The Dynamic Range of the Fuji S3 pro

Fuji's new Super CCD SR II sensor is touted to have an expanded dynamic range because it contains two pixel types. Is this true? How is it different from a standard sensor? (The Luminous Landscape has a "first look" review.)

I tested the dynamic range of the FujiFilm S3 pro digital SLR in two ways:

Versus itself and versus the Konica-Minolta DiMage A2
Versus itself only: standard dynamic range versus wide dynamic range at ISO 100 (its lowest available value) and ISO 1600 (its highest available value).

Right at the top, let me offer you two pictures that you can use to compare the Standard and the Wide dynamic ranges of the S3. See if you can tell the diffence before you read this. Then, after you understand what the camera does technically, see if you can then pick up the differences. Yes, they are subtle. Oh, and lest you doubt: the pictures were taken identically except for changing the dynamic range of the camera; in-camera 6MP jpeg images were written and no other post-processing was done.

Comparison of the Standard Dynamic Range and Wide Dynamic Range of the FujiFilm S3 pro digital SLR

Click on the word "Standard" or "Wide" to load a full-size image file.
Click on "Detail" to see a 500KB selection from each image that shows nicely the difference in hightlight detail between the standard and wide dynamic ranges.
I'd suggest that you download each into a separate tab for easy reference.

Standard Dynamic Range (Detail)	Wide Dynamic Range (Detail)

Characteristic curves of the F3

I re-shot the test a week after I first posted this page, using half-stop exposure increments (the smallest this camera is capable of) to show more clearly what this dual-pixel sensor does in the high values. Note that there is abrupt pixel saturation using the standard dynamic range, and a nice, film-like rounded shoulder using the Wide2 dynamic range.
Key:
Wdr = Wide2 Dynamic Range ("400%"); Sdr = Standard Dynamic Range
EI100 = ISO 100, this camera's slowest setting.
EI1600 = ISO 1600, this camera's highest setting.

Click for half-size or quarter-size images.

Comparison of the S3 with an ordinary sensor

Next, here's a graph of the results of my testing of the S3's dynamic range versus itself and the A2: There are three curves:

The blue boxes show the response of the S3 set to use its full dynamic range ("400%" in the manual), labeled "S3 Wide DR."
The red diamonds show the response of the S3 set to use a standard dynamic range ("100%" in the manual), labeled "S3 Std DR."
For comparison, a commonplace sensor, the results from my Konica-Minolta DiMage A2, labeled "A2." (I suppose you had that figured out already.)

Note two things:

Pixel saturation occurs one stop higher with Wide DR than with Standard DR or with the A2
Looking at the table of values below, note that the dark values -- Zones minus IV to Zone II -- scale a full stop brighter with Wide DR than with Standard DR.

Table of image intensity values (range 0 - 255):
The numbers below are the mean pixel intensities resulting from each exposure. If you open the image table in another window, you can view the test images, and measure them yourself as well. The test target is actually the left or darker panel; the right or lighter panel is one zone brighter. It is there to allow you to see whether a one-stop difference in subject brightness is detectable at any given exposure.

Zone	S3 Wide DR	S3 Std DR	DiMage A2
-V	2
-IV	4	3	3
-III	6	4	4
-II	9	7	6
-I	13	10	8
O	19	15	15
I	29	22	28
II	44	33	49
III	66	51	71
IV	96	77	103
V	134	111	129
VI	177	160	179
VII	224	211	236
VIII	243	254	253
IX	255	251	255
X	255	251	255
XI	255	251	255

Let's look at the "toe" of these curves in more detail. This shows that when using the Wide Dynamic Range setting of the S3, that image brightness or intensity gain is almost equivalent to the image brightness of a target one stop brighter than when using the Standard Dynamic Range. The A2 tracks the Wide DR brightness down to Zone O, then joins the Standard DR line. This produces somewhat less muddy shadows in A2 images through higher shadow contrast; conversely the S3 Wide dynamic range preserves information (texture) further into the shadows.

Summary:These curves show that the S3 gains about 2 stops of dynamic range over the A2 and over its own "standard" dynamic range. The image processing algorithm puts one extra stop at the top of the curve, diminishing or postponing "blocking" of highlights, and one at the toe of the curve, preserving shadow detail.

