Why is DXO taking so long on the EM5?

Started May 4, 2012 | Discussions thread
Anders W Forum Pro • Posts: 21,466
Re: Do you have a tribal juju rattle ? - Dr DxOMark, We Presume ???

Detail Man wrote:

Yes, that is correct. What I did do ("scaling ex post") does not conform with the increase in the EM5's Radiometric Exposure that you state would have made sense as a method to modify the EM5's "saturation level in terms of ADU units". I do not see doing such a thing as a valid approach.

What you personally consider a valid approach or not is beside the point here. In this post of yours


you pointed to the seemingly contradictory results obtained from a) a comparison between the E-M5 and the GH2 based on my DR estimates and b) a comparison between the two cameras based on pushing the deep shadows of RicksAstro's sample shots. Now the first of these two comparisons rests on the premise that the sensors are equally saturated. This is also the premise on which any comparison based on DxOMark's "landscape scores" (i.e., the max DR of the cameras in question) rests. The second comparison, by contrast, rests on the premise of equal exposure.

In my response to the post to which I linked above, I pointed out that the fact that the two comparisons rest on different premises make them less than fully comparable:

1. The comparison you make between the "deep-shadow" performance of the two cameras if not fully analogous to the comparison between the two DR estimates quoted above. The latter refers to the max DR of which the cameras are capable. The former refers to their performance when given equal exposure at their respective base ISO. In order to make the two comparisons fully analogous, the E-M5 should, in your case, have been given about 1/3 EV more exposure than the GH2 so as to reach the same saturation level in terms of ADU units.

Do you still have any problems understanding this response of mine?

As to the premises themselves, I find both of them acceptable. However, for comparisons at base ISO, I personally find tests based on the premise of equal saturation more interesting. The reason is that when I am shooting at base ISO and encounter a scene with high dynamic range, I am usually in a position to give the camera as much exposure as it needs to do its very best. For comparisons at higher ISOs, by contrast, I prefer comparisons based on the premise of equal exposure. The reason is that when using such ISOs, I am not in a position to give the camera as much exposure as it needs to do its very best (or I would be shooting at base ISO).

The equal Radiometric Exposure that RicksAstro chose was appropriate so as to be able to meaningfully compare the Signal/Noise Ratios of the EM5 and GH2 where it comes to the contribution of Photon Shot Noise sources (along with the Read/Dark Noise Sources) to the total composite image-noise. Modifying that equality (as you appear to have suggested) would have disadvanted the GH2 relative to the EM5, resulting in a rendering of the test results to be invalid.

See above.

The number of powers of two that RAW-scaling is increased is not so much the issue - how much of the visible image-noise arises out of Read/Dark Noise sources, and how much of the visible image-noise arises out of Photon Shot Noise sources is a relevant issue. The presumption is that we are interested in how much visible image-noise a person engaged in RAW processing will see in deep-shadow areas of image-frames, and will have to somehow deal with - and not so much whether DxOMark will (or will not) publish a numerical speculation that gratifies our senses of analytical satisfaction, or may (or may not) serve to confirm the seeming wisdom of our wishes surrounding the potential purchase of a new camera. Few among us would likely argue with that ?

I am not arguing with that. The point is that noone would, for purposes other than testing, want to push areas like the ones you selected (with original ADU-levels in the darkest portions on the order of a single unit) by as much as five stops. For this reason, the outcome of your test isn't very interesting either. It focuses on performance characteristics outside the range in which I (and presumably others) would want to use either camera. Both look awful when the very deepest shadows are pushed five or so stops.

It is true that the Radiometric Exposure was matched in all of his image-samples involved in the shadow-noise as well as the highlight-preservation comparisons.

In his shadow-noise comparison, my own eyes do also see more visible image-noise present in the lower-right area of the image. An interesting additional observation is that the GH2 seems to retain more detail on the srface of the girder in the left-hand portion of the image-frame, and also seems to exhibit a higher amount of contrast over the entire image-frame. All of the above observations may to some extent be attributed to characteristics involving the LR 4.x processing itself, however.

What you are seeing here is most likely the result of the different bottom-end clipping strategies used on the two cameras. The black-level offset used on the GH2 places the clipping point slightly higher than that used on the E-M5. On the GH2 the latent noise distribution in a black-frame shot (i.e., one containing read noise only) is clipped to the right of the mean. On the E-M5 it is clipped more or less exactly at the mean.

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +28 more
Post (hide subjects) Posted by
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow