skibum5
Offline
Dedicated FM
Upload & Sell: Off
|
 p.1 #6 · 7D & 'Mazing' (serious question) | |
omarlyn wrote:
Before I get to my question, I want to make it clear that I am not trying to fan any flames here between those that have posted (sometimes contentiously) about the 7D's 'mazing issue'. Without turning this thread into a back-and-forth arguement, I genuinely want to understand how to 'see' the 'mazing' that's been discussed ad nauseum.
My question/observation is the following: I've had my 7D now for just a few days and I haven't seen any 'mazing' as has been described in multiple posts here. However, I did notice that when I view a RAW image beyond 100%, I can see the actual pixels of the image. I'm not saying that some people are not genuinely having a problem but I do want to fully understand what I should look for. At low ISO on an OOF area, all I see @ 100% (using DPP) is OOF (with maybe just a few sharpening artifacts). When I view @ 200%, I see alot of blocky pixels so I don't know what difference any 'mazing' would make at this point. Is the 'mazing' pattern larger, smaller, or the same size as the blocks of pixels?
Omar
now this image had a TON of shrpening applied just to make the differences clear, notice that the upper left has a realmazing look to it with spiralling extended straight lines looking nothing like random noise, the one on the top right is a better copy of the 7D where the mazing is starting to blend into a normal noise pattern, the bottom right is a 50D, at this insane level of sharpening it has artifacts but they are largely random looking without more than a tiny trace of mazing structure and bottom left is a 5D2 sample comparing that to the one above it makes it clear how the mazing noise has an enitrely different structure to regular noise.
different converters reveal mazing to different degrees, DPP shows it less and tends to break it into dotted lines (it also is beginning to seem like DPP smooths over areas of low contrast); current versions of ACR (and a few other) show it the most
here is one copy of the 7D with very, very normal processing:
and 20D:
and a typical 50D copy:
when looking at raw data the 7D tends to show one of the green channels having less noise than the other and slightly higher avg values (on the same patch) than the other and many tend to have a repetitive pattern of varying gain column to column on the green channels. some have it low enough that it doesnt seem to mess too badly with raw converters, quite a few seem to have the variances enough to cause artifacts/more noise after demosaic. perhaps doing all sorts of tricky stuff will allow acr 5.6 to handle copies better.
|
