Photons To Photos published first Nikon D850 dynamic range chart


Based on some available NEF files Photons To Photos published their first Photographic Dynamic Range (PDR) chart for the Nikon D850 (click for larger view):

Pre-orders: B&H | Adorama | Amazon | BuyDig | WEX | Jessops

Nikon D850 Facebook Page | Nikon D850 Facebook Group

Nikon D850 directory

This entry was posted in Nikon D850 and tagged . Bookmark the permalink. Trackbacks are closed, but you can post a comment.
  • Mark

    Looks great – the D810 is so outstanding, to be able to match the dynamic range with more definition is an achievement

    • BG

      Nah.

      • Mark

        My point is that is the D850 can match the range of the D810 with better definition (and possibly better noise at higher ISO), this is an improvement.

      • Mark

        What are you refuting ? The chart or the claim that having the same dynamic range with more pixels is an achievement ?

        • BG

          I think maintaining the dynamic range was the minimum of what could be expected if they wanted to maintain the D8X0 series’ position as a prime landscape camera. Improvement would have been an achievement. But maybe I’m being overly critical…

          • CERO

            Nah, Its an archievement, considering all the other things that they shut together in the camera.. from 4k, to the AF, the nice firing speed, etc..

            • BG

              AF and frame rate are not of primary interest to me. Landscapes mostly don’t move all that much… and for the occasional shot of a living subject most modern DSLRs have more than adequate AF & frame rates.

            • CERO

              So, because its not your primary interest, therefore its irrelevant for everyone? lol.

            • BG

              No. It’s very relevant for many people. But it’s completely irrelevant to a discussion of sensor dynamic range.

            • CERO

              Except We’re not explicitly talking about dynamic range, since you were complaining that it should have BETTER while having almost 50% more sensor density.

            • BG

              You’re the only one in this thread who’s not talking about dynamic range.

            • CERO

              Jesus, are you a moron or just trying to troll?
              What part of “explicitly”?
              They increased the resolution and maintained the DR of the D810 which is oustanding.
              Specially where almost all the competition gets worse DR at the moment they increase the resolution.

          • alexdesign

            you know you cannot really go higher with only 14bits? so saying no to the “maximum reachable” just doesnt make any sense. They also do it on a 46MP sensor, it’s a big achievement

          • akkual

            Well, you are wrong. D810 is actually worse in DR than D750 at 1:1 pixel ratio. What makes D810 so great is the “oversampling” due to higher megapixel amount for 99% of usage. There is this “oversampling theorem”, which for example allows us to encode music with just 1-bit and still decode it back to high definition sound… we just use 2 MHz sampling frequency to get tons of 1-bit samples. The same works for cameras in megapixels. That is, if this camera has the same DR as D810 at 1:1, you will get about 0.5-1 EV more “usable” DR from D850 when printing or exporting in same size.

            • Mikycoud

              Glad someone finally explained that well!

            • BG

              Except he’s probably gotten it wrong…

            • Mikycoud

              If you say so. I for one hold his transposition as correct.

            • BG

              Bill Claff (the guy who does the measurements) just confirmed further down in the discussion that photographic dynamic range measurements are normalized. So I don’t think your explanation holds in this case (but I do understand the concept).

            • Piooof

              Well, Bill Claff’s PDR is normalized with respect to sensor size, not sensor MP count. So akkual’s explanation seems correct.

            • BG

              The explanation is correct but the numbers already include his considerations, so the assertion that the sensor is better than what the numbers say is incorrect.

            • Piooof

              Since sensor photosites are smaller on the D850, it would be a real feat to get the same pixelwise DR. But the D850 sensor is still a bit better than the D810 at high ISO, and most lower ISO values are lacking from this graph. Wait and see for a complete assessment.

            • Chris

              Oversampling has its own limits. You can’t oversample a bunch of 0 to get a 1.

              Oversampling only works when the individual pixel can validly see through noise.

              In real world, when you are shooting high dynamic range scene, it will be better to use bracketing.

          • Just because something was expected doesn’t mean it’s not an achievement.

    • CaMeRa QuEsT

      Exactly!

