GH3 = attempt at Photographic Dynamic Range

I think I probably did something wrong, however - here is an average of 3 series of measurements :

note that this is Panasonic and nominal ISO400 shall be ISO400 (in DxO measured sense), not ISO200 (in DxO measured sense for Olympus)

Since ISO 25600 is extended there is no way you can get the same DR for ISO 12800 as for 25600. DR at 25600 will be about 1 stop less. If your result is different you have severe error in methodology.

And I suspect ISO above 3200 will already be extended so with proper measurement you should get a chart similar to DxO chart for E-PM2. It's the same sensor after all.

BTW, according to manual you posted, GH3 has no real ISO 125, it's overexposed ISO 200, just like Canon has "ISO 50" on 5DMkII. So, unlike your chart shows, ISO 125 should have the same DR as ISO 200.

Bottom line - I agree with you, you probably did something wrong.

we shall see, I sent the data to William Claff (the author of the program)...

Except the nominal ISO for Olympus cameras are not the same as nominal ISO's for Panasonic cameras...

may be not... the wording "extended" in manual for ISO125 not necessarily means the same thing as in Canon's case...  Olympus nominal ISO200 DxOMark measures as ISO106, if Panasonic stays true to their approach then their ISO200 will be measured around ISO200 and hence ISO125 might as well be ISO125 and not overexposed ISO200.

the raw files are here = http://sdrv.ms/TOdeKB
5 series

everybody is welcome to find what might be (if) wrong

If ISO 125 was real why would Panasonic call it extended? And, as I understand, extended ISO needs to be enabled first. I can see no reason for that if ISO 125 was native.

It is true that Olympus rated ISO differently from Panasonic in the past. But now there is a new sensor, so we don't know. A french site (photo numerique?) posted RAW files from E-M5 and GH3 and IMO they were exposed similarly at the same settings. So, this might have changed, although there is not enough evidence.

you have the link to my raw files - feel free to find an error in how they were shot... I really need a fresh input where is the error if indeed there is an error on my side in how I was shooting.

there is no need for IMO - you just need to compare raw histograms to see the situation.

"IMO" is because I can't be sure if they used the same lighting conditions and lenses were different.
About your files - maybe DM will be curious enough. I can wait for DxO report. I know GH3 will be close to E-PM2.
BTW, if this info is not there DM will ask you about saturation level and minimum value level in RAW files.

and what was the point to make extended ISO = ISO125 and not ISO100 then (like exactly a stop below ISO200) ? so it is probably a combination...

Are you saying that the GH3 DR at ISO6400 is just under 4 stops? If so dxomark rate the OMD at 7.5 stops @ ISO6400. I am surprised to say the least. I would have thought the GH3 would be about the same as the OMD.

The Olympus E620 according to dxomark shows better DR than the GH3 across the ISO range.

--
Christos

They probably couldn't tweak JPEG engine enough to get rid of blowing highlights at full stop overexposure. Manufacturer's ISO is measured by JPEGs anyway. Not sure. Time will tell.

it is "Photographic Dynamic Range" as used by William Claff, the author of the program that was used for calculations = http://home.comcast.net/~NikonD70/Charts/PDR_Landscape_scatter.htm

Yeah, who knows the nature of their ISO=125 setting ? Not I. Am pretty sure that Bill Claff uses a SNR of 26 dB as a reference in PDR measurements. For a EM5 (rated ISO=200) Sat ISO=107 here:

http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Olympus/OM-D-E-M5

... Bill Claff's site shows the result of testing Anders' EM5 unit as PDR=9.27 EV here:

http://home.comcast.net/~NikonD70/Charts/PDR.htm#OM-D%20E-M5,Ideal_4/3

If we assume that GH3 (rated ISO=125) Sat ISO is around 125, that value is not far off from the EM5 Sat ISO=107, and your results suggests that the GH3 PDR falls short of the EM5 PDR by 0.35 EV.

That would translate to a numerically larger margin (in difference in EV of DR) if the measurement used a SNR of 0 dB (as in DxOmark testing). Nothing that you probably do not already know ...

I am not through looking at your files yet. But based on what I guessed before seeing the RAW data as well as what I have seen so far, I think the reason why the DR measurement fails (gives unreasonable results) for the last two ISOs is that these two are, in RAW-level terms, identical to ISO 6400. In other words, the signal is neither amplified before AD-conversion nor digitally scaled afterwards. Presumably, the only difference compared to ISO 6400 is that the camera will adjust the metering and the OOC jpeg gain plus leave an instruction in the RAW file to the RAW converter to add one or two stops extra to the default jpeg gain. No different from just pushing one or two stops in post.

Other than that, things look very similar to the curve for the E-M5, including the characteristic bow at the lower end of the ISO range, reflecting the strong reduction in read noise (as measured in electrons) between the first two ISO-points.

yes, I saw that in rawdigger, something like that... I even did 2 more series specifically for high ISOs only... still the same.
so Panasonic should be more honest then and designate ISO12800 as extended as well like Hi-1 (ISO12800) and Hi-2 (ISO25600)...

If you want more input for the purpose of diagnosing how things actually work, it would be helpful if you could upload two more series:

1. Black-frame shots, one at each major ISO. Use highest shutter speed.

2. A series where you shoot a target other than your computer screen (a white wall or something like that) at each major ISO using the same aperture and shutter speed all the way. Make sure the chosen aperture and speed are such that there will be no raw-file clipping at any ISO. The reason for this request is that it's a bit tricky to judge how the camera reacts to ISO increases in shots of a computer screen like those you posted, especially since the shutter speeds used are higher than the screen refresh rate.

stay tuned... uploading

black frame shots from GH3 uploaded to http://sdrv.ms/Sy1aP8

Hmmm. Looked at the ISO=125 and ISO=200 RAWs. RawDigger reports no black-level subtraction being used whatsoever(?), and shows a very clean noise distribution (all in the zero bin, not a thing in any of the higher bins). The Standard Deviation is (in both cases) somewhat higher than that of the EM5 (at base Sat ISO=107) and the GH2 (at base Sat ISO=167).

(Manufacturer rated) ISO=125 ~ 1.3 ADU

(Manufacturer rated) ISO=200 ~ 1.1 ADU

rawdigger does not have the most recent libraw version incorporated hence GH3 is formally not supported... libraw beta with GH3 formal support was announce only like couple of days ago... so I 'd assume you have to enter black level manually