the FujiFilm X series based on the X-Trans sensor. Digital sensors and amplifier circuits are far more advanced now than they were when ETTR was first introduced, the Fuji sensor is unique in many ways, and this may be one of them. And though there are many theoretical reasons why ETTR should work, testing with actual images offers the proof in the pudding: Fuji X shooters may be better off moving beyond ETTR.
If you want more theory, here are two articles offering both sides of the argument:
- Optimizing Exposure (with ETTR):
- Why Exposure to the Right is Just Plain Wrong: http://chromasoft.blogspot.com/2009/09/why-expose-to-right-is-just-plain-wrong.html
Let me know what you think, and by all means lets repeat with similar tests and compare results!
Update: I've followed this up with a post on the opposite extreme: Fujifilm XE-1 Push Processing
By Adriel Henderson