  • *estimated.

    • Ironheart

      Using actual data from publicly available raw files. As more ISO values become available I’m sure Bill will add the data points. It’s highly unlikely that the final data will change much.

      • hussey

        Yeah, I’m not sure how they derive these numbers using just random raw files.

        • BlueBomberTurbo

          Data is data. The only difference is that you might not get a smooth curve from a non-standardized test image. Same results, just arranged differently.

        • NJP

          Read his website. He explains the entire process. Its not exactly easy to understand but after hearing interviews with the guy and going back to the site, it certainly appears to be a lot more accurate than Dxo

          • Piooof

            I doubt it’s more accurate; it’s different (and less condensed, thus giving more information).

        • akkual

          You can derive estimation of DR from RAW file in editing software like Lightroom. Take as evenly exposed shot as possible (Histogram has wide nice curve), then “pull” it -5EV and see how much details the highlights have and reveal. Then push it +5EV and see how much detail the shadows have and reveal. If they reveal noise and nonsense instead details, you can estimate the DR is under 10EV. If there is detail and not much noise, you can estimate the DR is over 10EV. This is what dpreview provides as comparison tool on their pages. Typically you will find that towards shadows there is more room to push than towards highlights. With D750, you can take almost dark frame at base ISO and push it +5EVs to usable photo.

    • BlueBomberTurbo

      Bill’s usually pretty good about estimates. At worst, the curve may be lumpy/spikey, depending on what samples are available, but the average shape of the curve is reliable.

  • JJ168

    Everyone want to be the first to get into some clicks to their site 🙂

  • Does not look good!

  • rafakoy

    “Estimated” lol, I love these absolutely meaninlgess charts.

    • Ironheart

      The final values won’t differ by any meaningful amount. Raw data is raw data.

      • hussey

        But raw data doesn’t tell us what the scene was like. Would not the data change depending on the lens contrast?

        • CERO

          But you’re not comparing the “data” itself, but the results compared to the same “data” received from the same subject taken with a D810.

          • BlueBomberTurbo

            Why not? As long as you have data in the RAW from clipped highlights and shadows, it should be enough to figure out the DR. Especially when you have a decade or so of reference files to compare with.

            • CERO

              I still have no idea what you’re talking about.

              To make a perfect scientific comparison. They need to be shot (D850 and D810) in the same conditions, with the same lenses to the same subject.. then compare.

              Older data should be irrelevant and only used as older comparison information with less priority.

            • SF_Expat

              Bill is not measuring scenes but the range of data recorded, but any file that has data that somewhere in the file has the lowest value and some place that has the highest numerical values. Read up on how he does it, and why is is valid over a wide range of sensor types and image qualities. Any raw file can be used for data collection that matches other scenes with other files. He has been consistently, for years been the source for the charts that most reflect the results of real image files. Test a D3 now and it will match what was reported back when it was new, the data can be used to compare this one important characteristic regardless of model or age of the file.

            • CERO

              I never said anything about the “date” or the original file. I’m talking of the scene in question. Are they standardized on the same place with the same lens to produce similar comparable RAW data where only the variance is the sensor output/record ?
              Or the Raw data itself can be tested with any scene in question even if the next test has a different scene or lens?

    • Ivanku

      Photons to photons has a good track record, and wouldn’t be releasing this if there wasn’t at least some confidence behind the results. As long as this is labeled “estimate,” I’ll take it.

    • El Aura

      And don’t we love these people who make these absolutely clueless comments?
      (Ironheart & Ivanku already addressed the substance)

    • CaMeRa QuEsT

      I’ve seen Bill’s preliminary data before and it never strays far from the actual data. This guy is doing a great job helping the photo community and his work is to be commended at the highest levels.

  • Richard Haw

    OK…going to stop smoking and save some money for this

    • ZoetMB

      Don’t know how much cigs are where you are, but in NYC, I think the average is $12.50 per pack. At two packs per day, you can buy this camera in less than 19 weeks.

