Nikon D850, Early Impressions

[ShunCheung]

The mega pixel game is getting silly if you ask me. On the FX sensor, the difference between 36MP and 45MP is very small, and diffraction is going to be a bit of a problem from around f8 on the D850.

 

Personally, I would buy the D850 if you are after better AF (than the D800 and D810), better video capability and perhaps some of the new features such as auto focus stacking ... and XQD cards, as I dislike both CF and SD.

 

Concerning lenses, I am using the older 70-200mm/f2.8 AF-S VR II, and that is still an excellent lens today. Whatever lens that works great on the D800/D810 is likely great on the D850. E.g. my lowly plastic 18-35mm/f3.5-4.5 AF-S shines on the D850.

 

Here is a sample using the Sigma 35mm/f1.4 Art, at f8 and base ISO 64. You can read the larger prints on those business cards.

 

 

[kevin_beretta]

Thanks Shun, appreciated.

I think for me what attracts me to the D850 is the quiet operation possibility, the fact it has a tilt screen and the stacking. I think I am spoiled in terms of AF as I use center point basically always on the D700 and it works in near dark, so I never had any need for more. I next to never shoot sports, running kids or pets. Lots of portraits and landscapes, some street scenery. So pretty basic stuff. I would use my marco lenses and the bellows more if I had more pixels to play with. Being able to crop is a really good thing too. My main goal is travel photography ... My next purchase is LR 6.13 with the ACR update, then the D850. I already have more glass than I need (35 1.8 AF-S, 50 1.4 G, 85 1.4 G, 105 2.8 AF-S, 14-24 2.8, 24-70 2.8, 70-200 2.8 VRII - All Nikon). My real challenge is what to leave home when I am traveling. I think I'm going to settle on the 35, 24-70 and 105. I recently figured out how to manage files in the field without a computer (Android tablet, USB hub and 2 HDDs) so everyhting can pack realtively small. The D850 might be overkill... I shot some of my best pics on a borrowed Df. The stuff that camera put out was magical ...

[Andrew Garrard]

Hmmm ... al this makes it harder. So, more noise than my current D700...

 

People have kind of said this, but to add personal experience: comparing pixels with pixels, at the same ISO a D810 is noisier than a D700. If you down-scale the image to the same size, the D800 and D810 have, in my experience, most of a stop advantage over the D700 at higher ISOs - slightly more noise at each pixel goes away when the extra pixels are averaged together. This isn't surprising - less light is hitting each pixel, but given the same total amount of light spread over more pixels, the technology improved. If you want to work out whether you can create a print nearly twice as large with a D850 and retain the same detail as the D700, the noise at the pixel level may bother you; if you're printing (or viewing) at the same size, the D8x0 series are all significantly better than the D700. I got a D800e when they came out and shot it alongside my D700, and there was no contest.

 

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...

 

Well, "good enough" for what? You can certainly see more lens aberrations if you zoom into a pixel level on a higher-resolution sensor, but if you reduce the image size to what you had before, it will look the same. Arguably the biggest danger is that when some areas of the image look sharper, it shows up how much softer others are.

 

I use a 70-200mk2 with a D810. It's not quite a D850, but that's only a 1.1x difference, rather than 1.9x for a D700. At f/2.8, it tends to be a little soft and it's hard to hit perfect focus. At f/4 it's perfectly good enough, in my experience (and much better than older versions). Will I upgrade to the E version at some point? Yes, because I'd like to shoot at f/2.8 with less of a compromise to image quality without needing my 200 f/2. On a D700 you can shoot a lot of lenses wide open without worry; on any D8x0 body there are very few lenses that can hold up without dropping off a little - but that's only a problem if you want to use the extremes of the lens and still make the best of the sensor. Put another way, I'll be astonished if the 50mm f/1.8 AF-D isn't very sharp on the D850 by the time you're at f/6.3 - the problem is how it looks at f/1.8.

 

Sometimes the difference should affect your lens choice - I ditched my 28-200mm AF-D G that was one of my most-used lenses on a D700 because I could get better results with "digital zoom" on a less flexible lens. I don't think that's true for the f/2.8 AF-S lenses, but their imperfections are certainly more visible.

 

I'll also point out that some lens-aware digital post-processing can do wonders. My Tamron 24-70 VC is absolutely not sharp at the corners in JPEG, but gets most of the detail back once DxO has played with it. The same is true of the 14-24 Nikkor. You always gain a bit of noise through the sharpening process, and it's never as perfect as solving things optically - but then there's no perfect lens.

 

Dare I point out that the 105mm micro is known to be a little lacking at longer distances? :-)

[ShunCheung]

Since I mentioned the 24-70mm/f2.8 E AF-S VR (82mm filter thread), here is a side-by-side image with the 300mm/f4 E PF (77mm filter).

The 24-70 is also a lot heavier, since it is a zoom with more elements and the zoom mechanism.

 

[rodeo_joe1]

"Pulling back even a little shadow detail without trying to look unrealistic really pushes the dynamic range..."

 

- Bit of an non sequiter argument Andrew.

 

The lack of realism is purely a consequence of trying to pour a quart (of subject brightness range) into a pint pot (of display contrast). No amount of increase in captured dynamic range will improve that situation, and in fact may make it worse.

 

I also fail to see why a low spatial frequency is correlated with low brightness - only if some lossy compression algorithm is used, maybe.

 

My original argument still stands. As long as a lens system and camera dark-chamber of finite size are coupled with a reflective sensor, there will be a practical limit to the brightness range at the sensor. And I would argue that the limit has already been reached for an "average" subject. Unless and until dark chambers are properly AR flocked, and not just painted with a vaguely black paint, and the internal flare of lenses is reduced to near zero, then there is very much a limit to how much "dynamic range" can be utilised.

 

Not to mention how any current display can show it. Surface reflection from current screen technology is high enough to make claims of thousands-to-one contrast ratio laughable for a start. Nobody wants to sit in a black-walled, fully darkened room just to watch the telly!

[ilkka_nissila]

The human brain does local contrast enhancement, local color contrast enhancement, and opens up shadows as needed to see what the person needs to see. Photography and printing also can benefit from carefully considered local adjustments.

 

As for whether the extra dynamic range is real-world useful, in my experience, it is, in the sense that it makes the images look smoother and more beautiful. I haven't purposefully carried out testing by shooting identical images with two cameras simply because I'm more interested in getting the images that I need/want than testing and comparing equipment, but I will try to illustrate this at a future opportunity. I know that subjectively D810 ISO 64 images in sunlight look smoother than D5 ISO 100 images. I find that this perception of different quality is increased if I lift the midpoint of the tone curve to lighten midtones and decrease highlight contrast which is a technique I use a lot. I'll see what I can do about making a real-world comparison illustrating the benefits of high low ISO dynamic range without making excessive adjustments that might make the image look unnatural. I feel it should be easy to do actually.

[ilkka_nissila]

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...

 

In my opinion this isn't quite the right conclusion to make. The D850 should show equal or a bit better results than the D810 with any lens (at the image level, not necessarily at the level of individual pixels). The autofocus module is much better and so the real-world results relying on autofocus should be significantly better especially if you use the outermost points or shoot in low light (or both). There should not be any issue using the 24-70G or the 70-200/2.8G II that you didn't have using current bodies. The 24-70G has strong field curvature at 24mm which can be an issue but that issue is plainly visible already with 12MP cameras. While the 24-70/2.8E is larger and heavier than the G version, it does feel very well balanced in use. And the 70-200/2.8E is actually lighter than the G II and is less front heavy, which makes me often lift the camera with lens attached with just one hand. However, I agree these two new lenses are very expensive and it may make no sense to update to them from a value or personal or business economics point of view.

 

However if you choose between the D850 and D810 (staying with your existing lenses, which makes perfect sense), I would still go with the D850 because of the Multi-CAM 20k AF module. That is, if you use AF.

[Andrew Garrard]

I also fail to see why a low spatial frequency is correlated with low brightness - only if some lossy compression algorithm is used, maybe.

 

I clearly didn't explain that one very well. I'll try again.

 

The argument is, I believe, that if some portion of the image is of very high brightness compared with other portions, this will cause internal reflections that increase the brightness of the darker regions captured.

