PDA

View Full Version : GH2 Iso 160



Sage
12-24-2011, 10:44 AM
Its not so hot.

160
http://sagepictures.com/160.png

200
http://sagepictures.com/200.png

400
http://sagepictures.com/400.png

Kholi
12-24-2011, 10:54 AM
Just covered this last week: http://www.campcomet.com/archives/797

Sage
12-24-2011, 11:04 AM
Just covered this last week: http://www.campcomet.com/archives/797

Great blog.

I didn't realize how much better 200 was than 160, until this shadow test post quantum.

Kholi
12-24-2011, 11:19 AM
Great blog.

I didn't realize how much better 200 was than 160, until this shadow test post quantum.

Thanks Sage.

Man I've been back and forth on the iso thing. I feel the same way you do, as well. I think 200 would be the download limit, a place where you go if you have to or if you're in ETC mode it's perfect. It's hard to try to stay at 320-400 iso but to me, that's where the image is the sweetest balance of all factors.

If you do some more testing please continue to post your findings. I am away from the cam and lenses so I can't do them!

Sage
12-24-2011, 11:54 AM
I believe 160 is a 'software iso', a post exposure curve manipulation just like the Red, but done in a much less classy 8 bit process, permanently in cam. Done in an 8 bit downscale color space, without noise to act as a dither - huge increase in banding, I see.
200, as a native gain switch - most DR & least green shift, that'll be my sweet spot. :cheesy:

Phenixone
12-24-2011, 12:18 PM
Stay on the middle ISO line (200 - 400 - 800 - 1600) no matter what then. Use a ND if necessary to keep your super wide aperture to 0.95 ;)

Heartfelt
12-24-2011, 05:02 PM
Thanks Kholi!

dasbin
12-24-2011, 10:30 PM
I highly doubt this has anything do with a truncation process in the camera only being affected for certain ISO's.

It is much more likely that all you are seeing is the effect of dither. Higher ISO = noisier. Dither is just noise that has been intentionally added to reduce the appearance of bit truncation (in this case colour banding). The same concept applies to audio bitdepths too, though. If you are using a recorder only capable of, for example, 8 bits, it will sound better if you are recording a noisier source, because the effect of truncation has been masked by the noise. This actually increases the apparent dynamic range, too.

But it is all basically an illusion. In this case, the bands are still there, just less distinct at higher ISO's, as they are spotted with pleasant, random noise.

Please note in this case the colour *balance* is actually different between those three shots. The first is the most blue. The camera likely shows the most banding in blues. Whether this is an accident of timing (sun moved slightly out of a cloud or something), or a result of the higher ISO reducing colour saturation (I've noticed this in most cameras), in either case it does not mean the higher ISO is more accurate. It means the colour balance is being shifted out of the range where it is more likely to cause visible banding - it is irrelevant how this happens.

Kholi
12-24-2011, 11:10 PM
Color balance doesn't appear to be different in those three shots, the difference is the chroma noise polluting the mids and lows in the second two,and of course the added sensitivity raising the values in mids and lows. iDesaturate most of the chroma noise, roll your mids down in a curve and they're the same.

Dither is one thing, software ISO's are another: without hardcore data and facts from engineers we do not know which ISO is the camera's actually balanced for. Software enabled ISOs may also enable extra noise filtering (noise reduction?) done in camera or more. Higher ISOs do not reduce color saturation. Depending on which, the more or less exaggerated. It's greater in Canon cameras, and much more noticeable.

While dithering can help the process, it may be better to try to avoid the banding and posterization altogether in camera versus trying to clean it up in post. I think people should try Dithering a 160ISO image in the same lighting. Then try denoising an image at 200+ ISO in the same lighting. Let us know what your results are, of course.

yoclay
12-25-2011, 12:06 AM
Banding issues aside, the noise in the corner of the room near the bottom looks absolutely awful. Especially on the 400 iso shot. I am surprised how present this is, even at low iso's. Makes me want to go back to my 5D.

DrDave
12-25-2011, 12:15 AM
There are certain types of scenes that really show different kinds of noise, but fortunately we don't do a lot of vid of flat, grey surfaces in shadows. It is annoying, though, if it is in a patch of sky or intrusive in an indoor scene, like shadows in a bookcase.

Kholi
12-25-2011, 01:34 AM
There are certain types of scenes that really show different kinds of noise, but fortunately we don't do a lot of vid of flat, grey surfaces in shadows. It is annoying, though, if it is in a patch of sky or intrusive in an indoor scene, like shadows in a bookcase.

Well, actually, in my line of work I often get stuck with solid areas of color (walls) that have to be dealt with. In narrative, it is definitely not uncommon to be forced into a corner like what Sage has, and have to light talent and tame that wall. Different things for different people.

But, speaking of noise I am interested in what Driftwood has up his sleeve for the next patch. He's been hinting at something special.

Danielvilliers
12-25-2011, 02:17 AM
I have seen this post in PV by someone who has done test on the native gh2 iso on Dpreview. http://forums.dpreview.com/forums/readflat.asp?forum=1041&thread=40120298 perhaps it can bring some more light. Its conclusion is completely the contrary of here, saying the native ISO are more the 160, 320, 640 and then talks about analogue and digital boosted iso. For me, I am more incline to believe that the 160 is indeed native as it is the cleanest, unfortunately it shows more of the imperfection of the signal or treatment of it, which is in this case 8 bit banding. Now sometime the best technical thing is not the best visually. I would be more inclined to shoot a little higher ISO to break the banding.

Yes me also, I would like to hear about the driftwood comment. He sounded as something big was going to happen, like some breakthrough, not just a Quantum version 6 but something like when the first high quality gop1 where coming. Strange I did not see anybody talking about it.

DrDave
12-27-2011, 11:20 AM
In some respects the most interesting thing about the native ISO discussion is the presence of careful tests that reach completely opposite conclusions. A Schrodinger Cat scenario.