      • At two packs per day I don’t think ha can make trough 19 weeks 🙂

        • peter w

          died of cancer before that?
          You don’t mean…
          😉

      • Allan

        There has to be certified public accountants in your family. 🙂

        • ZoetMB

          No, but as I once said to an assistant who complained that she wasn’t an accountant when I asked her to track sales (it was a high-end, niche market, expensive product so the numbers weren’t large), “it’s not accounting, it’s arithmetic”.

    • Alexander Buturlakin

      That reminds me old joke…
      The young journalist (J) see the solid man (M) smoking cigarette near huge Manhattan skyscraper entrance, so journalist start asking the questions…
      J: Why do you smoke and how long?
      M: I like to have some pleasures and this is why I smoke during last 30 years…
      J: OMG… Do u know for money you spent for cigarettes you could buy this skyscraper!!!
      M: Well… I do smoke and this skyscraper is mine…

      • Richard Haw

        that made me chuckle!

  • Exynos

    On par with d810
    NIKON;
    i am; old tech

    • Jeffry De Meyer

      Wasn’t the d810’still the camera to beat by all other cameras?

      • KevHot

        It’s already been beaten though. Only chance it stands is as a back up body esp for someone with Nikon glass.

        • Jeffry De Meyer

          Which one has better range?
          The dxo list still has the 810 on top

          • David

            Jeff, these are just butthurt Sony trolls, look at their comment histories and move on…. these losers aren’t worth your time or energy.

            • Gosh1

              yes don’t waste Purity feeding such podgy infantile entities

        • CaMeRa QuEsT

          Tell us if you will, what FF camera has beaten the D810@ISO 64? Illuminate us, please!

          • BlueBomberTurbo

            According to Bill Claff, the A7R II’s DR at ISO 100 is extremely close to the D810 at ISO 64. As in margin of error close (0.15 stop). Now if you go ISO for ISO, the A7R II is neck and neck with the D810 till ISO 500, then beats it through the rest of the ISO range by around 1 full stop

        • bobgrant

          Spoken like a guy who can’t make enough money to buy a car from a few days of shooting. And that happens with my D810. Are you really THAT threatened that Nikon knocked it out of the park? Lots of great cameras out there, so go learn to shoot and stop worrying about specs.

      • Exynos

        A7r2

        • Jeffry De Meyer

          it is listed lower

        • tomskyphoto

          RAW files are poorer than those of the 810, more noise at base ISO, else IQ indistinguishable from camera to camera. I own both cameras. A7RII has a slight advantage at higher ISO regarding DR and noise.

          The rest about the A7RII? As a camera it’s a piece of s**t.

          • Stan Chung

            tomsky, even for video? Most guys I know use Sony for their video as they prefer it over Canon and quite obviously Nikon[crappy video AF but pros usually eschew AF]

            • tomskyphoto

              You’re right, video is a different story and an area where Sony is really quite good when it comes to a small rig with AF as mirrorless cameras have an inherent advantage over SLRs here. But then emphasis is on “rig” too; the poor ergonomics don’t really matter here.

              As I’m not doing video I was speaking for the stills part only, should’ve been a bit more precise about that. And apart from it’s great but not unmatched IQ there’s little where the A7RII can really excel. It’s a poorly built and designed slouch of a camera that’s unpleasant to use.

            • Stan Chung

              ok thanks Tom for explaining.

      • Don’t feed the troll.

    • CaMeRa QuEsT

      You are getting the same DR from smaller photosites, there IS an improvement, and, if it has the same DR@ISO 64 as the D810, there is still nobody else giving us that kind of DR@ISO 64.

    • David

      Guys, dont feed the troll. Look at his comments, he’s just a paid Sony flack.

      • brian valente

        I wish I was a paid Sony flack

    • Chris Phillips

      You are CLUELESS

  • Joe Russo

    I wonder if the files will still be ISO invariant?

  • Juraj Kosco

    Can someone explain the chart. Why does the 850 start at the 6th dot. Does that mean it’s short by about 2/3s of a stop in dynamic range at minimum ISO?

    • The preliminary results aren’t available for all ISO settings.
      So only the available ISO settings are shown.

      • CaMeRa QuEsT

        That was FAST! Kudos Bill, great job as always!

