Nikon D850, Early Impressions

What DxOMark call "SNR" is SNR for medium toned subject i.e. something like a gray card (but without the texture of the card as they use ND filters to achieve different luminosities). However, SNR is really a function of the detected number of photons. In some cases DxOMark have shown the full SNR curve for each ISO; it would be useful if they always displayed that. Different cameras have advantages in different parts of the curve (and different ISO settings). Anyway today the differences are most pronounced in the shadows. If you edit the image extensively, the differences may become visible. However, in my opinion most cameras today have very good image quality and too much should not be made of them. I do find the D5 automatic white balance to be more consistent across different lighting types and this is very helpful. The D850 seems to reproduce this behavior. Note that if you have a different white balance in the displayed image then the apparent noise will also be different (more neutral in low color temperature lighting => blue channel is amplified which makes the image look a bit more noisy; if the colors from the different cameras are adjusted to match, a more fair visual comparison of noise can be made).

 

That said, I routinely use up to ISO 25600 (and some cases have successfully used ISO 102400) on the D5 whereas with the D810 I don't go above ISO 6400 and am happiest using the D810 at ISO 64-1250 or so. There is a difference between these two in terms of which ranges of ISO they do best, but it is something one can best see in real-world use. Measurements confirm these differences but only real-world use in one's own applications can assess the significance of these differences in practical use.

 

Pixel Shift and the accompanying algorithms 'could' do away with a lot of pixel level noise. It's only suitable for stationary objects, but as a studio cam...!

 

In the studio, photographers usually use flash. To use pixel shift with flash I think you would need very consistent flash output otherwise there could be consistency problems between the images that are pixel shifted. With continuous artificial lighting you'd want to make sure the lighting doesn't fluctuate between exposures; is there synchronization with the 100 (120) Hz oscillation?

 

You say the AF isn't up to D5 standards, but how does it compare to the D800 or D810? And is it any more accurate in terms of hit-rate or absolute ability to nail focus without the fiddle of fine-tuning?

 

I don't have the D850 yet but will comment anyway. There has been a lot of progress in focus accuracy since the D800, even before the D850. There is also now the automatic fine tuning method available, which (while not perfect) has saved me a lot of time with the D5. While fine tuning is still needed you can expect it to take less time to do.

I found the D810's autofocus to be somewhat more reliable in good light than the D800 - whether this related to the issues with the AF alignment on early D800s (mine wasn't in that category) I can't say. In low light, the D810 struggles, possibly more so than the D800 - so I'm quite keen to have a step forward here. I'd love "Nikon's best autofocus", but I'm not going to hold "not being a D5" against the D850 too strenuously. I'm curious to see how the A7RIII's dynamic range compares, though. Not curious enough to switch systems, but...

 

Pixel shift and the techniques used by cellphones assume a longer effective exposure. The approach recent cellphones take is to capture multiple images and then align them to stitch into a lower-noise result. For a fairly static subject that can work well (although you can always do it manually); for a moving subject it's a harder ask. If you're trying to capture the moment, it's not really relevant.

In low light with the D810 you can use the group-area AF which is more sensitive and can handle focusing on subjects in dimmer light than single-point or dynamic area AF in my experience (and it focuses faster as well, and hesitates less). I found group-area AF to be one of the most useful new features in the D810 at the time, and I still use it frequently on the D5 as well.

 

From reports of more knowledgeable people than I am, I've read that the D810 includes correction for spherical aberration of known lenses which should reduce differences in focus between large and smaller apetures on a given lens. Additionally Nikon says the color dependent focus shift in the D800 was corrected in hardware on the D810. I find overall the D810 focus to be more accurate with fast primes than the D800. The D810 live view is of higher resolution and less noisy as well so it's nice for tripod-based manual focus work.

I could believe that explanation, Ilkka. And certainly true about the live view - the D800's live view limitations (particularly hanging until the image was saved, but also the quality) are one reason I updated. I should try group area AF more - I usually rely on 3D tracking, which is possibly the worst case.

My D800 and 50/1.4 AF-lots-of-things-G struggle with focusing..

