|Ok, those of you who read this blog occasionally are already aware that I really like the test images at http://imaging-resource.com for their consistency of lighting and setup. Especially the mannequin series for high ISO. That series is simply excellent because it stresses the camera for both noise and detail. And low-contrast detail to boot.|
You see, I really hate the “helmet-hair” effect when all the hair on a person’s head becomes one mass of the same color with virtually no detail on individual hairs. That creates what looks like a helmet, and that really sucks.
So they recently posted the K5 first shots (finally) and that means that I can now compare the D7000 and the K5. I’ve been waiting for weeks on this …
The K5 looks like the darling of the new generation of DX cameras. It has amazing DR stats and noise stats that get really close to the D700 above 6400 iso …. this is simply amazing. Could it be that the D700 is being seriously challenged by these two? I’ve already looked this over a bit, but here we are able to finally look at them all in low light.
First, DXOMark has this to say about the sensors:
Yes, you read that right … the D7000 equals the D700 sensor score, and the K5 exceeds it. So no need to buy a full frame cam any more, right?
Well …. check the ISO scores and you will see that the wonderful scores for these cameras are specifically driven by the amazing dynamic range of these sensors. And they really are amazing … but the D700 is already so good that this certainly would not “do it” for me … the ISO difference is still dramatic.
In fact, the D700 remains superior by a full stop at 200 iso through about 800 iso … after that things narrow a bit, but let’s look at it closer to see what is really happening.
No measurement system can be perfect, and I care about the kind of details that are not factored in … so I think there might still be a bit of a difference as ISO climbs, despite this graph.
All images Copyright © 2009, 2010 imaging-resource.com …
First, we look at the three cameras at 800iso. My favorite crop is a 500px wide crop from the eye through the hair to the left. This shows tonal gradients around the eyelid, fine high contrast detail on the eyelashes and fine low contrast detail on the hair. What you watch for is clumping in the hair … where hairs next to each other lose their edges and appear to mass together.
So far, the D700 is ahead by a bit, as predicted. The hair is pretty uniformly textured, whereas both DX cameras show signs of clumping already. It’s not bad, though, and it certainly would pass muster for large prints. So 800 is a free pass … use it at will. No noise here either, by the way.
Now, some of you raging pixel peepers might notice that the Pentax looks a tad sharper than the Nikons. Since these are jpegs, this comes down to default settings of the jpeg engine, and possibly a slight difference in the strength of the anti-aliasing filter.
But a slight tweak of contrast and sharpening (USM) is all it takes to equalize, so obviously the camera settings can be adjusted in the D7000 to give that K5 “look” …
Sidebar: Some of you might have noticed that the crops are almost identical in size. How can that be? The D700 gives up about 700 pixels on the horizontal axis. Well, that is the magic of bicubic interpolation. And this puts the D700 at a slight disadvantage. Remember that …800iso is a virtual tie between the DX cams and the D700 is better. But not by a huge amount (we’re not playing with RAW and difficult lighting here, so some of the advantages of the huge pixels cannot assert themselves in this test – so just be careful about the kind of conclusions you draw from this narrow difference.)
Moving on to 1600iso …
Well, I’m pretty impressed. All three look great. The D700 is stretching its lead a tad … saturation is excellent and the hair has hardly changed from the 800iso image. The two DX engines are showing a bit more clumping to my eye, and the Pentax’s heavy sharpening is degrading edges a bit. At least, for the way my eyes like to see gradients.
Sidebar: What I’m driving at here is that I am really attuned to round things looking round. Meaning the coins should have that lovely metal roundness to the letters, wool should have a 3D look that is unmistakable … and when sharpening and noise reduction are starting to get excessive, its these edges that break down quickly. That causes hair clumping at the beginning, but pretty soon wool begins to clump and metal loses its 3D look and so does glass. This is all easily visible on DPReview’s compare engine … check out compacts like the LX5 at 1600iso and you will see rampant edge destruction. Still good for a compact, but nothing like what we see here … and those images have much more light than is used here.
So … no real challenge to the D700 (but the curves to not get close until later on) … and the D7000 looks better to me than the Pentax, but that is because I put high weight on dimensionality. Your mileage may very well vary on that score.
Bottom line is that 1600 iso looks to be a free pass as well.
3200 iso should start showing some stress for the DX engines …
Well, they are all still very good. Maybe not a free pass, but pretty close.
That said, the D700 continues to show excellent hair texture and all the skin texture around the eye is still there. The D7000 has lower saturation, but that can probably be compensated in camera if shooting a lot in low light. The skin texture is still there and the hair texture is not bad at all. The Pentax, though, has wiped the skin texture away completely. Looks nice, if you don’t care about the texture, but it is starting to show its plasticity. The hair as well continues to clump more than the Nikon.
I think the D7000 has pulled ahead. But we are entering the part of the graph where the K5 looks great on the chart. So let’s continue … 6400 iso is next …
Tough call here … the early generation sensor of the D700 is showing chroma noise. Not a lot, but enough that you would need to process it. I’d shoot RAW and use ACR anyway, so it’s not much of an issue. And the hair still looks like hair. This is why we buy the big sensors …
The two DX cams are really stressed by now … serious clumping of the hair, which is looking more and more like a helmet. But … this is 100% … downsized, these still look amazing for 6400iso. For web use, I give them both a free pass. Especially if you process. But for large prints … I don’t think these will do it for you. Some subjects, sure … but perhaps not people in really low light.
Sidebar: A note on low light: concerts are generally *not* low light. The performers are often very well lit, so you can shoot at 400 iso with adequate shutter speeds like 1/80s. Low light means a living room with incandescent lighting or a bar. These are really tough light.
And finally …. let’s look at 12800 … this is the point at which the D700 is only about 1/3 stops ahead on the graph. Let’s see how real that is …
Fair warning: If you think you will be regularly shooting at 12800 with a DX sensor, then I suggest that you might need to seek help … in the form of a therapist or a D3s …
If you go back up to peek at the SNR 18% DXOMark graph again, you will remember that the Pentax has clear circles at around 1600iso. Those are marked in the legend as “Pentax K5 Smoothed” … and I think this series of crops makes very clear what that implies. Wow … the Pentax handles chroma noise wonderfully … better by far than the D700 in jpeg (older generation of sensor and processor) … but the smoothing is just rampant. Still … were you shooting in the dark and you have to shoot jpeg, you would probably choose the Pentax over even the D700 …
Or would you? Let’s see what a small bit of processing would do to these three.
Hmmm … still some chroma noise for the D7000 (eye is the wrong color) and the D700 (some blotching still evident.) But the D700 is still retaining more detail in the hair and the skin. So I would probably shoot the D700 in the dark over the other two. YMMV of course.
The Nikon jpeg engine in the D7000 is not quite as sophisticated as the Pentax jpeg engine, obviously. But the super-smoothed effect will not work on all subjects, so it is a toss up whether you would prefer one over the other. And as I mentioned earlier … if you shoot DX sensors at 12800iso regularly, you probably need your head read.
Now … let’s see these three at web sizes at 12800iso to see what they might look like on Facebook.
Pentax K5 12800iso
Nikon D7000 12800iso
Nikon D700 12800iso
The Nikons both have slightly cooler white balance, but could it get any closer? At web sizes, these images all look terrific. So maybe you Facebook shooters can have at it for your memory albums … 12800 in low light actually works with a bit of processing.
This generation of sensors is pretty impressive … for shooting at high ISO (not extreme ISO) they do not match the full frame cameras … but they are not dogs by any stretch.
So … here are my picks for low light high ISO shooting:
1st – D700 (never any doubt)
2nd – Pentax K5 (by a nose)
3rd – Nikon D7000
But I would never recommend that someone switch systems for such a small difference. The D7000 shoots very well to 6400iso and looks ok after processing at small web sizes at 12800. I’ll try to free up some time to do a similar exploration with RAW images for the two DX sensors. Unfortunately, the D700 was never shot at higher ISO in RAW though, so I can’t have as much fun in RAW :-)