@vicharris

There is a huge difference between shooting a commercial with a different camera than what you are advertising and showing stills that you imply have come from one camera but they are actually from another camera. One is just marketing. The other is fraud.

it degrades the precision of the encoded color data, particularly in near-black shadow details.

And that quantisation starts in RAW even before the encoder gets hold of it but the encoder really does play favourites with the data and low luma loses out big time.

I am doing some examination of frame grabs in Avisynth using the histogram("luma") function to show how much high luma detail is being favoured over low luma detail by different patches. One thing I am noticing straight away is how much the GH2 AVCHD encoder puts all the data into highlights and abandons low luminance sections of the image. I think this makes sense when you have bitrates around ~20mbps because perceptually you will notice low luma areas less but now that we have the hacks and very high bitrates it is very annoying to see the encoder still throwing so much data into highlight detail that you can't even resolve while still starving out low luma regions.

@Rafa I think the choices Panansonic engineers made in the GH2 AVCHD encoder targeting high compression techniques to achieve nice looking low bitrate modes actually probably gutted its potential for higher IQ at high bitrates(with difficult material). I don't think it is a 8bit/10bit thing at all. Bluray movies look pretty damn good and they are 8bit 4:2:0. You don't need 10bit to achieve high IQ you just need a good encoder. I do blame the encoder because 170mbps of h264 should produce an immaculate picture everytime and it would if the patch makers had more access to the encoding engine.

Hopefully the encoder in the GH3 has been radically upgraded and we can haz access.

Time will tell.

@LPowell Great exposition on the problems of using flat profiles when you have just an 8b code. From 10b up makes sense, but with 8b in fact you are losing DR.

72mbps is nice, but can it do 172? If it could I'd buy it.

At this price range i'll accept any increase in DR that they can give. The GH3 looks like it's already going to give us so much more as a device that i'm not going to be too harsh on them if they don't give us a perfect camera. We're getting so greedy now when it comes to cameras and really after a slow start all of a sudden the cameras are coming fast and heavy with new improvements. Seems like Nikon, Sony and Pany are all in on this market.

@Astro

Hehehe :)

re: the canyon, what I think danyyyel was referrring to was the still where highlights were intentionally let go; the shot was exposed pretty brightly. IN my opinion that shot showed some promising highlight roll-off... But I won´t analyze this any more now. Time to let it go and wait.

@kholi are you sure? I mean, how much would you bet on it?

@svart For me DR was always the missing link. That is why I did a test on the gh2 vs 7d DR below.

http://www.personal-view.com/talks/discussion/1795/dynamic-range-test-between-gh2-and-7d/p2

For me everything has to reach a certain threshold. True 1080p for resolution, very good 800 Iso to good 1600 iso for low light and at least 12 stop of DR. Unfortunately until now, you see people just focusing on resolution and low light. If a camera is not 4k or is not noise less above 1600 iso they go mental. What I am happy now is that general Joe is seeing the importance of DR (thus putting pressure on manufacturers). Perhaps because it was the last missing link after resolution, lSO, sensor size to achieve the film look. For me (photographer) DR is like matter, the more you have it (untill a certain level), the more your files have that richness (difficult for my English to explain).

Now everything has a threshold, once you reach it, it is just marginal gain. When the day we reach about 14 stop I think it will be a threshold (12 is still very good). Using a d800, I would call it near human vision and over and above it would start to get cumbersome because the image will absolutely need a lot of grading.

The only thing that really annoys me about the music is that I can´t remember who it is... I know, but can´t remember the name. Aargh!

Strike that: I solved it!

Im pretty certain it´s imogen heap:

Nm

Yeah, they could never set that shot up again with another camera. NM.

@svart @kholi

Hollywood lighting techniques were perfected in the film era and are optimized to make the most of the physical properties of film media. Digital image sensors differ from film in at least two important ways:

• Film responds logarithmically to light as does the human eye; image sensors respond linearly.

• Film preserves color data in analog detail; image sensors are digitized with coarse quantization.

The RGBG photocells in the GH2's image sensor are digitized to 12-bit precision in the camera's RAW still photo files. As with all digital sensors, this linear 12-bit range must be scaled by a logarithmic gamma curve to make the results look natural to the human eye. Image editors that work directly with RAW color data allow you to manipulate the dynamic range and color balance of the image in its original linear data format, before it's converted into a lower-precision format with a gamma curve that suits the human eye. This can produce image quality that rivals analog film.

With JPEG and AVCHD encoding, the GH2's 12-bit RAW data is remapped by the Film Mode's gamma curve and compressed down to 8-bit 4:2:0 color depth. While this does not reduce the camera's dynamic range, it degrades the precision of the encoded color data, particularly in near-black shadow details. Image editors that work with 8-bit color data are limited in the extent they can manipulate the dynamic range and color balance of the image without producing unsightly banding and digital quantization effects.

The "flat" gamma profiles that are popular with high-end cinema cameras produce high-quality results only so long as the image sensor's color data is preserved at near-RAW precision levels. If you attempt to mimic these low-contrast profiles on an 8-bit 4:2:0 camera like the GH2, its low-precision quantization will produce lower quality results than profiles that make use of the full dynamic range of the image sensor. That is the brutal reality of 8-bit color spaces.

There is one case, however, where reducing the dynamic range of a GH2 video can improve image quality. If you use fill lights to illuminate the shadows, you can boost the level of the darkest details above the coarsely quantized bottom end of the histogram. That will enable the AVCHD encoder to digitize shadow details with far better precision, and post-production grading will reveal much less noise and banding in the shadows. Using fill lighting in this way mimics the ability of cinema cameras to lift the pedestal of their gamma curves and flatten the DR of the recording. Since the GH2 lacks a pedestal adjustment, fill lighting is the best way to achieve this effect.

@Cde My last comment was more directed towards the thoughts that no company would do such a thing. That's just plain silly to think that way. I've seen footage shot on a red and messed with in post to look like a less cameras footage. You never know what happens when they film these things. OS not 100%, camera failure. I've seen it. We'll see when the finised version comes out. I've seen many industry professionals fooled by this. Not saying it's cool, just saying.