Low(er) contrast contrast wide-open has been mentioned as the culprit .. annoying it is, on an otherwise very pleasant combo.

That would really be something I would appreciate very much in a new camera ..like the D850.. : better AF!

If one is only considering the sensor, I think even the D800 from 2012 is already very good, and I decided not to bother with the D810.

 

As far as I am concerned, the 20MP and 24MP FX bodies such as the D750 is more practical. 36MP and 45MP give me these huge NEF files without much advantage.

 

I have pointed out a number of times that, for still photography, the D850's advantages are:

1. Improved AF from the D800, D810, and D750. The additional AF points on the Multi-CAM 20000 is clearly superior, but the D850 is not quite as fast as the D5. Moreover, the extra pixels means accurate focus is critical, and motion blur is easier to observe.
   
2. XQD cards. CF and SD have given me too much trouble.
   
3. Better controls: I really like the ISO button next to the shutter release. Touch screen is convenient to have.
   

On the video side, the D850 is currently the best Nikon DSLR for that purpose. I sure don't mind capturing 4K video with the entire width of the FX sensor, but those video files are huge.

 

There are other little improvements too. For example, In Custom Setting menu D, the exposure delay mode has more options. Beside delays of 1, 2, or 3 seconds, the D850 lets you set to 0.2 or 0.5 second. The 0.5 second option is very convenient.

As your images show the size of the pixels has a direct relationship to noise. Large pixels have a greater well capacity and can collect more photons quicker. The image processor is another factor, NIkon keeps improving them. The Nikon body with the least noise is the Df. You don't buy a D850 expecting it to have lower noise. The D850 will make a better mural when used a low ISO's. For low noise and good image quality the D750 with only 24mp is hard to beat. Unless your shooting humongous prints you don't need a D850.

I have handled one, it is a beautiful body about the same size as my D700, but larger than the D750. I'll probably buy one in a few months but use my D750 for available light and weddings.

RIck

The D750 is a fine, general-purpose FX body, but it is more "prosumer" type with 1/4000 sec top shutter speed and 1/200 flash sync. It is certainly good enough for most things, and I use mine a lot, but overall it is a step behind "pro" grade.

 

Here is an ISO 12800 sample from the D850 last night. I took some street shots in a shopping area in Fremont, California. A decade ago I was getting similar results from the D300 at ISO 3200. At the pixel level, it is not great high-ISO results but usable. For web displays, like Barry's football images, it looks quite good.

 

 

It's interesting that chromatic noise has all but disappeared from modern sensors.

 

I guess down sampling (binning?) from such a large number of pixels will clean up luminance noise even more?

 

Andrew G..

 

Do you reckon multi-sampling of the same pixel will be better or worse than sampling adjacent pixels... regarding noise for final image IQ?

Andrew, what's this obsession with dynamic range? Isn't > 12 stops more than enough? That's easily 5 stops more than film ever delivered.

 

The lens used probably has a greater influence on (subject) dynamic range than the camera sensor. So if you want nice transparent shadows, get an old single-coated lens, or fit an uncoated UV filter.

 

You might want to work out how little stray light needs to be bounced around the mirror box and onto the sensor to make an image-plane DR of more than 12 stops an impossibility. (It's around 0.025% by my reckoning.)

Do you reckon multi-sampling of the same pixel will be better or worse than sampling adjacent pixels... regarding noise for final image IQ?

 

You mean is it better to multi-sample spatially or temporally? There's always a trade-off between resolution and subject motion. :-) There's been some very impressive denoising done recently with neural net algorithms; neural nets always worry me a little, because you're never 100% clear what you trained them to do, but many techniques applicable to photographs also apply to ray-traced computer-generated images. There were quite a lot of papers on the topic in SIGGRAPH and HPG this gone summer, if anyone has access and is interested; having depth information also helps, and I'll be interested to see whether cameras with PDOS can take advantage of this. As I'm fond of saying, all computer graphics is cheating; in photography there's no substitute for photons at the capture point, but electronics go a long way towards the final image. This isn't new to digital photography - it's not like Velvia is realistic, and Ansel spent most of his time doing local contrast adjustments with things glued on a stick.

 

The Nikon body with the least noise is the Df. You don't buy a D850 expecting it to have lower noise. The D850 will make a better mural when used a low ISO's. For low noise and good image quality the D750 with only 24mp is hard to beat. Unless your shooting humongous prints you don't need a D850.

 

It depends on your measures, somewhat; at high ISO, the D5 is clearly better (and worse at lower ISO - the D4/Df sensor is a good compromise). The Df slightly outperforms the D850 at higher ISOs, but it's very close (except at the magic ISO1600 value, according to DxO, for which that sensor has always been good); the D4s has a slightly larger lead at very high ISO. I really don't pay much attention to per-pixel results unless I'm working out whether a body gives me better reach.

 

The "humongous prints" argument is, to my mind, legacy thinking. I very rarely print things at all. What I do instead is view on a computer or tablet, which gives me the ability to zoom in and scroll around. I absolutely like having the resolution of my D810 - and I've used it to look for detail in images I'd not expected. (I've tried to work out whether someone was at an event by staring at the faces captured in a fish-eye shot; that wouldn't have worked nearly so well with a reduced megapixel count). Resolution isn't everything, but it's not nothing, either.

 

Andrew, what's this obsession with dynamic range? Isn't > 12 stops more than enough? That's easily 5 stops more than film ever delivered.

 

There are those who'll argue you can get a lot of stops by drum scanning some negatives. Still, my basic problem is that I often shoot in circumstances where I don't want to use flash - either because I'm being candid and don't want to annoy the subject or just because I don't want the faff of trying to balance the light colour. With film, we shot fill flash and used reflectors, and for rare shots like some of Ansel's we did dodge and burn on the enlargement. It's now sufficiently easy to apply local tonal adjustments (semi-automatically) in seconds on a computer that I treat it as part of my standard workflow.

 

We're starting to see HDR displays, too. And that doesn't mean we don't do tonal adjustments, just that they are working with a wider range. PQ encoding (one part of Dolby Vision) encodes a 100-times maximum intensity output relative to standard dynamic range (nominally 10,000 nits, not that almost any displays can do that, vs 100 nits for a reference TV). Assuming you're trying to pack 8 stops of detail onto a standard display (8-bit content, although it's not linear), linear capture of 14-bit raw only just gets you there even before tonal adjustments.

 

Put another way: I have shots of Antelope Canyon where avoiding saturating the light shaft results in very little light covering the surrounding rock. Pulling back even a little shadow detail without trying to look unrealistic really pushes the dynamic range even of a D810 at base ISO.

 

You might want to work out how little stray light needs to be bounced around the mirror box and onto the sensor to make an image-plane DR of more than 12 stops an impossibility.

 

That does affect black level, and is something that should arguably be adjusted for up-front (or using ClearView or your choice of haze reduction technology). That's still something you can - mostly - correct for, because it's a low spatial-frequency phenomenon. If you still want to differentiate similar tones, you still need the dynamic range.

 

As with resolution, it's not everything - very nice shots have been taken on Canon bodies that, for years, had severe dynamic range capture restrictions compared with the Sonikon sensors. You don't have to capture 14 stops to work here, but it helps.

You don't buy a D850 expecting it to have lower noise. The D850 will make a better mural when used a low ISO's. For low noise and good image quality the D750 with only 24mp is hard to beat. Unless your shooting humongous prints you don't need a D850.

 

There are reasons why you may want to use a D850 other than its high resolution. The AF points covers a larger area of the frame so composition doesn't have to be compromised as much in many situations. I shoot a lot of verticals of people and the cross-type sensor points of the Multi-CAM 20k can be placed on the face with very little extra room at the top whereas with the D750 the furthest out points are linear and not that far off center. Another feature is the D850's support of Nikon's radio based flash control system. I have one third party radio trigger system which is moderately (but not completely) reliable and prefer Nikon's system. Other things which are missing from the D750 include the electronic front curtain shutter (mostly relevant to landscape photos with long lenses, but also with the newest implementation applicable for some hand held work), release+focus priority mode, auto AF fine tune (and more memory slots for fine tune settings), larger body etc. The D750 is a great camera especially for the price but one may still want the D850 for its features not just to make wall-sized prints but simply for everyday things such as getting shots in focus, a reliable and easy use of remote flash etc. While the D850 isn't optimized specifically for high ISO work there are many reasons why one might use it for high ISO because it can focus in low light and it may be the camera that you have in hand when high ISO is needed.