 

My counter-argument was that these internal reflections are likely to be diffuse (low spatial frequency), and that, while they'll raise the minimum intensity level at the sensor, they'll be adding a relatively constant amount to all the pixels. I would expect the result to be similar to raising the black level of the image, and that it could be compensated for by adjusting the black point (or, more locally, using "clarity" filters intended to remove atmospheric fog). Technically I guess this makes the significant measurement what DxO calls "color sensitivity", which should measure your ability to differentiate tones at various points in the tone curve - although it's probably a mix of both, since in my experience lens coatings and internal baffling is such that I've not seen huge issues with internal reflections. Since raw files are typically linear, the significant factor is the ability to avoid noise in the read electronics regardless.

 

Surface reflection from current screen technology is high enough to make claims of thousands-to-one contrast ratio laughable for a start. Nobody wants to sit in a black-walled, fully darkened room just to watch the telly!

 

True, although my living room is certainly quite dim when i'm watching telly. Most HDR TVs can't support large areas of high brightness - the intent is that the average brightess level remains within a small factor of traditional screens (which were standardised at 100nits, with wide variations in reality) and the ability to do very bright regions is restricted to small areas of the display or limited durations. Making the sun and some specular highlights, or car headlights, look very bright doesn't necessarily light up the entire room. The same is true in a photographic capture - just because I don't want to blow out the edges of some clouds doesn't mean I'm expecting large areas of the scene to be bright.

[rodeo_joe1]

"I would expect the result to be similar to raising the black level of the image,"

 

- Not similar to, it is raising the black level. And of course flare and scattered light are a low spatial frequency phenomenon. I don't see how that's at all relevant. In fact overall fogging has a zero spatial frequency within the bounds of the image.

 

".... and that it could be compensated for by adjusting the black point..."

 

- How, exactly?

Let's say you have 0.05% of the maximum sensor brightness reflected, diffused and uniformly re-reflected into the shadow areas of the sensor. That gives a contrast ratio at the sensor surface of 2000:1, or about 11 stops. That's it! That's all the DR that the sensor can possibly capture - full stop. You can't have negative light that subtracts from the base fog level.

 

Yes, you can lower the black level to zero display brightness, but you can't reconstruct image detail that (might have) existed at a brightness level well below that of the fogging.

 

It's like trying to dig any signal out from a level below a masking white noise. SNR is what it is, and there's nothing that can alter it apart from getting rid of the noise at source.

 

So, given that there must be some practical limit to the brightness range presented to the sensor. What good does increasing theoretical sensor DR do?

 

By all means increase A/D bit depth, because that increases the degree of differentiation between brightness levels. However, the image brightness range is effectively limited by the optical system in front of and surrounding the sensor.

[rodeo_joe1]

"I know that subjectively D810 ISO 64 images in sunlight look smoother than D5 ISO 100 images."

 

- What do you mean by 'smoother' Ilkka?

 

Such nebulous subjective descriptions aren't evidence of anything. Because to me, 'smoother' could mean showing less detail and making things more plastic-looking.

 

For example: The skin tones and textures in CGI renderings look smoother to me, but that just makes them appear less than realistic.

[ilkka_nissila]

Smooth means (relatively) free from noise (as in smooth tonal gradations). D810 images (correctly focused) are clearly more detailed than D5 images (both at low ISO); at the same time they have less noise at low ISO settings.

 

If you prefer measurements you can just look at dxomark they'll show that mid-gray SNR at ISO 64 for D810 is 46.3dB and for ISO 100 on the D5 it is 44.8dB; dynamic range 12.26ev vs. 14.76ev. It's easy to find evidence for the extra detail in D8x0 family images for example look at dpreview's studio comparisons. My subjective observations in this area of sensor performance of these particular cameras align with dxomark's measurements. Usually when measurements are correctly done they agree with subjective observations from a person who is experienced in evaluating image quality and understands what they are evaluating.

 

What gives the D5 the edge is its autofocus system, however, if one is shooting subjects which are not moving a tripod and live view is what I use, and in that case the D5 has no advantage compared to D8x0 family cameras, unless of course the D5 is the camera in your hand when the subject presents itself in good light and alternatives are not available.

Edited November 4, 2017 by ilkka_nissila

[Andrew Garrard]

"I would expect the result to be similar to raising the black level of the image,"

 

- Not similar to, it is raising the black level. And of course flare and scattered light are a low spatial frequency phenomenon. I don't see how that's at all relevant. In fact overall fogging has a zero spatial frequency within the bounds of the image.

