The logic behind D800?
06 March 2012 - 05:10 PM
For those of you how don’t know Nikon released a 36 mega pixel, 35mm sensor camera with a max ISO of 6400.
I thought it would be really interesting to discuss it here.
It really perplexes me that Nikon have made such a camera, as there new top of the range D4 has 16mp. So something tells me that if they could get great image quality out of the D800 then they would have increased the pixel count on the D4. Or lowered the ISO performance of the D800 like with the D3X.
What is the D800 for? With Tonys lecture from the good picture meeting. The only thing I could see the 36mp is useful for to crop into the image.
There can’t be that many lenses that can provide meaning information with that fine sampling rate.
So are there other advantages to sampling a lens at this fine rate other than to gain higher ferquency infromation?
06 March 2012 - 10:54 PM
Since the camera was announced I have been thinking why and have collected my thoughts together on the camera. I hope you find them useful.
The introduction of a Nikon’s D800/800E 36.3 MP full frame SLRs has certainly caused a few eyebrows to be raised as the number of pixels is very similar to that of medium format digital cameras. However the packing density of the pixels is very similar to that of Nikon’s D7000 thus at the pixel level you would expect the noise levels of the D800/800E cameras to be to be similar to the D7000. Though for equal print sizes images from a D800/800E will be of lower magnification and visually you would expect the noise to be less obtrusive. To be able to utilise 36.3 MP to the full will place a premium on the performance of your lenses, as at large magnifications the sensor will do a great job at capturing aberrations that would be lost if the lens was used with a camera with a significantly lower pixel count. Even with the very best lenses when used at small apertures, the limiting factor on the amount of detail that can be recorded is likely to be due to diffraction rather than a lack of pixels. It is interesting to see that Nikon have a D800E available which does not have the usual antialiasing filter fitted above the sensor. Nikon’s own literature describes the D800E as follows:-
Nikon engineers have developed a unique alternative for those seeking the ultimate in definition. The D800E incorporates an optical filter with all the anti-aliasing properties removed in order to facilitate the sharpest images possible.
This is an ideal tool for photographers who can control light, distance and their subjects to the degree where they can mitigate the occurrence of moiré. Aside from the optical filter, all functions and features are the same as on the D800.
Note: The D800E carries an increased possibility that moiré and false color will appear, compared to the D800. IR cut and antireflective coating properties of the optical filter remain the same with both versions.
Nikon (or more likely their advertising and marketing folk) have fallen into the trap of equating the visual attribute of sharpness with the objective measure of definition. It is only with very significant cropping or with very large prints examined closely, will the extra definition afforded by the removal of the anti-aliasing filter be seen in images from the D800E. The probability of seeing moiré will be higher in images from the D800E than the D800, but this is where an astute photographer could make diffraction work to their advantage, if the shot permits, by stopping the lens down such that the effects of diffraction mimic the low pass filtering effect of the conventional anti-aliasing filter. With the D800 the level of low pass filtering is fixed by the design of the anti-aliasing filter, whereas with the D800E the level of low pass filtering can be effectively varied as a function of aperture.
Whether the number of pixels is needed or not is a good question. The simple answer is “maybe”, depending upon the way a photographer uses the camera. I am sure for many it will be for “bragging rights” however, there will be some for whom the need for digital cropping is high (eg. surveillance, nature etc.), and there may be some who need large images that will be viewed at distances that are less than the optimum viewing distance. By the same token there will be some photographers who will prefer cameras with lower pixels counts because of their lower noise and faster operation and would never dream of the D800 preferring the D4.
I hope I haven’t upset anyone or put anyone off buying what is I am sure a very fine camera, and I’d like to stress they are my personal views. Someone coming from a different direction may well have different views.
07 March 2012 - 08:34 AM
The ISO performance is an interesting subject. Nikon have traditionally packed the brochure of the D700, D3S and D4 full of gratuitous full page high ISO samples, It is notable that the highest ISO image in the D800 brochure is 800 and it is a smaller image.
I found this noise comparison last night with the D700,
(Now as it is from Nikon Rumours it is entirely possible that it is a completely fake result fabricated by somebody with way too much time on there hands!) I never understand why camera companies are so guarded of there products and expecting people and companies to pay large amounts for pre-order cameras while providing little information about the images capabilities of the system.
From these images it looks like the noise control on the D800 is significantly better than the D700, is which is interesting.
It is quite often quoted that the most important factor in ISO performance and dynamic range it large pixel size. However the D2H disproves this has a pixel size of 9.4µm2 and manages to have terrible ISO performance (although this is a different type of sensor)
I am wondering other factors contribute to signal noise? I presume it is photon noise rather than read noise.
The images state that there was no noise reduction, as far as my understanding of CMOS systems is that the well depths are different across the sensor and it is necessary for the sensor to take a dark exposure so it can estimate the well depth (and thermal noise?) in order to correct the original image. I am not sure if this would count as noise reduction or if it is essential to create the final image.
Recently there has been a massive push to develop scientific grade CMOS chips, is it possible that Nikon have applied some of this technology to make higher tolerance sensors?
07 March 2012 - 08:42 PM
To my eyes the D700 images have higher colour saturation, and obviously if you have more colour saturation then chroma noise will go up. The downsampling will favour the D800.
To check the impact of downsampling I created two blank files in Photoshop, one 4256 x 2832 to simulate the D700 and one 7360 x 4912 to simulate the 800. I then flood filled them both with a mid-scale grey (128, 128, 128) I then added 10% Gaussian noise to both images. I then downsampled using bi-cubic interpolation the 7360 x 4912 file to 4256 x 2840 to match that of the D700 simulation. I then created a 250px X 250px square marquee and looked at the histograms from each image and got the following:-
Downsampled D800 simulation file sigma RGB - 13
D700 Simulation, sigma RGB - 26
Obviously the sigmas vary according to where you place the marquee but I have quoted you an “eyeball” average.
Re. well depth noise etc. I think the logic holds that for deeper wells/larger pixels you have less noise, assuming all other things are equal. I think the technology in the D2H will be different to that of the D800 so I don’t think we can infer anything directly from pixel size/well depth.
I’d welcome anyone else chipping in (no pun intended) on this thread!
ADDITIONAL THOUGHT - The Nikon Rumors test said the downsampling was done using Photoshop. So I repeated the test but used nearest neighbour for the downsampling method and the sigma was approximately 26.3 very close to the D700 simulation! So how that test was run is very important.