 

As the new AF system is particularly good in low light with wide-aperture lenses, it is relevant to wedding work (and many other types of events). While I moved to larger battery powered (400Ws) flashes for portraits outdoors in bright conditions, at other times when it is cloudy or there is not quite as much ambient light, a radio-controlled speedlight on a stand with an umbrella is easier to use, lighter to carry, and provides automation which means one can just shoot and move about for different shots with less sweating (and it means one doesn't need to have a car full of gear). I try to avoid extra battery management and Nikon's radio triggering system draws power from the camera and flash batteries and so there is no extra maintenance of batteries for the triggering and receiving systems. Furthermore the SB-5000 is compact and powerful and comes with built in active cooling so you don't need to be concerned with overheating if there is more ambient light. And, if you don't use a flash, being able to shoot at ISO 64 is just great for shots made in bright sun: the tones are smooth, one can push midtones up reducing highlight contrast, and manage the scene contrast in an easy way. ISO 64 also means you can shoot at f/2 or f/1.4 in bright ambient light conditions.

 

Of course the D850 is expensive and budget considerations are frequently present.

Over the weekend, I got together with a friend who uses Auto-Area AF a lot. He already has both the D5 and D500. He quickly noticed that on the D850, auto area is on par with the D500, but those cameras are not as fast and accurate as auto area on the D5. Since both the D5 and D850 are FX, that eliminates one difference.

 

Most likely Nikon is using different firmware on their top-of-the-line D5 or the strong battery on the D5 is making a difference.

or the strong battery on the D5 is making a difference.

 

Again, I'm hoping that someone gets to chance AF performance with a battery grip - both because of reduced black-out time and because of lens drive speed. I'm not expecting the D850 to match the D5, but if it makes a difference it's prioritise my decision to get a grip, as it were. Since there'll be at worst a significant step up from my D810, I'm not losing too much sleep about it not being quite Nikon's best - especially if it has to run with different accuracy requirements to hit the pixel count. Nikon have, I believe, been known to use the same hardware at a lower clock speed in prosumer devices, presumably decreasing cost (and power consumption).

 

[There are reasons why you may want to use a D850 other than its high resolution. The AF points covers a larger area of the frame so composition doesn't have to be compromised as much in many situations.

 

That's absolutely true. Indeed, the reduced coverage even of the MultiCAM 3500 variant in the D750 compared with the D810 points me towards the D810's version for extra spread despite the low light limitations. I absolutely value the dynamic range at minimum ISO of the D810, so combining that with (what I believe to be) some higher-ISO improvements in the D850 helps me, especially with reduced shutter slap. I'm not going to turn down 4K video. Compared with a D750, I'd like the buffer capacity, the faster store, the 1/8000s shutter, the AF joystick, the separate AF-On button (so I can still get at the other control points), the bigger viewfinder, the focus shift, the touch screen...

 

I have nothing against the D750. Although having been using a D700, then D800e, then D810, usually with an L plate attached, and an F5 otherwise, it did feel very plastic - which also means it's lightweight for transit, and I've no reason to doubt its robustness, but it doesn't add much stabilisation on its own. But there are several advantages even to a D810 (and counter-advantages), and quite a few more so with the D850.

I've been reading all that's been posted here. Now that I know LR 6.13 will come with ACR for the D850, I can't wait to be utterly amazed as I step up from a D700 to the D850 :)

Well, you'll probably be amazed that there's more pixel level noise, that you might be missing focus more often (because there's more resolution, so you can see when you miss), and if you're anything like me you'll be amazed that they swapped the + and - buttons with no way to override - that completely messed with my chimping process when I tried to use a D800e and D700 alongside each other. Oh, and I was amazed by where they put the AF mode and metering area buttons. But there are lots of upsides too. :-)

 

(One I didn't realise about the D8x0 series until I'd used it a bit is that the D8x0s have a separate aperture lever motor, so if you have a variable aperture zoom, hold down DoF preview, and zoom it, you'll get a constant aperture. The D700 leaves the lever where it was when you hit the DoF preview, so the aperture changes as you zoom unless you release the DoF preview button and try again. How's that for an obscure thing to be amazed by?)

the D850 has a lot going for it. Obviously it appeals to those chasing resolution. But it is also somewhat of an all-purpose body, in that it has some speed, some performance capability, and Nikon's best implementation of video so far. The increased AF coverage is indeed a selling point, at least for me, since the only other performance-worthy FF DSLRs i know of which can also do this is the Sony A9. (i know you can get wide frame coverage on various mirrorless cams and the D500, but those are APS-C).

 

when the D850 specs came out and i saw the DPReview write-up, i thought, hmmm...you could almost use it like a combination D5 and D800, by not always shooting at full-res (which is complete overkill for most action photography). in fact, in 25mp mode, it should have similar noise characteristics as the D5, at about $3k less.

 

of coursde, this is somewhat too good to be true., right? you dont get exact D5 performance, because Nikon uses higher-voltage circuitry in its flagship bodies (going back at least to the D200). that's why the D750 isn't quite as good as a D3s in most AF aspects except maybe tracking. So we should expect some sort of tradeoff for getting a consumer body.

 

The 850 is interesting in that its main strength and weakness are the same thing: that huge sensor. that's a strength because of impressive image quality and crop-ability, but a weakness in that all those megapixels have to be carefully micromanaged, which could alter one's shooting style.

 

the other downside, as Thom Hogan has reported, is that the 24-70/2.8 and 70-2o0/2.8 II lag a bit on that sensor. So if i was thinking i can save $$ vs. a D5 or A9, there go the savings because i would have to upgrade my two main event lenses -- a problem i dont havewith my D750.

 

I'm pretty sure I wouldn't mind having a D850, but my gut also tells me i don't need one at this exact moment. Still, it seems like a special camera with as much mojo as Nikon can currently muster, which will make many happy. i may check back in after i win the lottery or closer to end of life cycle.

 

oh, one guess about the AWB differences: the more neutral picture style was probably engineered as a nod to videographers. i believe the camera has 4k although not oversampled 4k like the A9.

Edited November 2, 2017 by eric_arnold

The color neutrality of the D5 and newer cameras (in AWB) is a very good thing in my opinion (you actually get three AWB options, in terms of how well it compensates for color tempererature variations in lighting; I use the middle one which produces more neutral results than the D810 but it doesn't force warm lighting into perfectly neutral so it leaves a bit of warmth to the image). It also helps when shooting in the arctic night as the light can be very blue late at night and with the D5 the results look good straight out of the camera, whereas they come out blue with the D810 and earlier cameras (in AWB). Also in tungsten lighting I felt older Nikons were too warm by default. The drawback of such changes is that different cameras now produce different output (though again you can adjust the AWB settings for a better match).

 

With regards to the 24-70G vs. E, the earlier version produced images with strong field curvature at the 24mm end which made it very difficult to place the focus on a large group shot and get even sharpness across the image. With the E version this has been corrected, making me very happy as I no longer have to use a separate prime for groups. However, even 12MP was enough to suffer from this issue, no need for 45MP to reveal it. The 70-200/2.8G II's main issue for me was its bokeh at medium to longish distances (double rings etc.). Many complained about the focal length shortening at close distances. The E FL version doesn't have these problems. The AF in the E FL version is quicker (though I don't know if is more accurate, just quicker and the stutter is gone) and the VR has the SPORT mode which I really like, as it behaves as it were not there, more stable viewfinder, less resistance to user will. The tripod mount is more stiff, and the lens is lighter and less front heavy. Optically I think the most important aspect is the improved bokeh behind the subject, the narrower field view, but lensrentals also report much improvement in the MTF at 135mm. Two drawbacks in the E FL version: the position of the zoom ring in the new lens is criticized by some, and the cost of the lens is high. But personally I really like both of the E versions of these zooms. I could use the old ones without problems but for my use the new versions are better. I wouldn't necessarily upgrade because of the D850 but independently of that, if you feel the need.