        The DR curve is looking exactly as I expected it will, given your findings on the D800, D810 and D500/D7500 sensors. As I wrote before, no apparent improvement is still an improvement, as the photo sites are smaller and there’s a huge rush of electrons going on at the sensor’s back, compared to the D810’s sensor. I guess this is the physical limit of the current state of the silicon photocell art, the next meaningful improvement will need the introduction of organic layer photocells.

    • JJ168

      Need more files. I take it Bill has not receive files shoot at those ISO.

  • Bill Slattery Jr

    Bill, does this mean that with the D850 one can run lossless compressed 12-bit and lose next to nothing, even in the shadows, when ISO is at roughly 600+?

    • akkual

      More bits, the more detail is stored on the memory card. The lowest level of detail is noise, so when you increase ISO, you increase amplification, and you increase the noise level. So as said by others, when the noise level drops below 12bits, use 14bits and vise versa.

      However, some signal processing algorithms are able to use information stored to the noise (there is always something there, which in invisible to human eye). Mostly these are related to noise reduction, noise shaping, and sharpening.

      So to be safe, for best possible outcome, store full 14bits always. However, if you only export 8bit JPEGs, I would say you can safely shoot 12bits all the time without actually loosing anything meaningful. For example, I personally drop my workflow to 8bit when I do any serious photoshopping (compositions etc).

      • Bill Slattery Jr

        Are you saying it is best to always shoot 14-bit for possible keepers because in the future there may be some software that is able to extract lost useful info from what is now just seen as noise when shooting 12-bit? Interesting.

        • akkual

          Not really. It is more like theoretical possibility. Adobe doesn’t reveal its noise and sharpening algorithms, so pretty hard to say will they ever try to take benefit from the noise or not. And there is obviously definite noise floor, which under you will not find anything meaningful information.

          But there is more to this noise issue than the definite noise floor. Someone said there that 2bits of noise and 12bits of info and after that you are good with 12bit. But that is not exactly how it works.

          Sensors measure voltage of pixels, e.g. from 0…100mV. That voltage is then divided to steps with either 12bits or 14bits. 12bits is 4k steps and 14bits is 16k steps. 0V is blackest black and 100mV is whitest white (in practice, it is never 0V due to thermal noise etc physical limitations). The noise on camera sensors is more prominent when less photons have hitted the photosensor (one pixel). That is, there is always more noise on lower voltage levels (blacks). With 12bits you will record less of that noise, but in the meantime it might well be that the well exposed parts of the picture could still benefit from the higher precision of reading of 14bits, as they have gathered more photons and are thus more stable/reliable on their voltage.

          Now, when you use higher ISO, you basically increase amplification. That means that you need less photons to hit those photosensor for it to render as 0…100mV voltage on the measurement side. So when you ramp up the ISO, you also increase the amplification of the noise on blacks, as there is even less photons that are recorded. But once againg, the well exposed section might have enough photons for more reliable rendering and have less amplified noise than the blacks. This is the reason why some people suggest to ETTR especially when using higher ISOs (ETTR = exposure to the right – meaning histogram has most stuff on the right – also meaning, you are might be “overexposing” for what you are thinking of as best possible exposure).

          But bottom line: There is no definite answer to your question. Use 14bits to be safe, but in most cases you are completely fine to use 12bits, especially when you go higher ISO.

      • peter w

        In Geophysics, we recorded raw data to the amount of detail that could be attained. – That would be 12 bit at 400 iso in the world of D800 files. – Then we would proces and store the results of all individual processing steps at amuch a higher resolution – like a 16 bit TIFF – , this because the processing may shift valuable information to values that would fall in between the 12 bit gaps… (sorry for my English, it is like twenty years ago). Then we would store the final result for geological interpretation to the level that shows relevant detail, which would be somewhere near the original depth, – that would be 12 bit again. – Then we would make a copy suitable for printing, or whatever, the parallel for which a 80% JPEG would be fine, to be safe make it 100%.
        Well, what to do now, 12 or 14 bit… probably I’ll choose to be safe, slow and silly.

        • Yes, 14-bits is only required at the low ISO settings and 12-bit is sufficient at higher ones. FWIW, the cut-off for the D800 is ISO 500 but I’m sure you did fine. Too bad we can’t select bit-depth based on ISO. Maybe someday someone will add that to their menu.

  • pauly

    Bill is very good at what he does I would trust his work over DXOmark

  • Stan Chung

    Great job on the post, My next camera unless the DF2 comes out with sexier specs.

  • Graham Blaikie

    Am I missing something here? Nikon claim that with the BSI CMOS sensor it “captures light more efficiently, resulting in a wider dynamic range” but I don’t see that in the chart. It seems to be level-pegging with the D810. From the hype I was expecting it to approach D750 levels. https://uploads.disquscdn.com/images/a7cfb4e7fcfe94c150e0cd1afb2430d58be2a1dae785ee541edcd95dda9acfe2.gif

    • Chris Phillips

      The D 810 has better DR than the D750.

      • GeoffB

        But Bill’s chart shows the D750 to have greater DR.

        • Chris Phillips

          Whatever the case , the D850 if for argument sake, had even slightly lesser of DR , we would all agree on it being great still considering the high MP count. To be honest I was very pleasantly surprised to see it muching and /or exceeding that of the D810. BTW I go by DXOmark results on DR.

      • akkual

        D750 has higher DR at 1:1 pixel ratio than D810. But when both are scaled down, lets say to 8mpix, the oversampling (more resolution) of D810 increases the DR over D750. I am pretty sure D810 scaled down to 24mpix is pretty close the same as D750.

        • Chris Phillips

          I think they are both great cameras , let us not split hair here. Point is that the D850 has a great DR too and smokes the competition .

    • PhoThor

      Well, i don’t know the site but i could swear the only used raw files as data, so you have to think about this: There is a difference in DR in 1:1 Pixel Raw and downsized for internet, or pictures. So while the DR favours the D750 in 1:1 Pixel magnification, the D810 favours downsized pictures, because more pixels are made to only one pixel in the downsizing stage. It is important to check the values at the same downsized MP, because there you can compare the performance. If you are checking for same resolution the D810 takes DR at Iso 64. Since the D850 will have more than over 20% more pixels, there is probably an increase in DR to overall D750 levels, but you cant see it on Pixelbasis, but will see it on matched resolution.

  • JOHN TANG
  • Eric Calabros

    Compared with A7R2. Very close pixel size, both BSI, both Sony made, and still the old one beats the new one?

    • BSI is only one of many factors. For example, higher read out rates introduce more noise and lower dynamic range.

      • Eric Calabros

        Ok, compare with A99mark2, it has equal or even higher read out rate than D850.

  • BG

    Bill, thanks for doing those tests. I’ve been wondering – are your measurements normalized to an output size (i.e. if camera A and B are measured at the same dynamic range that means that they will produce comparable output at a given print/viewing size), or is this “per pixel”, meaning that a higher-resolving camera with the same score will actually have more available dynamic range at a given print size? Thanks!

    • Yes, Photographic Dynamic Range (PDR) is a normalized measure. If you’re curious to know the details look at the Further Reading section below the chart.

      • BG

        Thanks!

  • David Gottlieb

    Excellent work. Preliminary DR is almost equal with all the other D800 series give or take a tiny fraction. So, the original D800 is still a great camera when it comes t DR. No surprise. It’s the 45mp, the D5 focusing and all the other incredible features that put the D850 above all the rest….. But to be sure, I need to get my hands on one…

  • David Gottlieb

    So anything faster than ISO 320 (between 64 and 320) we should use 14 bit….????

    • At those lower ISO settings the camera would capture meaningful information at 14-bits (rather than 12-bits plus 2-bits of noise); whether you would actually see a difference is another matter.

      • David Gottlieb

        Thanks. Very logical. That’s what I always thought. It doesn’t hurt to shoot at 14 bits. You save space at 12.

    • Thom Hogan

      I call the cutoff slightly different than Bill (ISO 400 on the D810), but yes, you don’t get a benefit from 14-bit past a certain ISO. My book recommendations for years have been to drop back to 12-bit lossless if you’re shooting at higher ISO values.

