@whiterabbit
No correction, i just changed the way the image is rendered (in fact this is realtime). No loss in quality, you just squeeze all the latitude of the coded picture in a flatter container.
This is SiLog, not a much extreme curve, with Rspace space you get an even flatter picture.
PS: I don't know if we are offtopic, i think that this is just a useful part of the conversation.

@kholi, it is ok, I did not take it bad at all. As you can see I posted something about the same time as yours. I was just talking that people have the right to be disappointed, if one parameter or the other is not as what they thought. For me DR was the only thing that I could not quantify with the gh2/hacked. Because I knew it was sharper, nearly no moire/aliasing to degrade the image etc but the DR no comparison or test. I always find people bashing it on forum saying that it has lower DR etc without any test/comparison/footage to back it up. On the net suffice to say something 100 times and it becomes reality. At least now we have at least some... ok, if not far from perfect comparison.

Unfortunately I don't have any pro camera like the EPIC to use and compare to the gh2 personally and found a personal opinion. Which make these kind of test valuable to me. If you read my post above, you will see that if you look and test (even on so lowly compressed format) the gh2/hacked compares much more favorably in terms of DR to most of those cameras.

Now for your question about DR and ISO, normally it is true that as higher you go the DR will decrease. But the gh2 is benefiting from true downscaling/binning which as with noise tend to make things better. If you look at testing site like DXOmark you will see two DR value, the native DR and the one they have standardize of a 8 megapixel camera. Noise is the same, that is why when you use the ETC mode the image is noisier, because it is just cropping a 1080p at 1.1 pixel level of the sensor and you are not getting the benefit of downscaling/binning. Everything is relative, the DR will decrease but not as the same rate than in photo mode.

The best part of this test is that you and me don't have to pay a single cents to watch it. PB decides to do it with his own time, money, contacts and effort to keep everybody happy, informed and entertained. WTF are we complaining about?

@Elenion, any chance you would be so kind as to explain your Cineform curve process a little further? Much appreciated, thank you.

hahahah @kazuo thats becuase you are thinking only from your reality, at least i have no real possibility to test bigger cams

@itimjim i see no difference between bias or not very well executed, as it was a subjective test, just by setting the cameras one way or another by eye immpression, you are getting a bias,

if there was or not an economic bias, well, that statement can be as it cant be, i dont care really, as there is no way i can know for sure, so his shootout will remain for me as his shootout, nothing more, a valuable opinion, but not an absolute truth or whatever...

@Vitaliy_Kiselev of course, there is better knowledge in books, but this way is more graphic and easy to digest in a fast way, all i´m sayng is that it is just a blog post, and an interesting one, it is not a scientfic study by any means, its just an article with interesting (bias) knowledge available, that is worth seeing

haha at least from my side, i was having fun, trying to figure it out wich camera was used on each frame

@kholi somwhere in here http://www.personal-view.com/talks/discussion/comment/30852#Comment_30852 you will find a dpreview of the jpeg output of the gh2... but still you can notice it at naked eye on the lcd screen, just record the same scene, one at iso160 and the other to iso 12800, you will notice the difference

@lolo

Has anyone ever actually tested this "higher ISO lower DR" theory? I'd LOVE to see that, because it sounds backward to me.

Low ISO means banding and posterization of brighter areas, and a quicker clip to white. The only thing that should see the maximum amount of range the camera has to offer is knowing what ISO it's actually optimized for, and what film mode setting.

1 5D
2 NEX5
3 AF100
4 C300
5 D7000
6 F3 AB
7 F3 S-Log
8 FS100
9 GH2
10 7D

keep in mind that the dynamic range depends also of the iso used, higher iso means lower dynamic range,

and about how to get and improvement in dynamic range, i do it in post, if the camera was in a tripod i shoot two takes (lets say that i´m in a room with a window), first i shoot exposing for the window outside, with no subject at all, and then i make other takes where i expose for the subject indoor, later in post i mix the two images

What is point to listen what some Philip do ? Just go and shoot with your gh2 or whatever camara you have.
If result is good everything is fine. If Olivia make a video about cams I will Hmm watch.

In fact the picture at first site is a quite deceiving. I have downloaded the images and doing some test ... I know they are re-sized lower resolution and jpeg compression. Even then the gh2 is noticeably sharper and much much more detailed. In fact were the majority of the other ones are mushing the shadows the gh2 keeps a lot more, that can be boosted. I am quite surprised/amazed how this bad quality compressed shot stand up to correction in post, at least compared to the others. The histogram also tells another story, if you look at most they clip at different level (upper) while the gh2 does not. In fact the gh2 is one of the best (at histogram level) and seems a tad underexposed compared to most of the dslr. You can easily simulate the flat look of the other camera just by lowering your white and raising your white point in photoshop curve. (For copyright reason I think it is not good to post it here).

Hopefully he will upload some uncompressed screen grabs of these for some better judgement. There are too many variable to conclude about the true DR of these camera. The model is sometime moving and if the light source is near it can have and impact on the lighting and DR (White t-shirt could blow for example). Philip Bloom has done this test for skin tone and not DR.

Hey guys.. we already know the DR of the gh2 is not huge.
So getting dissapointed from heavily compressed JPEGs from shots where the lighting does not look favourable to the camera itself and with unknown white-balance is a bit wierd. My guess is he shot on the same K instead of balancing each camera individually according to their color profile. Then we also have the issues of how all our computers handle a specific color profile and how they are displayed..

For this kind of shot I would try to light as to be able to expose to the right on the gh2, which IMO would yield a similar result as in image 2 (as I see it). That said, someone might think another way of doing it gives better results. For the reason of the test the lighting might be done as it is - to display the DR of the camera, rather than accomodate for the DR of the camera.