 

The Nikon 70-200/2.8 E FL does really well in lensrentals comparisons by the way, e.g.,

 

An Update and Comparison of the Sony FE 70-200mm f/2.8 GM OSS

 

MTF Lens Tests of the New Tamron 70-200 f/2.8 SP Di VC USD G2

 

I like that test site because they measure and analyze multiple samples. But optical bench testing is done at infinity focus so close-distance performance should be verified by another source.

Edited November 2, 2017 by ilkka_nissila

FWIW, updating my 70-200 from the mk2 to the E FL version is on my to-do list, partly with the D850 in mind.

Speaking of the 24-70mm/f2.8 E AF-S VR, I tried that on the D850 recently. At 24mm, there is a bit of barrel distillation, as expected.

 

If you pixel peep, you can see some clear purple fringing at the edge of the white uniforms towards the edge of the frame. 46MP means the image is greatly magnified. Overall my main reservation of that 24-70 E is big size; it is a bigger lens than the 300mm/f4 E PF AF-S VR, and it is very expensive for a 24-70. Otherwise I am quite happy with that lens.

 

 

Edited November 2, 2017 by ShunCheung

Thanks, Shun - it doesn't look too bad, and something that DxO (or your choice of lens-aware converter) ought mostly to sort out.

Hmmm ... al this makes it harder. So, more noise than my current D700... My 24-70G and 70-20 VRII are no longer "good enough" (I read Thom's blog too now..). There is no way I am going to upgrade to the 70-200 E and if the 24-70 E is even bigger than the "old" 24-70 .. good grief. I just checked the prices of the new 24-70 and 70-200.. YIKES! I might be better off with a used D810...

My requirements are not as stringent as others' may be -- my most basic requirement is a D8x0 with the AF speed of the D500, which the D850 meets. My primary lenses for this are the Sigma ART lenses (24-35-50) mostly in well-lit conditions. It would be rare for me to go beyond an ISO of 800. So I do plan on getting one as soon as I can put together the money and obviously depending on availability. My target is Q2 (or maybe Q3) 2018. Edited November 2, 2017 by photo_galleries

@kevin_beretta: have you compared D850 photos downsampled to 12 Mpix when you concluded that the D850 has more noise than your D700? I ask because I do not know myself how well the D700 fares today, only that it has aged well. Personally, I am very happy with the five year old technology in the D800, which still produces excellent photos (I prefer that to the more modern D610(D750). Having said that, the D850, will most probably be my next camera hence my interest in this thread.

 

Perhaps Shun can also test some of the lenses that he has access to and give his input on which Nikkors are up to the job as well as post comparisons? I do not know If Shun has the previous 24-70 to compare with, but I am sure that I am not the only one that wants to see how the 70-200/2.8 VRII holds up. Perhaps it is still good enough for sports and portraiture but not for landscape?

 

Regarding noise; for the third time (third time's the charm, they say), are they shot with in-camera High ISO NR set to off, low, normal or high? I only recently discovered that this setting affects the RAW files as it is available when only RAW is selected. I have not tested it on my cameras to see what effect it actually has on RAW files but I do what it does to small jpgs. As so much of the discussion here is about noise, I think there is value on knowing how these were shot. Same goes for a brief description on the settings used in PS to downsample as results vary with the settings. Sorry for nitpicking, but the better the testing methodology is decribed, the less questions will be asked about the results as they will speak more clearly for themselves.

 

Again, providing a few selected RAW-files are so valuable as everyone can be their own judge on what is good enough for them. :)

I never had the High ISO NR to default (normal?) for the entire life of the camera. I didn't know it would make a difference in RAW mode ... Hmm. And if the 24-70 G is not up to snuff anymore, what about the 14-24? Grrr... so many choices and decisions to make. I think what I will do is rent a D850 for a day and stick it on a tripod and shoot a lot of my lenses on it and then make a decision.