DxOMark - The natives are becoming restless ...

Started Jun 16, 2012 | Discussions thread
Anders W Forum Pro • Posts: 21,466
Re: DxOMark - The natives are becoming restless ...

bobn2 wrote:

Anders W wrote:

bobn2 wrote:

Anders W wrote:

bobn2 wrote:

Steen Bay wrote:

MAubrey wrote:

Yes. You'd better watch what you are drinking. Bold statements like the E-M5 beating the 5D3 for base ISO DR. And beating the D800 for high ISO shadow noise whenever DoF is in short supply. Interesting times, aren't they?

What does that have to do with the question at hand? I resent that you even bring that up here since I definitely said nothing like that. Don't put my observation in the same camp as that kind of nonsense.

It's very likely that DxO's test of the E-M5 will show that its base ISO DR is higher than the 11.7 Evs/stops (print/8mp) the 5D3 has,


and also likely that the E-M5 has more DR than the D800 at so-called 'equivalent settings' (same DoF and shutterspeed, and two stops higher ISO on the D800).

It will be close. The read noise of the D800 is pretty good but the QE is very good.

Lets assume that the E-M5 is performing with the best of Panasonic sensors at a pixel level. That will give it a QE in the mid 40%'s and a read noise of around 3 e- per pixel. The D800 is operating in the mid 50%'s and a read noise about the same. That will give it a read noise of about 4.5 e- per 16MP pixel. So at the same photon count per equivalent 16MP pixel it will gain 1.2x at the top end but lose 1.5x at the bottom, so a DR of 1.25 times or 0.3 stop.

I am pretty sure that the E-M5 performs better than the best (known) Panasonic sensors at the pixel level, which explains why things look the way they do for example here:


The way things look there almost exactly bears out what I said above. The D800's shadow noise is more than a stop better but not two stops better, and its bright to mid tone noise is a bit more than two stops better - which equates to the D800 having better QE but less read noise per area (about the same per pixel).

OK. So we agree that the difference in shadow noise is only about one EV. As to the difference with regard to bright- to midtone noise, could you please illustrate what the more than two stop difference you are talking about actually looks like. Some perceptual evidence please.

Look at your own example, the white patches. The E-M5 noise at 1600 ISO is greater to my eyes than the D800 at 6400 ISO. But of course the trouble with perceptual evidence is that it is very prone to experimenter expectancy effect, so you are likely to disagree with my subjective assessment. I'd like to think mine is objective, because I don't have a horse in the race.

The white patches all look just fine to me. Essentially no perceptible difference, which is precisely my point. While it may well be that there is a two-stop measureable difference, the perceptible difference is virtually nil.

As to horses in the race, neither of us have one. I tend to adapt my decisions to the facts, not the other way around. And I made the same points long before I actually ordered an E-M5.

Fully consistent with best of Panasonic performance as Jim and DM have been saying.

Which Panasonic sensors are you comparing with and at what ISO?

Say the GX1 sensor which has a QE of 44% and a read noise of 2.7 e- per pixel, and is up there with say the NEX C3 (45%, 3.3) or the D7000 (48%, 2.5). Where I'm thinking that the E-M5 has improved is with a better ADC to give better base ISO read noise. The GX1 is 11.1 e- to give 10.1 stops DR. If Olympus has managed to make it ISOless then it will gain an extra 11.2/2.7 = 4.11 = 2 stops DR to take it to 12.1 stops DR.

The E-M5 sensor isn't ISO-less. Based on the evidence we have right now, the read noise is 1.0 ADUs at ISO 200 and about 1.7 ADUs at ISO 800 (early measurement by kenw in the latter case). On the other hand, the read-noise levels are considerably below those of the G3/GX1 and GH2 sensors in absolute terms. Depending on exactly how deep the wells turn out to be, we are talking perhaps 3-4 electrons at base ISO and something like 2 electrons at higher ISOs.

That's if the EM-5 has a QE up there with the best of Panasonic. One of the ways that Olympus tackled the visible noise issue in the E-5 was to put in a very strongly cut CFA which cut chroma noise but whacked the QE. If they have done that again, then there won't be any advantage.

Fortunately, they haven't.

Not so much, the chroma noise seems too strong for a tight cut CFA.

Too strong compared to what?

Too strong compared to cameras with tight cut CFA's like the E-5. However, having just had a look at the DPR studio comparison tool looking for examples, I'm beginning to thing there is some raw NR in the E-M5 - there seems to be some spreading of the chroma noise, which is usually a sign. Something else we'll need to wait for is DxO's correlation tests.

No more spreading of chroma noise than on any other sensor that we know (if we trust DxOMark) not to do any raw NR. Besides, Oly has never done this before, although it might have been tempting for them to do it in view of the poor (by today's standard) performance of the old 12 MP sensor. So why would they start now?

 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
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow