NEW DATA Is the D750 still the low ISO champion???

[Frank Fremerey]

OLD DATA as of 9 OCT: Here is the comparison D750 (green) vs D850 (blue) when it comes to PDR

NEW DATA on 15 OCT suggest this new graph d750 (black), d850 (blue):

[Mongo]

Thanks Frank. Interesting stuff. Can you tell us the site so Mongo can look up other comparoisons.

[Frank Fremerey]

http://www.photonstophotos.net

[Michael Erlewine]

Here is the comparison D750 (green) vs D850 (blue) when it comes to PDR:

Frannk, how do you read that chart. To me it looks like the D850 is farther to the left (ISO 64), while the C750 is more to the right. What an I not seeing?

[chambeshi]

Thanks Frank! I've logged these charts. The real-world datapoints of the D850 at ISO 64 stand out. These dots cannot be finite points wrt each model. We are well acquainted where variance in quality shows up in a lens :-) So, I question :

1. What is the intra-populational variation across sensors for a camera, including between batches? If there's variation , how much do these differences matter between different copies of the same camera?

2. Is it mainly the differences in noise that stand out in what we actually see in images taken at ISOs above 3200? And there are software solutions to reduce noise in post processing.

http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D4,Nikon%20D5,Nikon%20D500,Nikon%20D750,Nikon%20D850

http://www.photonstophotos.net/Charts/RN_ADU.htm#Nikon%20D4_14,Nikon%20D5_14,Nikon%20D500_14,Nikon%20D750_14,Nikon%20D810_14,Nikon%20D850_14

http://www.photonstophotos.net/Charts/PDR_Shadow.htm#Nikon%20D4,Nikon%20D5,Nikon%20D750,Nikon%20D810,Nikon%20D850

I've been shooting my D850 since last Tuesday and am in awe at the results across all ISOs even 12800. Admittedly, I don't have a D750 or D5 for direct comparisons of the same scenes. Above and beyond any pedantic details in its inner workings, I'm utterly delighted with this new Nikon to compliment my D500 and Df. So far, I have been mainly shooting flying birds. I'm keen to see how the AF performs against the excellent D500.
I cannot yet pick up any significant differences between these 2 AF systems (as we know the D5 system). The brilliant viewfinder of the D850 stands out. But the full spread of focus-points across the D500 screen remains a definite advantage in action shooting. One also gets high continuous shooting (10fps) in a D500, without the tax paying for the Grip, MH18a battery (+charger) - pricey prerequisites for the D850 to shoot at 9fps

[Ilkka Nissilä]

I guess these small differences are probably not very significant from the point of view of practical applications (compared to the large differences in feature sets, body size and handling etc.) and they show different emphasis in sensor design.

My favorite metric is DxOMark's color sensitivity measure.

https://www.dxomark.com/Cameras/Compare/Side-by-side/Nikon-D850-versus-Nikon-D750___1177_975

Click measurements / Color sensitivity to get to those graphs.

Ultimately practical use and evaluation of the results in one's application reveals the value of the different tools.

[Frank Fremerey]

Frank, how do you read that chart. To me it looks like the D850 is farther to the left (ISO 64), while the C750 is more to the right. What an I not seeing?

The higher the point, the more PDR is measured by the site owner who creates these charts (other sites deliver different results using different methods).

Here is the interesting part, zoomed in ( http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D750,Nikon%20D850 )

[Frank Fremerey]

My intention posting these is to keep anyone from thinking it is necessary to get the D850 for better Dynamic Range.


There are millions of arguments to get a D850 like AF, 7/9 fps, color accuracy and differentiation, battery life....

[JKoerner007]

Frank, your view is in conflict with everyone else on the planet.

You might want to note the actual language on the Photons to Photos site you reference(in red even): Preliminary data are actual measurements from raw files but not those taken under the usual controlled conditions.

In other words, you're citing unrefined data. Photons to Photos then recommends the reader "see DxOMark Photographic Dynamic Range Chart." I'll post the screen grab for you:



Moving from Photons to Photos to DxO Mark, you might want to read their latest article. To quote, they say:

• "The D850’s key strengths are its outstanding color (Portrait score) and dynamic range (Landscape score) at base ISO where it again ranks as the number one among all commercially available cameras we’ve tested for these attributes. Offering outstanding dynamic range of 14.8 EV and color depth of 26.4 bit at base ISO, combined with its massive 45.7Mp resolution, the D850 is a mouthwatering prospect for landscape, studio, portrait, as well as high-end editorial or advertising photographers who are  seeking top-notch image quality for large-scale reproduction and display."

In DxO's comparision, the D850 beats the D750 pretty much across the board. Here are some screen grabs from DxO Mark showing the differences (SNR and Tonal Range are near identical; however, Dynamic Range and Color Sensitivity favor the D850 throughout):








Granted, the D750 is a very good camera, arguably the better value.

