skibum5
Offline
• • • • • •
Upload & Sell: Off
|
The tricky thing about sensor tests is that you need to realize that you can't read too much into any one thing and to try to keep everything in mind and about what aspect may matter more than others and when (and also to realize that lots of these differences in SNR and DR are not necessarily so easily evident to the eye).
1. if you compare jpg output who knows what in the world the settings were, very hard to compare sensor performance
2. if you do look at RAW data that elimates jpg engine issues but you need to remember what the tests do and do not do
a.
for example, let us accept the DxO result that the 50D and 7D have the same DR at ISO12,800 (although I still keep getting more for the 7D, with every single body myself)
now that is a valid piece of information, the same DR
and yet if you only look at that and don't look at tests of pattern noise/various sorts of banding that is NOT the whole picture, in fact, the pattern/banding differences are often more relevant in practice and yet just about no review sites look into this aspect; the differences here between a 7D and 50D or 40D and 50D or 5D2 and 1Ds3 are greater than the impact of the measured DR differences
you can potentially have a camera that rates better for DR have less usable DR when you go to process pics since certain forms of pattern noise are rather objectional to the eye and don't scale away easily either (although some aftermarket programs can help, which complicates things further)
furthermore, even if a single pattern/banding number tested with data is listed even this may potentiall miss the real story, case in point the 2nd and 3rd 7D bodies I've look at carefully; body 2 would appear to do a little better in this regard with a one size fits all RAW test than body 3 and yet if you take a quick peak at a converted image from both you'll instantly notice that while body 3 has much more banding over the left 1/6th of the image and noticeably more over the bottom 1/6th of the image it has the same or less over the rest of the image, in particular it has noticeably less in the entire center region of the image which is perhaps the most important part of the frame so in real world images processed the noticeably worse performance over the bottom and left 1/6th of the image is only very rarely even half as objectionable as the modestly worse performance of the second copy over the rest of the image
SO:
a simple look at a RAW DR test might peg the 50D and 7D as even performance at high iso and yet the real world visual result is that the 7D generally does, effectively, noticeably better in this case
and a 7D body that tests a little better with a single broad stroke pattern noise/banding RAW test results actually looks WORSE on a large majority of images shot (well, images shot where shadow detail is important, otherwise it is all moot)
b.
some cameras, really only the 7D in terms of DSLR, have rather unbalanced green channels and more column to column variation
a basic RAW test would show better SNR due to the green channel being a bit more transmissive in one channel and a few other improvements when comparing the 7D to the 50D (ignoring for now the column to column stuff and pretending it doesnt exist) and yet a real world look at an image processed with a program that did not take into account the 7D's special behavior would show much worse SNR right up to the very bright parts of a picture as well in a rather objectional looking form from the 7D than the 50D at even ISO100
many noticed this and complained on various forums
OTOH a program that did take into account the special treatment the 7D requires would maintain such a hypothetical measured SNR advantage and possibly even ever so slightly increase it and yet if you didn't test for micro-contrast and detail you wouldn't notice that some of the extra detail grab of the 7D compared to the 50D has now been given back (although it still does grab more)
the DxO RAW test for SNR doesn't seem to show any advantage for the 7D over the 50D really and yet this might be because it is picking up certain types of 7D imbalance as noise but if you have a converter that can undo it reasonable well the real world processed images may suddenly do better than the DxO results would imply and the 7D might actually have an advantage large enough to notice over the 50D, haven't really checked but it seems quite possible, although it won't be anything at all crazy like all the 1-2 stops better nonsense!!!
c.
if all you look at are normalized SNR and DR RAW tests, don't forget about resolution and detail retention and the detail that a 7D can capture may in many cases be more important than minor SNR/DR differences, it all depends
keep in mind that at worst you could get the same detail as a lower MP cam only with less demosaic artifacts too
try to get all the detail you can out of the 20D/D300s and you get jaggies and stair-steps and moire and you can easily beat that detail level while maintaining a very nice and natural smooth look with a 7D
you can apply special NR algorithms and preserve same visual amounts of noise while retaining detail better in many cases
for single shot high dynamic shots at lowest ISO some of the nikons with better DR may be more useful and yet for a bird or animal off in the distance or even any general picture where absolute max DR is not THE key the 7D files tends to come out noticeably better than the D300s file
d.
detail, SNR, DR, pattern/banding are still not the whole story as some cameras, notably many recent canon bodies 'cheat' with the CFA a bit so you can potentially end up with mushy color, more monochromatic shadows or larger scale chroma clumping; otoh chroma noise is usually easier to remove without detail loss than luminance noise which the more transmissive CFA may help a bit so it's all tradeoffs when it comes to more transmissve CFA and more MP, wanting to push it far but not jump the shark with it all
side note, i will be a snob  , and say that the 7D doesn't peform as well overall as the 5D2 unless distance limited (where it actually does better) (I see some insist the 'clean pixels' of the 5D2 may it do better even for birding especially at high ISO, not that I see....; I see some insist the 7D fully matches the 5D2 at the image level for detail, DR and SNR, not that I see aside perhaps from DR at times)
side, side note, freezing rain here, you go out and shoot in it!
|
