Saturday, October 22, 2011

Fuji X100 – Review Part 4 – D7000 comparison Round 3 – Exposure differences, why?

In the previous review part, I ran into some truly confusing exposure differences. It is clear that the Fuji’s sensor is sensitive up to 1600 ISO and no further. What this really means, I believe, is that Fuji has analog amplification of the photo count up to 1600 ISO and after that it is all digital tricks.

This becomes very apparent to RAW shooters, but is well hidden from JPEG shooters. So let’s start this more formal examination of the issue with this group of images that I shot and reshot until I got what I wanted, which is (in order of appearance here) three RAW images from the D7000 at 6400, 3200 and 1600 ISO, followed by three RAW images form the X100 and then three JPEGs from the X100.

image

What you see here is that the three D7000 RAW images and the three X100 JPEG images are all looking identical in exposure. That’s the first three and last three respectively.

But the three X100 RAW images who three different exposures approximately 1 stop apart. Note that I still have not gone onto Google to find out what is known about this issue … I want to explore it further before cheating :-)

So something is up inside that camera at ISOs above 1600. All three sets of images were shot at –0.33EV, a hold over from the previous test, but since they are identical, they mean nothing. Aperture priority mode is used with both as well. We continue …

Camera ISO Aperture Shutter
D7000 1600 f/5.6 1/400
X100 RAW 1600 f/5.6 1/180
X100 JPEG 1600 f/5.6 1/80
D7000 3200 f/5.6 1/800
X100 RAW 3200 f/5.6 1/340
X100 JPEG 3200 f/5.6 1/170
D7000 6400 f/5.6 1/1600
X100 RAW 6400 f/5.6 1/680
X100 JPEG 6400 f/5.6 1/340

Hmmm … forgive me, but those number look like nonsense to me. The two sets of RAW images were shot really close together with the JPEGs shot about 15 minutes later, so I’ll discount the jpegs as an issue when we looks at exposure differences. If the sun went behind clouds (as the images appear to show), it is easy to imagine a 1 stop difference from the X100 RAW images.

Yes, I should have shot RAW+JPEG to eliminate the difference in exposures, but the point towards which I am heading is perfectly well backed by these shots anyway.

But here we see the Fuji shooting at 1/180s when the Nikon shoots at 1/400s and the exposures are reversed from what you expect. Since the sensitivity and aperture are fixed, the shutter being slower by more than a stop should be showing us an exposure that is in fact brighter by one stop. What we see instead, is an exposure that is clearly badly underexposed … by one or more stops. Which means that, in total the X100 exposure is really off by two+ stops.

And since the sensitivity as measured at 1600 by DXOMark is essentially the same as it is for the D7000, what we see here makes no sense whatsoever.

Looking at 100% crops of the three JPEGs, we see that there is definitely a difference here in image quality, even if the exposures look absolutely identical when viewed together in CS5.

The shadows get progressively noisier and the details get progressively more smeared. So there is not much doubt that the X100 respects the exposures chosen.

But what about the RAWs? They look weird as thumbnails, so do they process as weird as they look? The answer is yes … they are very hard to get to a similar exposure with the 6400 ISO version already there and the 1600 ISO version way, way off. This is the reverse of what makes sense given what we know.

Well, the RAWs looks a little less processed, at least as far as the mottling is concerned. They are a tad cooler, and the shadows are not quite as open. But they are better in the noise / detail area and that is to be expected. However, I had to push the 1600 ISO image by 2 stops to make it match the 6400 ISO and the 3200 ISO image required about 1 stop of push. I cannot fathom why it seems to be reversed from what I would expect.

Perhaps it is time to have a peek at the D7000 RAW files. They turned out to require no special handling and the behaved perfectly well in ACR. The 6400 ISO image was pretty noisy, but I did what I did with all of the RAW files, I used the various sliders in the NR panel of ACR to get a balanced result. As little noise as I could achieve without smearing becoming too obvious.

The D7000 versions are much more open … I could have increased contrast quite a bit and the noise would have dropped considerably. Oh well … that’s what happens when each set of images is processed separately.

I’ll have to do a formal ISO ladder to try to untangle what the X100 is doing. I find these results again equivocal. So far I think I would be a confirmed RAW shooter with the X100 and I would probably be happy to shoot high ISOs above 1600 … ACR seems to handle them well enough. The noise is quite a bit higher, but that is true of the D7000 as well, and it is about as good as it gets these days.

So what does the DXOMark measurement mean? I suppose it means that the X100 performs a classic trick for 3200 and above such that it sets an exposure for 3200 or 6400 ISO, which obviously must underexpose the sensor since DXOMark measures no extra sensitivity at those settings. Clearly, Fuji then pulls up the data digitally. Which works well enough, just as it does with Nikon’s LOW and HI settings. The curious thing is that Fuji chooses to hide these peculiarities.

Anyway, I will do some more exploration of this phenomenon through ISO ladders later on. I would say that practically, the X100 is pretty good at high ISO and will shoot about as well as the D7000. Of course, the formal ISO ladders might say something else, but for now I have no data that says otherwise.

p.s. For those who are wondering what the subject of these images is, it is a bedsheet that covers a brand new window in my bedroom. The window was replaced when the window and patio door in the kitchen were replaced. And here I ripped out an old set of horizontal blinds and plan to replace them with curtains, once the renovation of the bottom floor has come to a conclusion. Of course, that only means that the top floor comes under scrutiny as there is new carpeting awaiting installation …

p.p.s. For those who obsessively want to pick a winner in this comparison, the Fuji can have the nod. I think I could create a tie for the D7000 with a reprocess of the RAWs, whereas the Fujis are definitely as good as they will get. But the Fujis are good enough that I won’t bother to steal their thunder. The camera continues to impress, even if its RAW files at high ISO are rather inscrutable.