Dynamic Range

The S3 Pro's unique selling point is that it is the only digital SLR to have an extended dynamic range, it achieves this by the use of two interleaved photodiodes. Fujifilm refer to these as the S (normal sensitivity) and R-Pixels (less sensitive). The camera's image processor can then combine the values of these pixels to extend the dynamic range of the image. (This is supposed to be similar in effect to film which has different sizes of 'grain' which are sensitive to large or small amounts of light). The S3 Pro has 6.17 million of each type of pixel (12.34 million effective total), however it's important to understand that the R-Pixels are only used to extend dynamic range they do not increase image resolution. Note that for ease of reading we're using Fujifilm's nomenclature of an (input) Pixel however the correct term would be photodiode.

The S3 Pro provides four different options for dynamic range, firstly you have to select between 'Standard' and 'Wide' dynamic range on the camera menu. Selecting standard uses only the S-Pixels (hence would be pretty similar to the S2 Pro in terms of dynamic range). Selecting the Wide option mixes data from the two 'Pixels' to create the final image. In Wide mode you have access to three different modes for 'mixing'; Auto where the camera decides (depending we assume on how much R-Pixel data there is), Wide 1 (described as 230%; just over 1 stop more) and Wide 2 (described as 400%; about 2 stops more). To be honest I think the whole 'percentage' thing is deceiving, most people don't think logarithmically, and most people won't realize that you only gain at the highlight end of the light scale.

It's worth remembering that the S3 Pro's extended dynamic range is at the highlight end of the brightness scale, it doesn't add any more 'shadow detail' at the bottom end, this means that to get the most out of the camera's dynamic range logic dictates we should be over-exposing slightly with wide dynamic range mode (as we won't lose highlight detail and this will lift the shadows).

Aren't those R-Pixel's quite small? The answer to that is yes, they are, in fact by our estimations they're about the same size as a normal consumer camera's CCD pixel which shouldn't be much of an issue in most high dynamic range situations (ISO 100, bright outdoors or studio flash) but they will obviously get noisier the higher up the sensitivity range they are pushed.

Standard vs. Wide (JPEG)

Below you can see two images, the one on the left was taken in Standard dynamic range mode, the one on the right in Wide dynamic range mode (Wide 2 - 400%). We have overlaid an animated 'blinking highlights' display on these thumbnails as you would perhaps see in digital camera playback, those pixels which blink red were either overexposed or almost overexposed. As you can see the S3 Pro's Wide dynamic range approach clearly works, maintaining more detail in highlight areas of the image and avoiding overexposure clipping.

Standard Dynamic Range Wide Dynamic Range (Wide 2 - 400%)
ISO 100, 1/180 sec, F6.7, 6 MP ISO 100, 1/180 sec, F6.7, 6 MP

Another example below, same setup, wide dynamic range mode is good for maintaining all types of detail, here you can see that the blown out color which we have in the standard dynamic range mode is maintained and detail recovered in wide dynamic range mode. That said it's not clear that there is a 300% gain of dynamic range in the second image.

Standard Dynamic Range Wide Dynamic Range (Wide 2 - 400%)
ISO 100, 1/60 sec, F6.7, 6 MP ISO 100, 1/60 sec, F6.7, 6 MP

Measuring the difference

We've tried to measure dynamic range in the past but it's always been a little hit or miss, mostly because our previous methods required multiple exposures, each of which could be subject to different automatic tone adjustment. Our new test involves shooting a calibrated Stouffer Step Wedge (13 stops total range) which is backlit using a daylight balanced lamp (98 CRI). A single shot of this produces a gray scale wedge from (the camera's) black to clipped white. Each step of the scale is equivalent to 0.3 EV (a third of a stop) and so if we find the 'middle gray' and measure outwards we can define the dynamic range captured in the shot. We have chosen to stop at the dark end of the scale where there is no longer any measurable distinction to the next step and at the high end where the step becomes clipped.

I've chosen to present this data in two forms, firstly visually, these are reduced size crops of the actual step wedge shots taken in Standard and Wide 2 dynamic range modes. Depending on how well your monitor is calibrated you may or may not be able to see a difference between the two, but essentially the Wide 2 tablet carries more detail into the highlight.

Next in graph form, this is a little easier to see, as we have described earlier the S3 Pro's dynamic range 'extension' occurs at the highlight end, our test supports this as in both modes the amount of dynamic range available below middle gray is approximately 5.0 EV. However above middle gray we can see that Standard dynamic range mode clips at about +3.0 EV and Wide 2 dynamic range mode clips at +5.0 EV.

So it's fair to say that Standard dynamic range is about 8.0 EV (8 stops - about the same as the Canon EOS 20D) and that Wide 2 dynamic range is about 10.0 EV (10 stops). This is in line with Fujifilm's claim and is typically more than most other digital SLR's, you should remember though that the gain is made in highlights only. Perhaps it would be more accurate to name it 'expanded highlight dynamic range'. It would also have been more interesting if the wide dynamic range modes used a different tone curve and/or automatically exposed for shadows.

Exposure value (EV) or stops

* In our previous tests of dynamic range this has been the average for most digital SLR sensors, we are developing a standardized dynamic range test which will be used for all our D-SLR reviews in the future.