  • peter w

    Hey, this is interesting. Can please everybody upvote this?

    Do you have any further intermediate-technical explanation for this concerning D800 type camera’s? I see your graph attaining a value of 2 at the iso 320 point, but my mind fails to correlate this to 12 bit recording.
    I assumed with the extreme dynamic range of D8x0 camera’s it would make sense to use 14-bit. It makes sense that 14-bit is only necessary at the iso’s with highest DR.

    (I see further improvement in camera’s: automatic selection of bit-depth recording associated with iso-setting)

    • The y-axis is logarithmic. So when the value reaches 2 the read noise is 4 Digital Numbers (DNs). At that point the last 2-bits of the 14-bit value is random noise and superfluous.

      • peter w

        that is mighty clear 😉
        I noticed the logarithm, so it had to be free of uh, unit. That it would be as simple as that…

        thanks

  • AKEEM

    Now we can finally shoot 12MP images with 4:4:4 chroma lol

  • Fdfas Jlkjl

    TBH this chart is disappointing. The landscape guys should be happy, of course, but I was hoping for better at higher ISOs. It seems like it isn’t going to be as good for me as the D750. I have gone from “definite buy” to “see if there’s a D760”.

    • D750 has 12800 at highest level. This has one stop more with more size. So D850 is obviously D850 is going to be better at high ISOs. That is if nikon is consistent in its ISO interpretation.

      • Fdfas Jlkjl

        No, that isn’t obvious. Just because the setting is available doesn’t mean it will make nice pictures. Have you seen what the highest ISO settings on the D500 and the D5 produce? I haven’t given up hope, I just want to see comparison shots.

        • D500 is Dx so should be treated accordingly. Plus it’s high ISO gives best performance in jpegs. If one does perfect NR then you can get good result in raw. Besides D750 and D650 are comparable being non specialized bodies and nikon has been consistent about their ISO performance ratings and even under speccing that performance unlike canon. Also when scaled down D850 will have more advantage .

          • Fdfas Jlkjl

            In order:
            The highest ISO settings on either the D5 or the D500 give horrible performance in jpegs or raw.
            Perfect NR is nonsense on stilts. What a wonderful world it would be if perfect NR were possible.
            A D650 is a soldering gun and therefore not relevant.
            A particular ISO setting does not in any way relate to a perceptual level of “graininess”, which is what I care about. When I take a picture of a bird in flight, just after dawn, with a high shutter speed, how will the darker parts of the image look? You can’t deduce that from the available ISO settings, but comparing the DR curve to other cameras usually gives a strong indication. The most important ISO settings for me are roughly from the low hundreds to the low thousands. The preliminary results in Bill’s graph show virtually no improvement over the D810 in that area. Let’s hope he eventually revises it upwards.
            A number of photographers who have been testing the camera say the noise performance is very good, so I will keep my fingers crossed.

            • As you very well know , D5 and D500 are low light cameras and hence tuned to perform accordingly. Nothing wrong with that.
              D850 on the other hand is tuned to perform better at low ISOs.. So according to what you use, you surely in luck there.
              And then again, if you resample the image down, that should help .

  • Duncan Dimanche

    Why does the D850 stops at 64iso when the D810 goes beyand with extanded mode ?
    I am a bit confused at that graph…
    what does the white circles represent?

    • Mosawr

      These are preliminary values. A lot of the data points are missing as this isn’t the final chart.

    • Max

      He only had a few raw files to test.
      None of those were shot at iso 64 (nor were there any for certain other iso values).

  • Mr Majestyk

    I find it hard to believe they went to the trouble of using a BSI sensor to achieve no improvement at high ISO in DR. I don’t believe it at all. Look at the A7RII, it has 1 stop more DR than D810 at high (ISO800+) than D810, and I expect the D850 to match that, otherwise they’ve conned us and wasted a lot of effort.

    • TinusVerdino

      BSI is needed to prevent deterioration of high ISO and DR. Still not sure of the effect. In aps-c only Samsung has used bsi-cmos on it’s 28mp sensor (smaller pixels than the D850). It’s DR wasn’t better than the Sony 24mp sensor in the D7200.

