Canon SX50, 50D, Sony A58.
Amin Sabet: Great review, and I love the demonstrations of exposure latitude and ISO invariance. Really well done!
Hope to see these tools include D810 soon.
"Hope to see these tools include D810 soon."
And 6D too. In the conclusion they say that "The main competitor to the D750 is Canon's EOS 6D.", so think it would have been natural to use 6D in the exposure latitude test, instead of the 5D3 that has less DR and more banding than 6D. Wouldn't make a huge difference, the D750 would still be far ahead, but still.
ottonis: First of all: terrific job with that comprehensive and sophisticated review, DPR! Hats off!
I have one question, though:
When using the studio comparison tool it appears to me that the D750 is on par or even slightly better than Sony A7s in low-light up until ISO 6400 -12800. THis is a huge feat, IMO, considering the A7s has been specifically designed to provide top-notch low-light sensitivity, low noise floor and high-ISO IQ.
I don't know how to read out the exif data, so someone may help me out: have the aperture and shutter speed been the same across the D750 and A7s for this particular studio comparison?
Thanks in advance and keep up the great job, guys!That being said:
Hold the mouse over the 'i' below the crops to see exif. The f-stop and shutter speed are the same for D750 and A7s, but the 6D gets 1/3 stop less exposure (uses 1/3 stop faster shutter speed). Don't know why.
Dimitris Servis: How does it work at iso 50?
"ISO 50 costs you 1 stop Raw DR compared to ISO 100. If you can't see it; your scene doesn't have enough DR."
ISO 50 has 1 stop less 'headroom' (above metered middle gray) and 1 stop more 'footroom'. The total RAW DR is the same at both ISOs.
Steen Bay: The 'Exposure Latitude' and 'ISO Invariance' tests (page 13) are very interesting. Hope that you (the review team) will add more cameras, like e.g. the 6D, Canon's currently best FF camera regarding DR. It won't be dramatically better than 5D3, but 6D has a bit more DR (0.4 Ev according to DxO), and it also has less banding than 5D3.
You write about 'ISO-invariance'.. "However, this way of shooting confers an image quality advantage if you make the effort to work around these inconveniences.". Wouldn't put it like that. With an ISO-invariant (or ISO-less') camera you can get more safety highlight headroom when shooting in low light, but that'll just be an advantage if you otherwise would have blown important highlights. If you just 'expose properly' (increase the ISO, but not so much that important highlights are blown) when shooting with an 'ISO-variant' camera like 5D3, then you'll be fine.
"1, 2, 3, 4, and 5 stop pushes are already in the ISO-invariance widget"
And a 6 stop push too, but that's a completely different thing than 1-6 stop pushes at base ISO. In the ISO-invariance wignet the exposure is 6 stops below full saturation (ETTR) at base ISO in all 6 examples. That's why the D750 in all 6 examples have higher noise than it has with a 5 stops push in the Exposure Latitude wignet above.
iAPX: The ISO-invariant part is incredibly interesting, having a D610, I will test it with it, to see if this technic gives good results with -2EV or -3EV of exposure, divising ISO by 4 or 8: 800 ISO instead 6400 ISO for example gives a far better dynamic, and at same time while shooting live shows, it enable to shoot at much faster speeds, increasing the potential of a 70-200 f/2.8.
I am glad I learned a new way to do things, that could give immediate benefits!
The f/stop and shutter speed aren't supposed to change if you shoot 'ISO-less', just the ISO. If you shoot at ISO 800 instead af ISO 6400, then you'll get 3 extra stops of safety highlight headroom. If you don't need that extra headroom (if you don't normally blow important highlights when shooting at ISO 6400), then the end result will be the same as before.
P.S. - Maybe the 5 stops push in the Exposure Latitude test is a bit extreme. Many (or even most) photographer will probably never want/need to lift the shadows that much. How about also showing how e.g. a 2-3 stops push will look? Think it could be useful/relevant, because 3 stops latitude probably will be enough for most people in most situations.
The 'Exposure Latitude' and 'ISO Invariance' tests (page 13) are very interesting. Hope that you (the review team) will add more cameras, like e.g. the 6D, Canon's currently best FF camera regarding DR. It won't be dramatically better than 5D3, but 6D has a bit more DR (0.4 Ev according to DxO), and it also has less banding than 5D3.
Just another Canon shooter: Thank you for adding a page about DR. It was time to make this "more official". Canon cannot pretend anymore than they do not know what everybody is talking about.
BTW, why comparing the 7DII to the D7000? Why not to the D7100? The latter is newer, and ... suffers from banding (still better DR than any Canon).
Let's say you're shooting in low light with an ISO-less (or ISO-invariant) camera at base ISO (ISO 100). Don't think that any cameras are quite ISO-less yet, but let's say it is and that you choose 1/60, f/4, which will give you an exposure 5 stops below ETTR. Another photographer with a Canon camera will also use 1/60, f/4 and increase the ISO to ISO 3200, or maybe just to ISO 2500 to be on the safe side. Or maybe he'll use Auto ISO with -1/3 or -2/3 EC.