 

Only that it implies relatively easy correction. Local contrast variations are harder to correct.

 

".... and that it could be compensated for by adjusting the black point..."

 

- How, exactly?

Let's say you have 0.05% of the maximum sensor brightness reflected, diffused and uniformly re-reflected into the shadow areas of the sensor. That gives a contrast ratio at the sensor surface of 2000:1, or about 11 stops. That's it! That's all the DR that the sensor can possibly capture - full stop. You can't have negative light that subtracts from the base fog level.

 

Yes, you can lower the black level to zero display brightness, but you can't reconstruct image detail that (might have) existed at a brightness level well below that of the fogging.

 

It's like trying to dig any signal out from a level below a masking white noise. SNR is what it is, and there's nothing that can alter it apart from getting rid of the noise at source.

 

All true. But let's say you have 14 stops of dynamic range. You have a large highlight that results in uniformly raising the base brightness level at the sensor to half of the maximum brightness (say digitally 8192, or 2^13). Result: So long as you can distinguish level 8193 from 8192 as effectively as you can distinguish 1 from 0, you've still got 13 stops of dynamic range left. That's a fairly big qualifying statement, but on the other hand raw files are traditionally linear. I'd claim the actual influence of a small highlight on the sensor tends to be smaller.

 

So, given that there must be some practical limit to the brightness range presented to the sensor. What good does increasing theoretical sensor DR do?

 

I'm standing by my "small highlights in a very dark room" argument, because I've seen it in images (note the banding in the walls) - but I'm very under-slept, and that may be making me less scientific than usual.

 

By all means increase A/D bit depth, because that increases the degree of differentiation between brightness levels. However, the image brightness range is effectively limited by the optical system in front of and surrounding the sensor.

 

I'd say both. Dynamic range tends to imply well capacity and read noise. These things are good. I notice the A7R3 has some apparently impressive (slight) improvements over its predecessor, making it better at ISO 100 than a D850 (but less good than the D850 is at ISO 64). Both the A7R2 and 3 have better high-ISO behaviour than the Nikon, sadly (in the graphs I've seen). I'm more than a bit bought into the F mount and what you can buy and hire for it, though, so I don't think Nikon need to worry about my custom just yet.

[ShunCheung]

I managed to borrow a D810 and went back to the same location for some high-ISO test images. I used exactly the same 24-70 E VR lens but the focal length is a bit different. Even the D810 is not bad at ISO 12800

 

 

[bob_bill]

Shun, if I read the description correctly at B&H they indicate 1. with the grip and battery, 2. in crop mode and 3. using silent live view 30 fps is possible. Have you tested that. 30 fps sure seems amazing to me.

[ShunCheung]

Shun, if I read the description correctly at B&H they indicate 1. with the grip and battery, 2. in crop mode and 3. using silent live view 30 fps is possible. Have you tested that. 30 fps sure seems amazing to me.

30 fps? Sounds like you are shooting video.

 

I don't have the MB-D18 grip.

[bob_bill]

Shun, that's what I thought on first read. Here it is:

Benefitting the sensor is the apt EXPEED 5 image processor, which affords a wealth of speed throughout the camera system, including the ability to shoot continuously at 7 fps for up to 51 consecutive 14-bit lossless compressed raw files in a single burst. When working with the optional MB-D18 grip and EN-EL18a/b battery, this shooting rate can be increased to 9 fps, and up to 30 fps shooting is possible when working in a DX crop mode during Silent Live View.

Haven't seen too many typos at B&H so wondered if it was true.

[Andrew Garrard]

I don't think the 30fps mode requires the grip. It captures "JPEG normal" images at 3600x2400 with a DX crop and fixed focus/exposure - effectively video, but all i-frames and only for 3s. You don't get the usual JPEG vignette control or distortion correction either.

 

I'll probably try it in an emergency, but capturing 3840x2160 video at the same frame rate without such a small time limit or sensor crop, then pulling frames, feels more useful. If it let you fill the internal buffet with raw files at that rate (even with the rolling shutter) I'd be a lot more interested.