      • Thom Hogan

        Again, BSI on pixels that large with already good fill factor just isn’t going to make for a meaningful increase. BSI works best on small photosites, which is why it became the standard for smartphone image sensors first.

        BSI allows you to put more electronics on the sensor because you have more space for it now that it’s behind the image collection area rather than in it.

    • Thom Hogan

      Nikon says they used a BSI sensor for bandwidth, not for DR. Given the high fill factor on the D810, I didn’t expect any real gain from BSI.

      The real question here is how the amount of data that’s being moved impacts read noise. It appears that Nikon has put their attention to keeping that change to zero while making substantive changes in volume of data moving off the sensor, and that’s a significant achievement. It’s the reason why Sony is using BSI and Stacked sensor technology, too.

      • Eric Calabros

        Please compare with A99m2 which is also a high speed sensor, not only there is nearly 1 stop difference but even in shadow improvement chart it has a considerable advantage. If its scaled up version of D500, it should be better. D500 noise at high iso is much better and finer than A99m2 in DX crop.

        • Thom Hogan

          I’m only going by what Nikon is saying, and what they’re saying is indeed believable (that they went BSI for bandwidth).

          That said, it’s premature for me to speak to actual results as I don’t have a camera to actually test. Sony is doing at least dual gain on their 42mp sensors. Looking at my data from the A7rII, it’s even more complicated than that, as there are other things going on at various ISO values. It’s impossible to tell from the early PtoP data whether the D850 is invariant or dual gain.

          But, you’d expect some tradeoff in DR for frame rate.

          • Mr Majestyk

            Never heard of BSI being used for faster sensor readout, even if that is a consequence. BSI is about improving QE and will improve noise levels and DR. Sony A7RII improved DR by a stop at ISO 800+ going from the non-BSI 36MP sensor to the 42MP BSI sensor compared to D810 and K1. If they are using copper traces that will help speed, but we should see real increase in DR too. Sounds to me like they should asked Sony to do the BSI design for them. D810 is not that good at high ISO, so it’s disappointing if DR has not improved even is noise is a tad better.

            I’m in no rush to grab the D850, so I might hang off and see what Sony brings to the table next year. IMO A7RIII will be more like a mirrorless A99II, but stacked sensor, A9 AF, battery improved ergonomics etc and we may finally have a supertele soon.

            • Thom Hogan

              BSI was a way of making small photosite sensors have larger fill factors. Large, full frame sensors already have very large fill factors, so BSI really doesn’t provide much in the way of light capture improvement over what we’ve had in FX.

              A side effect of BSI is that you’ve put all the electronics behind the light collection area (rather than in it) and you now have more room and can more easily address individual photosites with electronic changes. This is the reason Nikon has stated they went to BSI for this sensor.

              I’ll repeat: I don’t expect to see any difference in DR. Even at the 1″ sensor size we only saw a one-third stop gain when Sony made the switch, and the gain in full frame would be scaled far less.

              As I’ve written many times now, I think people are expecting too much with the A7III models. Sony really now needs to take costs out of them, not push them higher in capability. The fact that they named the most recent camera A9 and not A7III should be an indicator: there will be two lines, lower and higher.

          • There is complete Read Noise in DN data at PhotonsToPhotos and it’s definitely dual conversion gain. Very much like the D500.
            https://uploads.disquscdn.com/images/f5456f486978052a1bee925b4b9b7204e9db1018167440307809091bf343edb6.png

  • Max

    Will you be doing more testing later?

    • Yes absolutely. I look forward to getting final results from a properly collected set of files.

  • Nikkor300f4VR

    Well done, Nikon!

  • TinusVerdino

    Looks about equivalent to the D810.

  • John Chandler

    personally, it does not encrouage me to spend $3300 to upgrade my from 810, to little of upgrade, i really dont see a big advantage when, too little too late – i would rather spend it on a7rm3.

  • Chu Eseka

    wrong comment pls..