Also, if you have a camera with EVF (or use live view), then things like 'exposure simulation' (WYSIWYG), live histogram and maybe zebras/blinkies make it much easier to get the 'correct exposure', including ISO, without blowing important highlights. Like I said, don't think that the possibility to shoot ISO-less is such a big deal.
Notice that I said "unused headroom" and "final image". If you shoot at base ISO with an ISO-less camera and your exposure is x stops below full saturation (below ETTR), then you have lost x stops of DR.
@Rishi Sanyal - 5 extra stops of unused safety highlight headroom if shooting with an ISO-less camera in low light won't give you more DR i the final image. The extra headroom is just an advantage if you otherwise would have blow important highlights. If exposing 'properly' when shooting at higher ISOs, without blowing important highlights, then the extra headroom that ISO-less shooting gives you won't be needed. OK, the extra headroom could maybe be nice to have if in a hurry, but I really don't think it's a big deal. Just protect the highlights when you shoot at higher ISOs (don't increase the ISO too much), just like you have to do when shooting in good light at base ISO. Then you'll be fine.
Jack Hogan: We look forward to seeing the relevant EM1 images in the New Studio Scene updated to reflect performance with the new mode. The current ones show it sompm' fierce.
Right. The ISO 800 image shot at 1/320 sec is clearly sharper than the ISO 200 image shot at 1/80 sec.
D4s gets 1/3 stop less exposure than D4 in the comparison tool. For example 1/4000 vs 1/3200 sec at ISO 6400 (f/5.6 on both cameras). Here assuming that the lighting is the same, like it's supposed to be in the new comparison tool.
fenceSitter: "50 Mpixel CMOS sensor with almost twice the physical size of the largest 35 mm DSLR sensor."
On which planet are 1.67 "almost 2"?
Phase One IQ250 has (most likely) the same CMOS sensor and the price is $34,990, so this Hasselblad is actually quite a bargain.
falconeyes: This is both interesting and isn't.
It is interesting because the RED can capture 19 MP images at 85 fps, quite a beast. And of course, at 1/42th s, you just combine two frames into one, thereby halving effective base ISO for slow shutter speeds. For such combined double frames, DxO's measured performance is disappointing actually.
It isn't interesting because for stills with a focal plane or global shutter or flash photography, you can't use this technique without embedding a digital frame memory on the sensor. And this is nowhere near in time.
DxO knows it, which is why they don't list the RED in their rankings.
DPR's headline is WRONG: D800E still is #1 in the DxO ranking, it hasn't changed.
Think that the Red Dragon combines more than two frames at base ISO. The D800 has a 78% larger sensor and a lower measured base ISO than the Red Dragon (ISO 74 vs 104), but despite that the base ISO SNR comparison (see 'In Depth Analysis') shows that Red Dragon has a much better SNR (1-2 stops!).
Steen Bay: Well, if something sounds too good to be true... As far I can tell the QE of the Red Dragon must be quite a bit higher than 100% in order to get such a high SNR, and that's hardly the case.
D800 has a 78% larger sensor than Red Dragon and the D800 also has app. half a stop lower measured base ISO (ISO 74 vs 104), meaning that it can 'handle' app. half a stop more exposure, but despite that the base ISO SNR comparison (see 'In Depth Analysis') shows that the Red Dragon has almost 2 stops (5-6 dB) better SNR in the midtones. That's simply not possible, since the QE of the D800 already is about 50%.
Well, if something sounds too good to be true... As far I can tell the QE of the Red Dragon must be quite a bit higher than 100% in order to get such a high SNR, and that's hardly the case.
Buhl213: How can the Tamron be ENTIRELY useless when cut in half?
Should be OK for half-format.
Steen Bay: Rather disappointing performance, I think. Lumia 1020 has a 1/1.5" sensor with twice the area of a 1/2.3" sensor like used in for example Panasonic FZ70, and yet the FZ70 has better SNR and DR if comparing at the same exposure (same 'measured ISO'). Shows that the 1020 sensor has a rather low quantum efficiency (less than half the QE of the FZ70 sensor).
@HowaboutRAW - I didn't, I was just commenting on DxO's test results.
Bervilat: So, this will be actually worse for base iso landscape photography?
Depends.. DR is better, and the QE is probably quite a bit better too, but the higher base ISO means that SNR 18% probably won't be better at base ISO than it was with the CCD sensor.
AndyHWC: I stopped checking DXO for a long time because its result don't always match what my eyes told me. Is it comparing overall picture quality or pixel noise. To fit 41mp in a tiny sensor, you expect noise to be bad in pixels level. Anyone don't like their 1020, I love to trade my HTC 8x with you.
The sensor score is about things like noise and DR, it doesn't take the resolution into account. Could be interesting if DxO also tested the resolution of the 1020 ('lens score'), which of course is its strongest point (assuming the lens is good enough).