GH2 What is the dynamic range of the GH2?
- Thread starter Hutchman
- Start date
See ColorGhear website
The DxOMark website rates the GH2 dynamic range as 11.3 f-stops, but this is for still photos.
GH2 vs Nikon D800
GH2 vs Nikon D800
Also keep in mind that DSLR dynamic range shrinks as the ISO goes up, so a camera might have 10 stops of dynamic range at ISO 200, but this might become 7 stops of dynamic range at ISO 1600.
The new Canon Cxx Cinema cameras ( C100 / C300 / C500 ) are one of the exceptions, where they maintain their 12 stops of dynamic range right up to ISO 20,000.
The new Canon Cxx Cinema cameras ( C100 / C300 / C500 ) are one of the exceptions, where they maintain their 12 stops of dynamic range right up to ISO 20,000.
j1clark@ucsd.edu
Veteran
Long ago, for the GH-1 Barry Green produced some images of grey chart step wedges, and I think for digital filmmaking purposes, 7-8 stops is the effective dynamic range of the GH-1. I don't think GH-2 did significantly different.
It all boils down to the fact that the resulting image is encoded as an 8-bit quantity for each of R, G, B. and so that will only give 8 'doublings'... or stops.
There are tricks to create an 's' curve, that mimics in some way Film film response, but until 10 bits of bit depth are at popular prices, there won't be much differences on the dynamic range.
In the case of still images, 'raw' does deliver deeper bit depth from the sensor... so one can have more Dynamic Range thereby.
It would be 8 stops if gamma were a linear function whereupon there would only be 8 doubling of voltage to an A/D converter, but isn't that why the gamma table compresses logarithmicaly?
Dgital video is encoded in linear steps, not logarithmic. Exactly the same number of bytes are used to encode a one stop change from zone 1 to zone 2 as from zone 9 to zone 10. The limitation on dynamic range is still governed by the analog swing voltage of the original NTSC standard. Ideal NTSC analog on a proper CRT can display up to 10 stops DR. Most current digital displays can't do that. As the Canon C series show, you can encode wider dynamic range in 8 bits with a high quality codec, but 4:2:0 AVC produces banding artifacts if you try to cram to much DR into a low bit rate hi compression format.
Last edited:
The C300 is 8-bit & has a dynamic range of around 12 stops.
PaPa
Major Contributor
No one has really shown this yet. From what I've seen in comparison videos, it's only speculation.
8 bits does not limit you to 8 stops or even 10. An 8 bit byte can represent any number between 0-255. Each number could represent anything, an f/stop, part of an f/stop, a british sterling to a british thermal unit...anything.
The important thing when encoding, is that whoever is decoding the secret message on the other end understands what he is receiving, be it the German Enigma code or mpeg4 h.264. Thats what a codec is for, code/decode.
The encoded byte is not just passing along raw sensor voltages. If it was, that would not be encoding and the poster would be correct that only 8 doublings would be possible to represent the 256 shades of pixel broghtness.
During the internal encoding step, levels of light are being translated into 8 bit values in accordance with the gamma curve, to later be decoded into a pixel brightness value at the receiving end.
So at the limits of what you can do with shaping the gamma curve in 8 bits, each byte could represent a maximum of 256 f/stops, or a minimum of just one f/stop divided into 256 parts. The former would give you a lot of banding, the latter would give you a very narrow dynamic range.
The important thing when encoding, is that whoever is decoding the secret message on the other end understands what he is receiving, be it the German Enigma code or mpeg4 h.264. Thats what a codec is for, code/decode.
The encoded byte is not just passing along raw sensor voltages. If it was, that would not be encoding and the poster would be correct that only 8 doublings would be possible to represent the 256 shades of pixel broghtness.
During the internal encoding step, levels of light are being translated into 8 bit values in accordance with the gamma curve, to later be decoded into a pixel brightness value at the receiving end.
So at the limits of what you can do with shaping the gamma curve in 8 bits, each byte could represent a maximum of 256 f/stops, or a minimum of just one f/stop divided into 256 parts. The former would give you a lot of banding, the latter would give you a very narrow dynamic range.
BobbyMurcerFan
Well-known member
I'm confused?? Then what does the "Log" stand for in S-Log and C-Log? Also, doesn't a gamma curve make the endcoding non-linear? And finally, doesn't using a knee or some type Cinegamma/Hypergamma that rolls off the top end also make the encoding non-linear?
Can the GH3 guys confirm this with footage? The GH3 footage I've seen so far is a long long way off my C100 at 12 stops. Much closer to my GH2 at 8 or 9 stops. And I haven't seen GH3 footage which has more than a 1/2 stop improvement over the GH2. I'm not looking to ruffle feathers...the GH3 would be very handy for its 1080 60p...but my eyes just don't see the high numbers being quoted.
PaPa
Major Contributor
I dare say the GH3 has barely a stop more, but again, no one has done any test to really show this. I'm sure it will come when more people have them in their hands! I am excited to be shown just how much better the GH3 is in DR than the GH2, but I've seen no evidence of that quite yet.