  • harvey

    I’m thinking you don’t buy the D850 for a big DR improvement over the D800s/D810 but for handling improvements. Even the resolution increase is not that much. But. The improved AF, the button choices, the flippy back screen, increased frame rate, 4k video, improved AF, hopefully a better self-adjusting AF …

    • akkual

      I am one of those, who will be consider D850 seriously for exactly these reasons, as I have been a bit dissapointed with D750 in some aspects on the usability. If I had D850 body with D750 sensor, I’d be even happy with that (obv. that should be less expensive).

  • akkual

    Depends how you define impressive. To me, looks like ISO12800 and ISO25600 are a tad worse when scaled down to 24mpix than on D750. Impressive for such high megapixel count, but not improvement over D810. Of course, need to play with RAWs and postprocessing etc. for final judgement, but I am afraid that D750 will still continue to be the best Nikon FX camera for low light. However, ISO6400 looks nice and clean, when taken into 50% (~23mpix), which would be improvement over D810 and D750. So seems like ISO6400 will be the sweetspot (as with oh so many cameras.. you cannot crush the physical limits I guess). Also ISO1600 seems to be clean at 1:1, which is major improvement over D810, which was 1:1 clean up to ISO800. Of course these are JPEGs, so NR is question. Cannot judge DR from this well exposed pics.

  • Shutterbug

    Question for Thom – does the electronic shutter mean we will get shutter sync speeds like in the D70S?

    • MB

      All cameras that can do video have some kind of electronic shutter and it does not necessarily mean faster sync speed …
      D70s (and venerable D40 🙂 ) had CCD sensor that was able to do that trick …

  • Exposure does not influence the measurement of dynamic range.
    Those PhotographyBlog NEFs you cite are the ones I have already analysed.
    I generally compute PDR for DX Crop Mode but the smaller Raw files are a different story.

  • MB

    I think current CMOS technology has reached the limit as far as DR can go couple of years ago and there are not much difference between D800 and D850:
    https://uploads.disquscdn.com/images/4ec8e93fa18445ed9dc6093aa8a7e5da7363e00dd4fb50261b4181f6b73698ac.jpg
    But real and important improvements are made in signal gain at high ISO:

    https://uploads.disquscdn.com/images/a48210b7005bfd6e1f157cc2e09bbcc57cbf1c3ca1c311f3c32143bf03cb1731.jpg

  • Bill Ferris

    That Nikon appears to have at least maintained dynamic range (versus D810) while increasing both resolution and burst rate, is impressive. The increased resolution, while subtle, should allow a bit more aggressive use of noise reduction in post while still preserving detail. It will be interesting to see comparisons of the D850 with the D810 and other hi-res full frame sensors. As a D610 shooter, I’ll be interested to see direct performance comparisons with the D750.

    (I’m assuming it’s more likely we’ll see comparisons with the D750 than the D610, especially if Nikon merges those two 24MP systems into a single entry level/enthusiast full frame body.)

  • Purdyd

    Just to be clear, Bill Cliff is measuring the read noise from totally dark sensors. The noise is calculated as the standard deviation of the values from those sensors.

    Since Nikon has historically provided some optically blacked out sensors on the side of the sensor, these values can be read from any image.

    Digital to analog converters introduce some noise into the data chain. The faster they are run, the more noise they introduce.

    They also use an digital to analog converter to create a comparison voltage.

    So the higher numbers of bits us, the slower the process. Remember the D300 which would shoot only 2 fps in 14 bit mode?

    Increasing the d2a conversion speed and maintaining the same dynamic range (read noise) is impressive.

    What is not clear to me why Nikon doesn’t offer a much slower frame rate with a 15 bit raw conversion with another stop of dynamic range

    Perhaps there are issues with running the a2d that slow

    • I don’t use read noise to measure Photographic Dynamic Range (PDR) although this is something I measure when performing full sensor tests. (DxOMark does use read noise; that’s a different topic.)
      It is true that several years ago I successfully predicted the PDR of the Nikon D800/D800E using read noise values taken from the optical black area. Nikon no longer writes out the optical black pixels are part of the raw data.

      • Purdyd

        Thanks for correcting me Bill.

  • Back to top