First, DXo does an ISO adjustment that is total garbage for th purpose at hand: it "corrects" ISO speeds on the basis of the amount of high-light head-room and then shifts SNR and DR graphs horizontally to "correct".
BJL,
I can't believe you are so aggressive towards DXOMark. I hope this is not because you are a 'fanboy' of the micro-4/3rds system and cannot tolerate any criticism.
Is correcting ISO sensitivity on the basis of the amount of high-light headroom available not consistent with the principle of ETTR? If there's been one topic that has cropped up on this forum almost ad nauseum in the past, it's the principle of 'expose to the right'. If one doesn't fully utilise the highlight headroom, by using a shutter speed that is sufficiently slow, but not so slow that highlights are blown, then image quality will suffer, whether it be shadows that are slightly noisier than they otherwise would be, or mid-tones that are not quite a smooth as they otherwise would be.
But to compare performance at a given light level, the only normalization that should be applied is using equal shutter speed and equal f-stop to get the data used for a given ISO setting.
If the ISO sensitivities of two cameras are different, it's not possible to compare ETTR exposures from both cameras at equal shutter speed, equal f-stop
and equal manufacturer-nominated ISOs. Either a different ISO setting has to be made to one of the cameras, or one of the cameras will over-expose or underexpose the image.
Now it's true that the DXO figures for the qualities of DR, noise at 18% grey, tonal range etc at the
manufacturer-specified ISOs for these two cameras are actually
better for the GH2 (with the exception of ISO 800 where DR still has the edge for the GH1).
If one were to align the red dots vertically as you suggest, according to manufacturer-specified ISO, the dots for the GH2 would mostly appear slightly above the dots for the GH1, indicating marginally better performance at the
incorrect manufacturer-specified ISOs.
However, in my view, the purpose of such graphs is to describe performance at the
correct ISO values, not the
incorrect values. If one wishes to compare image quality at equal shutter speed and F-stop, then the GH2 should be set at a
higher ISO, either 1/3rd stop or 1/2 stop higher. Alternatively the GH1 should be set at a
lower ISO, either 1/3rd or 1/2 stop lower.
What the ISO increments actually are on the GH1 & GH2 is another issue. Most cameras allow ISO settings in increments of 1/3rd EV, but not all cameras allow for
real 'intermediate' ISO settings. My 5D apparently allows for real ISO settings of ISO 1000 & 1250 etc, but some cameras have real ISO settings only in increments of 1 EV. The 1/3rd stop increments in between may be merely sofware extrapolations at the end of the in-camera processing chain.
If this is the case with the GH1 and GH2 models, ie. no real intermediate ISO settings, then in situations where one reduces the ISO of the GH1 to get equal ETTR shots with the GH2, at the same shutter speed and aperture, one will in effect be comparing (for example) an underexposed image at ISO 800 on the GH1, which is extrapolated in software to ETTR status simulating ISO 1250, with a correctly exposed ETTR image on the GH2 at a true ISO of 1600.
Likewise, if I'm comparing a shot taken with the GH1 at ISO 1600 with the same scene taken with the GH2 at equal shutter speed and F-stop for an ETTR exposure, I will have to increase the ISO of the GH2 to the intermediate ISO of 2000 (or perhaps higher if 1/2 stop increments are available). I will in effect be comparing a slightly underexposed image from the GH2 with a correctly exposed image from the GH1.
If the intermediate ISO settings are actually
real settings derived from appropriate amounts of analog gain of the signal in the early stages of signal processing, then the line on the DXO graphs joining the dots should give one a fair indication of performance at the intermediate ISO settings.
In all situations for still images, the GH1
appears to have the image-quality edge, according to DXOMark results. But that edge appears so
slight in my view, that I think for all practical purposes image quality will be about the same. If there is any noticeable difference, I would expect it to be a slight increase in resolution and detail from GH2 at large print sizes. However, the full HD video of 1080p at 24fps, and faster processing are amongst the main improvements that make the GH2 a worthwhile upgrade.
I hope all this is at least clearer than mud, although I expect for many it won't be.
I see that Michael is a bit mystified as to why the DXOMark results indicate the GH2 has no fundamental image quality improvement over the GH1.
Perhaps this could be a test case to explore the reasons for any discrepancies between rigorously tested results and visual impressions that appear to be in conflict with such results.
It might be worth noting some explanations from the DXOMark website on their results.
All Sensor Scores reflect only the RAW sensor performance of a camera body. All measurements are performed on the RAW image file BEFORE demosaicing or other processing prior to final image delivery. DxOMark does not address such other important criteria as image signal processing, mechanical robustness, ease of use, flexibility, optics quality, value for money, etc. While RAW sensor performance is critically important, it is not the only factor that should be taken into consideration when choosing a digital camera
Sensor Overall Score is logarithmic and a 5-point difference on the scale corresponds to a gain or loss of sensitivity of 1/3 of a stop.
Sensor Overall Score is normalized for a defined printing scenario—8Mpix printed on 8”x12” (20cmx30cm) at 300dpi resolution. Any other normalization, even with higher resolution, would lead to the same ranking, given that any camera that could not deliver the chosen resolution would be eliminated from the comparison.
This last point clarifies something that I was not sure about. An 8mp image seems a rather small size for normalisation of a 60mp MFDB image and a 24mp D3X image, but the message here is that the ranking would be the same for any normalised size (in respect of the P65+ and D3X) as long as it were not larger than 24mp.