Under the A900's Hood by GaborSch...

[Javelin]

he claims the A900 raws are ... not raw ? here we go again!

[douglasf13]

Agreed that it's an interesting read, Sonolta. You should check this thread out, because Iliah Borg weighs in with rebuttals:

http://luminous-landscape.com/forum/ind ... t&p=267987

[douglasf13]

Yeah, there is definitely something strange going on, although I'm surprised Gabor is still using ACR to even look at the A900 files, considering Adobe's shoddy conversion history with Sony. Iliah's comments are interesting to me in that he isn't arguing that there aren't issues, but rather that the cause of the issues aren't what Gabor thinks that they are. Interesting stuff. I hope that Iliah lets us in on what Sony has to say about the tests he (and his group) sent them. In all honesty, I hope that Gabor is right, because that means it's fixable. If Iliah is right...and that's usually the case, then it may not be fixable.


p.s. I found it particularly interesting that Iliah claims the intermediate ISO settings, like 320, are handled better in the A900 because of the pre-amplified signal. I'll have to try that out (although I really haven't had much issue with low ISO in the first place.)

[bfitzgerald]

Hmm a few things here.

We have known for a while that high ISO on, does indeed do NR to raw, I have no idea why this happens. You will have to ask Sony about that one. IMO a bit dumb, but as I never use high ISO NR on anyway..it's not a deal breaker.
ACR does NR on it's own (and unavoidable), so he really should not be using that software, try something like C1 4 and RT.
He says high ISO off does a small bit of NR, I cannot say..not really looked at many A900 raw files.

High ISO NR should just work on jpegs, and IMO attack the colour noise, which is the fatal weak point of sony's jpegs. Instead they are bashing the luminance noise to bits, wrong strategy..let's hope Mark Weir wakes up.

[David Kilpatrick]

A lot of this is connected to the gain applied to the red and blue channels (especially the red channel) in connection with white balance adjustments.

Example: here is a picture which at ISO 400 has relatively low noise in my final conversion at 2850K (tungsten, and also the Auto WB selected by ACR based on the spotlight bulbs which create the white highlights) - a 1000 pixel version only is given here:



But the camera's own Auto WB setting for this was 7000K, and apart from diminishing the intensity of the blue effect lighting, the noise level is dramatically worse:



This is the kind of noise which, when people see it in a blue sky at ISO 400, leads to heavy criticism of the A900 along the lines of 'that's what I expect from my 5D MkII at 1600' etc. But if ACR's AutoWB (or Tungsten) setting is used instead, as I did for my final conversion, this is the result:



I realise the underlying coarse blob noise pattern is still present. I have found that the camera, and ACR's 'as shot' WB, tend to pick too high a Kelvin value all the time. This produces rather over-warm yellowish images. Even 5200 or 5600K can be too high. Just dropping the WB manually in ACr to 4900, 4800, 4500, even as low as 3900 for some daylight scenes can clean up the colour and reduce the noise. I do not always do this but I have become much more aware of the need to fine tune WB with the A900 and ACR, and the lower visible noise levels which result especially in sky blues. Also, applying different camera profiles in ACR can have a similar effect.

David