[Andrew Garrard]

Fwiw, i button (or A? button if you have an iPhone) in live view, then look for "silent live view photography mode 2" if you want to try it, Shun. The manual says nothing about needing the grip to do it. I assume the grip is needed for 9fps to power the shutter/mirror mechanism at speed (not doing anything for this mode), although I've never quite understood why Nikon found the need to have their frame rate bound by the battery and in all their shutter redesigns haven't managed to make one run at full speed off the internal power source.

[mike_halliwell]

Andrew, cynically, one thinks it makes you buy the extortionately expensive battery grip to get a feature that should be there already.

 

The 'mark-up' on grips must be HUGE. It also wouldn't surprise me if there's a custom chip in the grip that's quite hard to clone too, just to slow down the copiers! Same as the batteries having to communicate with the body and preventing clones being used at 1/5 the price.

 

Grip is £370 and a new 18b is £199. That's nearly $750!

 

Sure, Nikon's got to make money, but deliberate crippling the fps seems a funny way to do business.

 

Equally, it's not as if you can't draw enough amps from the normal battery.

Edited November 18, 2017 by mike_halliwell

[ilkka_nissila]

The EN-EL18(a/b) battery has higher voltage which is used in the D850 to move the mirror at higher speed. This is necessary to do the 9fps on an FX camera with a reasonable blackout time. Remember also that the processing the camera has to do for 45MP images at 9fps must be consuming a lot of electrical power as well. However, the camera designers said in an interview that the reason for requiring the larger battery is to run the mirror motor at higher speed.

 

The D500 presumably has a smaller and lighter mirror and it has a comparatively long blackout time (it runs 10fps on the smaller battery). I recall the D500 blackout at 10fps is 45% longer than on the D5 at 10fps. This means the AF sensor is not seeing the subject for a longer time before the shutter opens and you also get less visibility in the viewfinder.

 

There are (to my knowledge) no moving-mirror FX DSLRs that achieve > 7fps with a small battery. As for the pricing, it is necessary for Nikon to make a profit, otherwise investors will eventually abandon the company. If you want 9 or more fps for less money you can get the D500. It’s not like a lot of people ”need” 9fps & 45MP at the same time, the amount of data must be staggering. So when a feature is not really a need for most D850 users, but something of a want it is a perfect place to ask for exta money, without doing too much injustice.

 

Some of us who are getting the D850 also have a D5 or D4(s), so we are at least spared of the extra cost of the battery and charger. The grip is about the same cost as this type of grips were in D800/D810 (five + years of use from that grip already) and D700 before.

Edited November 18, 2017 by ilkka_nissila

[ilkka_nissila]

The counterbalanced shutter in the D850 is also a lot more complicated than the ones in the earlier models. It would not be surprising if it consumes more power, though the mirror is said to be the limiting factor.

[Andrew Garrard]

For some reason I was convinced that Canon had managed to go faster with an unexpanded 1v than they actually did - although I'm inclined to think that cranking the film should be the limit there. Canon did manage to drive the original 7D at a rate that needed a grip for the D300, but since Nikon fixed it with the D500 I can't argue.

 

Still, in all the many years of shutter development, I would have expected them to find a way. Of course they did, with the 30fps mode - kind of.

[paul_b.1]

Thanks Shun. I thought D810 was better at low ISO. Obviously is not .

 

"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"

 

As far as can I remember (from min 20 years ago) it was always the same. And it was a Canon lineup who make me jump to this conclusion : only top pro cameras (both Canon and Nikon) deliver the best AF, best auto exposure and best flash exposure algorithm , noticeable better than the rest of the lineup.

Edited November 18, 2017 by paul_b.|1

[ShunCheung]

Some friends and I have been comparing the D850 against the D5, D500, etc. Even though they have the same AF module, the D5 still has the best AF among all three, and that would extend to all Nikon SLRs. The D850 has very good AF, but not quite as excellent as the D5.

[heimbrandt]

While a lot has already been written about the D850, it would still be interesting to read the full review and your thought on its strong and weak points. Having read many reviews I still would like to learn more about its high ISO performance, its DX crop mode and lot more that you have observed during your time with the D850. Having a D800E, it seems like the natural upgrade path, when it is time.

 

When do you expect the review to be ready to publish?