However, when you factor in the AF, the sensor-size, the FPS, combined with the absolute best you can do in color/DR, the D850 earns its #1 spot among all cameras available today ... and the D750 its #11 spot.

[Frank Fremerey]

Frank, your view is in conflict with everyone else on the planet.

You might want to note the actual language on the Photons to Photos site you reference(in red even): Preliminary data are actual measurements from raw files but not those taken under the usual controlled conditions.

In other words, you're citing unrefined data. Photons to Photos then recommends the reader "see DxOMark Photographic Dynamic Range Chart." I'll post the screen grab for you:

Moving from Photons to Photos to DxO Mark, you might want to read their latest article. To quote, they say:

• "The D850’s key strengths are its outstanding color (Portrait score) and dynamic range (Landscape score) at base ISO where it again ranks as the number one among all commercially available cameras we’ve tested for these attributes. Offering outstanding dynamic range of 14.8 EV and color depth of 26.4 bit at base ISO, combined with its massive 45.7Mp resolution, the D850 is a mouthwatering prospect for landscape, studio, portrait, as well as high-end editorial or advertising photographers who are  seeking top-notch image quality for large-scale reproduction and display."

In DxO's comparision, the D850 beats the D750 pretty much across the board. Here are some screen grabs from DxO Mark showing the differences (SNR and Tonal Range are near identical; however, Dynamic Range and Color Sensitivity favor the D850 throughout):
Granted, the D750 is a very good camera, arguably the better value.

However, when you factor in the AF, the sensor-size, the FPS, combined with the absolute best you can do in color/DR, the D850 earns its #1 spot among all cameras available today ... and the D750 its #11 spot.

My view is the view of the Author of Photon to Photos William J. Claff.. He writes:


DxOMark Investigations and DxOMark Derived Data (Ball (magenta)interactive charts Ball (cyan) interactive tables Ball (blue)articles). Data is derived from data available at DxOMark. In general these results are less reliable than those actually measured and presented in the previous Investigations section.


In opposition to what you say, Bill's own investigation seems more trustworthy to him than the DXO-Data!

[Frank Fremerey]

PS: I had my D850 preordered and after one month of heavy shooting I am still very happy with her. you can read my comments on the experience with my D850 all over the place.

[Ilkka Nissilä]

In opposition to what you say, Bill's own investigation seems more trustworthy to him than the DXO-Data!

DXO give engineering DR and SNR for 18% gray. From these to get the "photographing DR" (using threshold SNR = 20) I believe some kind of modeling / interpolation may have to be used, so presenting PDR from DXO data is an approximation. That doesn't invalidate the data DXO presented by DxO themselves (they're just different measures).

I believe Bill Claff's site marks those cameras which are estimated from images that were not shot in a controlled way using a symbol such as (e) or (p). Initially the D850 had one of those letters behind it in the legend, but then those were replaced by new data which is apparently measured using a controlled procedure and that symbol indicating estimated or preliminary data was dropped. However, I don't know how controlled the procedure is; I would think a company such as DXO which do all the measurements and analysis in house can control the process to be more precisely the same for all cameras.  But any measurement procedures over long term can be subject to human error and variability in conditions. I do believe each site make their best effort to provide valuable data.

Also it is good to remember that there are differences which are not measured by these procedures. For example the D800 strongly clips blacks at high ISO, leading to problems in its use for astrophotography (where averaging is commonly used to image faint objects). In the D810 this issue was greatly alleviated. This is discussed in Jim Kasson's blog in depth, it is worth reading. I remember also seeing comparisons that show that the D750 handles long exposures at high ISO really well, but I don't remember which web site it was.  Some other cameras are reported to produce increased noise after being used for longer exposures.

I would not recommend using a single source of sensor data to make camera-buying decisions. Study multiple sources of information, and consider also the feature set, not just numerical data. For most people the feature set is probably more important than small differences in sensor image quality. And for those special applications where the differences in sensors do matter, one would be best off to test the cameras in the application itself to be sure.

[bobfriedman]

Frannk, how do you read that chart. To me it looks like the D850 is farther to the left (ISO 64), while the C750 is more to the right. What an I not seeing?

dynamic range here is shadow to highlight detail measured in bits (powers of 2).. i personally don't put much stock in these curves and question the method of their construction as somewhat obscure and arbitrary.. however, i would read these as essentially the same.. since they are within 1/2 bit of each other if they were accurately calculated... fractions of a bit...

what i would like to see is more bits.. exceeding 14 bits.. like 20 bits.. i am waiting for this.

[Ilkka Nissilä]

To get 20 bits of dynamic range you can estimate the ISO required by extrapolating from the graph; something like ISO 0.1 would get you started But you could just do series of exposures and average them in software.

[bobfriedman]

To get 20 bits of dynamic range you can estimate the ISO required by extrapolating from the graph; something like ISO 0.1 would get you started But you could just do series of exposures and average them in software.

Yeah. there needs to be a technology innovation to adjust the slope. But you get my point.