Why is D90 betten then D300s at DxOMark Sensor test?

Started Jan 4, 2010 | Discussions thread
cluna
Senior MemberPosts: 1,403
Like?
Re: What makes you think...
In reply to DStan, Jan 6, 2010

DStan wrote:

cluna wrote:

DStan wrote:

What makes you think their software is any good. I've used their RAW conversion software and IMO it stinks and was buggy. Scientific, IMO not in this case. Not all programs function correctly or consistently.

Im the one also questioning the result So what make you think I think the test is accurate ...

I guess then we're on the same page.

I never meant to be supportive of it, I know it is wrong per the 2EV DR at high ISO example. However, we learn a few years ago that the Nikon RAW hide some info from us....IE:WB info encryption. There is no reason DXO , Adobe, or myself cannot process a RAW and get the same result but there is a bit of 'cooking' of the data going on by Nikon, when it shouldnt be other than compression/truncation if the output file requires it. Id guess DXO is needing to profile each sensor point for compressed 12/14 vs a raw lossless uncompressed 12/14bit that the D300s can produce. Maybe that the missing link, but if RAW data is truely raw, even a MATLAB (or even excel) analysis of the file can produce S/R, DR , etc.

I think Nikon is being sneaky for Capture NX sales or DXO is needing to test each raw output type setting (and EXIF) permutation and only post he best output and note those settings for the reader to see.

I was very disappointed by their RAW conversion software and have a weak confidence in anything that they may do. I based my assumption of your position on the test's accuracy by the below posts:

Im unbiased here and only inspecting this output from an engineers perspective. I started with RAW files and even did my own demosaicing/denoising code way back with my Canon G3 's Raw file. It was straightforward then but if Nikon is being cryptic with raw output per body type, then it is reallly Nikon causing it for their own purpose. I understand the color profiling differences being special, but not RAW pixelsite differences. They should be consistant across the bodies with the same RAW format output and sensor type bin....but that isnt the case as we are seeeing or led to believe.

-C

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow