alundeb wrote:
The A900 watch face is still not as sharp as the sharper parts of the NEX image (mostly everything)
I agree. The NEX-7 DpReveiw sample does look sharper in most areas - and displays more artifacts - than the a900 example but there is clearly an issue with the test as shown so all bets are off so to speak. Typical DpReview stuff.
Tariq Gibran wrote:
I agree. The NEX-7 DpReveiw sample does look sharper in most areas - and displays more artifacts - than the a900 example but there is clearly an issue with the test as shown so all bets are off so to speak. Typical DpReview stuff.
We can conclude however that the NEX-7 is capable of producing fine detail at the pixel level, and that less sharpening would balance micro contrast, noise and artifacts a bit better.
wolfloid wrote:
Do you have special noise glasses? I can't see the noise here, even when enlarged.
At 100% I can see a little blotchiness in the sky, but I'm not sure I would call it extreme. Combined with the the watch face in the DPR's test image, might it reasonable to suggest that the NEX-7 has a little bit of an issue in the blue channel?
wolfloid wrote:
Do you have special noise glasses? I can't see the noise here, even when enlarged.
No special glasses required. It jumps right out at me. Here is a 100% crop of the original and below that is the same crop from the sky image, converted to B&W with the blues slightly pushed to 148 to lighten them up - clearly revealing a ton of noise for an ISO 200 file. http://www.gibranstudio.com/1604476b.jpg http://www.gibranstudio.com/1604476c.jpg
@ bluetsunami: I did the same thing; the watch photo post generalizing the output of the two cameras was a bit misleading IMO.
One thing I did find interesting in the comparison between the 7 and 5N at higher ISOs is the fact that sizing the two images to the same output, the 5N noise advantage becomes null. Pixel eeping aside, since I shoot pictures, not pixels, this is not entirely a surprise...
I guess things will get better when raw-converters learn to use the base material better. There's more noise than there would need to be. The correct conversion strategy would be to apply almost twice as much NR to the blue channel as to the other two (at 5000K and lower WB temperatures at least).
When the raw converter tries to fill in the missing two values per pixel, in a surface where the blue is the strongest input channel (even though it's the weakest channel in relative terms) noise WILL be present.
Consider that the mix (from raw channels to Adobe RGB) of input that results in the "green" Adobe RGB often has a matrix looking about like [-0.1R + 1.5G - 0.4B] = ARGB "G". What will happen if the green input is weak, and the blue strong - but still noisy? The blue noise will dominate, and spread into the green channel - where most of the luma information is. That is why NR has to be applied to the blue channel before Bayer interpolation. There's often not much detail information in the blue channel (alone) anyway.
..........
BTW - that A900 shot looks pretty weird to me. At F11 just about any crap lens should be sharper than that if correctly focused. I would consider anything produced by my A850 (in a studio situation) as soft as those shots as "failures". Strange about the watch though, seems some areas (where's the watch face in the image?) are sharp, others not so sharp. I don't like DPR's test methodology, not their "consistency" either.
theSuede wrote:
BTW - that A900 shot looks pretty weird to me. At F11 just about any crap lens should be sharper than that if correctly focused. I would consider anything produced by my A850 (in a studio situation) as soft as those shots as "failures". Strange about the watch though, seems some areas (where's the watch face in the image?) are sharp, others not so sharp. I don't like DPR's test methodology, not their "consistency" either.
Completely agree. The watch is located near the bottom right of the frame - closest to the camera- and I would estimate there is at least one to two feet of depth in the studio setup. If you download the original raw file, it's very clear that the queen of hearts playing card - located furthest from the camera in the center - is completely oof and the dof, even at F11, is not sufficient to bring everything into critical focus using the 85mm lens that close to the subject on the FF a900. Again, this just proves how lacking DPR's methodology is.
ITT people post 100% crops of images and complain about shot noise as pixel counts get higher and higher. DPreview's own conclusion sums it up -- the NEX 7 has the highest IQ of any APS-C yet.
Doubtful that what is being seen in the blue sky is due to jpeg artifacts. As far as I am aware, jpeg artifacts do not take on grainy, organic shapes at the pixel level - the jpeg artifact pattern is often linear and mosaic looking. Jpeg artifacts also do not restrict their effect to just one channel either. IF the image of the flag with the blue sky is opened in Photoshop and the Channels palette is used to view the individual channels, it's clear the flag and base are really smooth and noise free while the blue sky is full of noise.