But will this be noticeable in photographs?

Well, that's going to depend on the subject brightness range and your exposure technique:

Only on a sunny day with snow, linen, or other bright values and shadows in which detail is important will you see effects of this expanded brightness range. Cloudy days and shadow typically have quite a restricted range, as you can see on the histogram if one's available to you.
The pixels of this sensor saturate just like any other pixels, and the exposure metering and image-processing algorithm only give you one stop of additional latitude.

What about that long, long "toe" on this curve?

Zone Zero is the subject value that traditionally was assigned to the deepest black that photographic paper could produce. The fact that both of these cameras record detail 3 to 4 stops below this value means two things:

First, that the low values in your prints will look "flat" unless you take care to modify them by cutting off low values with the curves tool or by making a contrast mask or using unsharp masking.
The expanded dynamic range of a good sensor does not expand the brightness range of your printer, of photographic paper, or of your monitor. Thus, when a camera such as the S3 is able to record an image brightness range that is 2 additional stops, it means that more values must be squeezed into the same print scale: the print looks flatter, not snappier.
The S3 compensates for this by keeping the mid-range contrast about the same, and compressing the top values (note the "shoulder" of its characteristic curve in the top graph).
Second, all these extra stops in the "toe" of the curve are needed to accommodate the higher ISO numbers the camera can produce. ISO is based on the lowest image luminances that produce a detectable image on film. The analogy is the Zone I exposure value. Thus:
- Zone I is the "threshold" value for ISO 100;
- Zone O is the "threshold" value for ISO 200;
- Zone -I is the "threshold" value for ISO 400;
- Zone -II is the "threshold" value for ISO 800;
- Zone -III is the "threshold" value for ISO 1600.
You can see by inspecting this graph that the A2's highest useful ISO is likely to be about 400, and the S3's highest useful ISO is likely to be about 1600. And this is true. The A2 offers ISO 800, but it's useless, as the lowest values are too noisy; the S3 offers ISO 1600 and it's a little noisy but very usable.

This graph shows that the differences between the Wide and Standard dynamic range settings are to gain one stop of highlight texture and one stop of shadow texture. That is, the dynamic range is extended two stops. The necessary side effect of this is to produce a lower gamma (contrast) in the mid-range values. Unsharp masking or contrast masking can compensate for this, but this does require post-processing time.
Note that this graph shows that the higher ISO setting does not make the sensor more "sensitive." The ISO setting simply instructs the image processor to translate pixel intensities to higher image values. Note that the ISO 1600 curves saturate 3 to 4 stops before the ISO 100 curves, and that the ISO 1600 curves lighten shadow values about 4 stops (as they should! -- ISO 1600 is four stops faster than ISO 100).
Thus, when you shoot at higher ISO values, you throw away the high values and have a constricted dynamic range. If you want to preserve the whole dynamic range of a contrasty scene, you must shoot at the camera's lowest available ISO and carefully place the important bright values below the pixel saturation threshold (by using the camera's histogram or it's pixel-saturation warning).
Then, in processing you can choose what's worth keeping and what can be obscured in highlight or shadow.

Let's look at the characteristic curve again for one more lesson:

What's the best way to calculate exposure with a digital camera?

First, remember that exposure calculations are standardized on middle gray, and that the algorithms of these cameras place this middle tone quite near the pixel-saturation value. (They do this to keep the bright values "snappy" but the cost is muddy shadows.)

Second, remember that pixel saturation and loss of image texture occur in these digital cameras three stops (A2 or S3 std. dynamic range) or four stops (S3 Wide DR) above middle gray.

Third, at the lower ISO settings, there's a long "toe" that preserves detail. This is a safety reserve that you can use.

These things mean that you can "underexpose" images by one or two stops (three if the brightness range of the subject is small) in order to guarantee against highlights blocking up. The price of doing so is more work in post-processing; all your images may print too dark unless manipulated. The advantage is that it's much easier to create a terrific print if you leave yourself room on both ends of the characteristic curve.

Text and images, except for the photo of the CCD sensor, Copyright 2005, Daniel L. Johnson. All rights reserved. Permission is hereby given for non-commercial use of these words and images as long as they are distributed without charge and without alteration.