Five E-System DSLRs Compared: Dynamic Range in More Detail

Five E-System DSLRs Compared (RAW, Silkypix Developer Studio): Dynamic Range in More Detail

[Back]

Since I do not have the necessary equipment to get a complete EV scale on a single image, I placed a white sheet of paper under a neon lamp with a daylight-like colour temperature, mounted the Zuiko Digital 2/50 Macro and took a series of shots with exposures ranging from 4 seconds at F5.6 to 1/4000 sec at F16 in 1/3 EV steps. This of course has the disadvantage that the resulting plots will include possible errors caused by inaccuracies of the camera's exposure system, be it the shutter or the aperture diaphragm. But, since I always had the impression that the Olympus DSLRs do expose quite precisely throughout the complete shutter speed and F-stop range, I thought it ought to work well enough. To obtain the greyscale value for each measuring point, i.e. for each image, I wrote a small Java program which calculates the mean grey values for JPEG images from all of their pixels, and I processed all resulting images with it. I set the 0 EV origin to where the value was nearest to the medium brightness of 128.

What I did not expect, though, were considerable bumps and bends in the graphs standing out below -3.0 EV. Because the unexpected readings correlate with unexpected visual aspects of the associated images (single images came out darker than expected and looked greenish instead of neutrally grey, although a fixed white balance was used), I suspect the neon lamp to be the cause, perhaps something related to the mains frequency in conjunction with some 'critical' shutter speeds? The second image of these three constitutes such an anomaly (1:1 crops):

I'm inclined to rule out a camera malfunction, because I never saw such a thing in real-life photography, and because it happened with all five cameras. Next time I'll experiment with different light sources, but for now I'll simply point out the fact... and paint the affected leftmost segments of the graphs with a different grey background, setting them apart from the 'good' segments. Anyway, it's highlight range I was mostly interested in, and I did not yet specify a definition for the dynamic range's lower bound either, from which I'm now saved for the time being...

There's nothing to gain in comparing those leftmost segments between graphs, by the way; I marginally tried to smoothen them out, and I was not equally successful on all three graphs, and so they're probably even less comparable than they were before.

Out-of-camera JPEG, camera set for lowest contrast

Now, let's first have a look at the out-of-camera JPEGs. To get the largest dynamic range possible, I chose a contrast setting of -2, which is the lowest possible setting for all cameras:

And we see what was to be expected after reading the major E-510 reviews, some of which published similar graphs: The E-510 JPEG seems to have at least 1/2 to 2/3 EV less highlight range than its predecessors, even when maximizing dynamic range by choosing a low contrast setting. What we also see is that the four other cameras behave nearly identically, from the venerable E-1 to the three years younger E-400. The E-1 deviates only insofar as it yields almost negligibly darker results than the rest of them, although the marginal brightness difference is the reason why the E-1 seemed to have a slight advantage in highlight range by example, part 2 and part 3, while it actually hasn't any. It's just the fact that it's impossible to exactly match exposures of two cameras when they're, say, 1/6 EV apart and the smallest available exposure step is 1/3 EV.

RAW developed with Silkypix Developer Studio, "Average contrast" (Silkypix default)

How compares the RAW developed with Silkypix default settings to the out-of-camera JPEG, when it comes to contrast and dynamic range?

The E-510 now gains about 2/3 EV in highlight range, and the other cameras gain, if anything, maybe 1/6 EV. Moreover, we see that for most parts except for a small central interval between something like -2/3 and +2/3 EV the "average contrast" RAW curve is actually similarly or less steep than the lowest-contrast JPEG curves of all five DSLRs. Or, in other words: What RAW converter makers like Ichikawa Soft Laboratory (who by the way used to create the RAW engines for Olympus, maybe still do) consider to be an "average contrast", is in fact similarly or even less contrasty than what Olympus offers its customers as the least contrasty option for the out-of-camera JPEG. No wonder that some people regard the E-System cameras' default contrast as being too high, and no wonder some people starting off with their new E-System camera are surprised about blown highlights at default settings, and this is to be said for the five of them, to some extent, not only for the E-510.

But, how compares the E-510 to the rest of the gang, now we've switched to RAW and chose Silkypix as our converter software? Except for the E-1's faint overall brightness difference, which is quite the same here as it could be observed in the out-of-camera JPEG, the curves are almost identical now. In other words: Working with Silkypix and its default contrast settings, we not only have a gain in highlight range, we simply cannot distinguish the E-510's contrast behaviour from the other four cameras anymore – and we can no longer observe significant differences in tendency to blow highlights.

RAW developed with Silkypix Developer Studio, "Lowest contrast"

Although it says "lowest contrast", this setting is far from being the lowest possible contrast setting within Silkypix, it is simply the lowest within the predefined options ("tastes" in Silkypix' diction). But it might well approximate the lowest practical contrast setting, as long as we desire neither black nor white to turn grey. Do we learn something new from its curve?

Actually I think we don't. As expected, the overall curve is considerably less steep than the "average contrast" curve, but in order to keep highlights from turning grey, from somewhere around +1 2/3 EV onwards it compensates for it by getting even steeper than "average". With the result that there's maybe another 1/6 EV difference between the absolute upper ends of the contrast range, marking the points where highlights start clipping. For the E-510, this lets the RAW advantage over the out-of-camera JPEG definitely reach, maybe slightly exceed 2/3 EV, while it adds up to around 1/3 EV more highlight range for the other cameras, and thereby to an accordingly reduced danger of blown highlights.

I've tried to word a conclusion on the starting page.

[Top]

RS, 28-Feb-08