Monday, October 15, 2012

Nikon D600 vs. Fuji S5 Pro Dynamic Range Retested

Last week I tested the highlight dynamic range of the Nikon D600 against that of the Fuji S5 Pro.  That test had flaws, among which was that official raw support for the D600 had not yet been released.  Now that there is Adobe raw support for the D600, albeit in beta, I ran the test again.

The test is simple -- I arranged a test scene that included:
  • a white card
  • a gray card
  • a black card
  • a subject with approximately midtoned red, green and blue
  • a mannequin head
  • specular red, green, blue, white, and silver subjects.

The correct exposure was 100 ISO, f/5.6, 1/100.  I shot the test scene from 5 stops underexposure to 5 stops overexposure, then I checked if any of the subjects showed clipping by normalizing the exposure and checking for lost detail or color.  In addition, for the S5 I shot up to 10 stops overexposure, and for the D600 I shot up to 10 stops underexposure.  The tests were taken with the subjects facing north during an afternoon with uniform cloud coverage.

In the case of the D600 I tested without Active D-Lighting and with ADL set to Extra High.  For the S5, I tested both the standard 100% dynamic range mode and the 400% dynamic range mode.  Here are the results.

1. The maximum image size allowed by Picasa is 20mb per photo.  At that limit, I had to reduce the quality of the JPEGs to 80 and the long edge to 3200 pixels.
2. I only tested at 100 ISO.  The dynamic range will decrease as you increase ISO.  In the case of the S5, the extra photosites that allow 400% DR mode don't function above 800 ISO.

With respect to the highlight range, the standard mode of the S5 has lost color accuracy at 3 stops above middle gray.  Although the gray card is somewhat acceptable, the red, green and blue blocks have shifted colors:

Note that this limit needs to be adjusted for the tone of the subject.  Lighter toned subjects obviously lose detail at a lower threshold.  At 2 stops overexposure, the light-toned objects (card, ribbon, bear) are effectively around 4 stops above middle gray therefore you can see that they have lost detail.

With respect to shadow recovery, I was surprised that the S5 could get a usable image out of even 6 stops underexposure, at least at web-viewing sizes.  However, on my S5, you can see there is a long vertical line near the middle.

The vertical line shows up starting 3 stops underexposure.

Here is the web album with the Fuji S5 Pro at standard (100%) dynamic range.

We've seen this before so I thought it would be a useful control for the comparison.  With respect to highlight range, the S5 at 400% DR mode starts to lose color and detail at 5 stops above middle gray.  See the shot below where the colors of the blocks are starting to lose accuracy.  (Again, lighter-toned subjects will start to lose detail at proportionately lower thresholds.)

As you can see, one problem with the S5 is that when highlights get overloaded, they look magenta, which requires more editing to correct than a normal highlight blowout which can sometimes be cloned.

As with the 100% DR mode, the S5 was able to recover tones that were underexposed by 6.3 stops (!).  I didn't test beyond that. Of course the threshold of acceptability is very subjective, and depends on your intended audience and usage.

Here is the web album with the Fuji S5 Pro at full (400%) dynamic range.

The Nikon D600 loses color accuracy and detail at around 3.3 stops above middle gray.

In terms of shadow range, the D600 retains full color accuracy and detail at up to 6 stops below middle gray, with controlled amount of noise (better than the S5 at ISO 800 let's say).

Between 6 and 7 stops below middle gray, it keeps good amount of detail but there is a greenish tint to shadows.  Here is what it looks like at 6.7 stops:

The greenish tint is quite heavy at 8 stops below middle gray, although there is still plenty of detail.

Here is the web album for the D600 shots without Active D-Lighting

I also tried the D600 with Active D-Lighting set to Extra High.  I could not see a difference in dynamic range with or without Active D-Lighting.

Highlight range
+3.3 stops without Active D-Lighting
+3.3 stops with Active D-Lighting
Shadow range
-7 stops without Active D-Lighting
-7 stops with Active D-Lighting
Here is the web album for the test shots with Active D-Lighting.

Despite having a full frame sensor and being three camera generations newer, the Nikon D600's highlight range is not as good as that of the Fuji S5 Pro.  However, the D600's shadow range covers an even wider range than the S5 Pro's legendary highlight range, with noise that is controlled after recovery even from as low as 7 stops below middle gray.  In my next post, I'll show how I used that to achieve images similar to that of the S5.