Introduction
Since Nikon announced the D800 back in February, many people jumped all over it, claiming it to be a medium format killer, and a lot of non-empirical comparisons (for example the Nikon D800 vs Hasselblad H4D-40 video that's been circulating on Youtube) have been circulating online.In addition to existing comparisons, DXOmark has also tested the D800 extensively, and found that it is currently a very high performer as far as ISO performance, bit depth and dynamic range are concerned.
However, I have not yet seen any direct comparisons between the D800 and medium format digital as far as resolution is concerned. This was a big question on my mind considering the huge price difference between the D800 and even the low end medium format cameras...
When shooting film, the size of the film grains in a stock are the same whether shooting medium format or 35mm; medium format simply has more grains over a greater area. Thus, a larger frame (given similar quality optics and the same film stock) would always yield a higher resolution image.
While this same principle applies in the world of digital, there is now another variable in the number of pixels on the sensor.
Clearly, at some point, the number of pixels in a 35mm digital sensor will exceed the potential resolving power of 35mm lenses. I wondered if this point had been reached at 36 megapixels with the Nikon D800, and whether there is an advantage in resolution to be gained by shooting with a larger sensor with a similar number of pixels.
I was very interested, then, when given the opportunity to directly compare a Nikon D800 and a Sinar Hy6 with a 33 megapixel eMotion 75 back...
Note: I wanted to include the test methodology and analysis for those interested, but for the tl;dr version, just scroll straight to the conclusion.
Test methodology
I photographed a Siemens star chart at a distance of 100x the focal length of the lens being used. This allows a direct calculation of camera resolution in lines per millimeter, which can then be multiplied into the physical size of the sensor to calculate the actual resolution of the sensor.I shot the Nikon with a 50mm f/1.4G and the Sinar with a Rollei 80mm lens, bracketing through every f/stop to find the aperture which achieved maximum sharpness.
I used a 1 second shutter delay with the D800, and shot with the self timer to minimize camera shake. I used autofocus, and shot two shots in a row, as the autofocus for some reason was less accurate on the first frame.
While the Sinar Hy6 has autofocus, I found that it was so awful that I had to use manual focus. I was as accurate as possible with my manual focussing, and moved the focus back and forth and checked at 100% until I achieved the most accurate focus that I could.
I shot in the studio to achieve the maximum sharpness that I could.
I then compared the images at every aperture to find the aperture which yielded maximum resolution, calculated the resolution in lines per millimetre, and then multiplied this by the height and width of the sensor for each camera.
Notes and caveats
There are probably other ways to test camera resolution, but this is a method which I was taught a while ago, and something that was easy for me to repeat in the short amount of time that I had with these cameras.
While I would have loved to shoot a full resolution chart, all I had access to was a simple Siemens Star Chart. This meant that I was only able to test centre sharpness, but as I was primarily interested in seeing what each camera's resolving power is in optimal conditions anyway, and the highest resolution point of most lenses is in the middle of the frame, I did not consider this an issue.
As you will see, using this method, the target appears a different size in each frame. However, the projected image of the target will have been the exact same physical size on the sensor for each camera, and this is what the comparison will be based off.
What this does mean, though, is that there is a bit of maths required to actual come up with a final conclusion and comparison.
Nikon D800 test:
Sinar Hy6 with eMotion 75 test:
Crunching the results
Sinar Hy6I found that the Sinar achieved its highest resolution at f/5.6.
In order to calculate the resolution in lines per millimeter, we need to figure out the physical size of the pixels in each sensor. The physical dimension of the eMotion 75 back is 48mm x 36mm and the pixel resolution is 6668 x 4992. Taking the horizontal axis, by dividing 48 by 6668, we find that each pixel is 0.007198mm (which is also confirmed in the datasheet for this back). On the vertical axis, 36mm divided by 4992 is 0.007211.
Therefore, the size of each pixel is 0.007198mm x 0.007211mm, which we can round to 0.00722mm.
Knowing this, the images can then be analysed in Photoshop to find the circumference of what I'll call the "mush circle" - the circle of unresolved detail when zoomed to 100+%
This "mush circle" was found to be 18 pixels in diameter.
knowing that each pixel is 0.00722mm, we can find the diameter of the "mush circle" in mm by multiplying 0.0072mm by 18, which yields a diameter of 0.1296mm.
If we multiply the diameter by pi, we can find the circumference of the "mush circle" - 0.1296 x 3.141592654 = 0.40715mm
Knowing that this is a 30 line star chart, we can divide 30 by 0.40715 to find the resolution in lines per millimeter to be 73.682 lines per millimeter.
This metric does not tell the whole story, however, as we still need to take into account the size of the sensor. Knowing how many lines per millimeter the sensor is able to resolve, we can multiply the height and the width respectively by the resolution in lines per millimeter, to find the maximum number of lines which the sensor is able to resolve.
The sensor is 48mm wide by 36mm high, so the sensor would be able to resolve 3536.73 lines on the horizontal axis, and 2652.522 lines on the vertical axis.
Since each camera has a different aspect ration, we can multiply the horizontal resolution by the vertical resolution to find the total resolution of the camera in an arbitrary unit which we can call "lines per square millimeter" - 3536.73 x 2652.522 = 9378601.61 "lines per square millimeter"
---
Perhaps a more reasonable metric for comparison is maximum print size at an arbitrary, standard viewing distance at which the eye could resolve 5 lines per millimeter. If we divide the lines per millimeter by 5, we can find the amount of enlargement which we could make to a print, and from there find the print size in centimetres; a much more approachable metric.
73.682 / 5 = 14.7364, so the image could be enlarged for print by 14.7364 times. This means that the maximum print size for an arbitrarily standard viewing distance would be (48mm x 14.7364) x (36mm x 14.7364) or 707.3472mm x 530.5104mm or 70.73cm x 53.51cm
This information could also be used to find the maximum print size, assuming a standard viewing distance, for example...
Nikon D800
The process is repeated for the Nikon D800 file.
The Nikon D800 sensor is 35.9 x 24mm, with a pixel resolution of 7360 x 4912. Highest resolution was achieved at f/8.
35.9 / 7360 = 0.0048777mm and 22 / 4912 = 0.0048859mm. These are close enough to square pixels that we can round these to 0.004882mm
The diameter of the "mush circle" for the Nikon D800 is 20 pixels.
20 pixels multiplied by 0.00488mm = 0.0976mm diameter
0.0976 multiplied by pi = 0.306619mm circumference of "mush circle"
Again, as this is a 30 line star chart, 30 divided by 0.306619 = 97.8412 lines per millimetre
Factoring in sensor size, 97.8412 x 35.9 = 3512.502 lines horizontal resolution and 97.8412 x 24 = 2348.1888 lines vertical resolution.
Since each camera has a different aspect ration, we can multiply the horizontal resolution by the vertical resolution to find the total resolution of the camera in an arbitrary unit which we can call "lines per square millimeter" - 3512.502 x 2348.1888 = 8248017.85 "lines per square millimeter"
---
Again, to come up with a more reasonable metric for comparison, we can calculate maximum print size at 5 lines per millimeter.
97.8412 / 5 = 19.56824, so the image could be enlarged for print by 19.56824 times. This means that the maximum print size for an arbitrarily standard viewing distance would be (35.9mm x 19.56824) x (24mm x 19.56824) or 702.499816mm x 469.63776 or 70.25cm x 46.96cm
Conclusions and Discussion (or "the tl;dr version")
I expect that this is what most people will read...
To summarise the above, the resolution in lines per millimeter of the Sinar with 80mm lens was 73.682 lines per millimeter, and the resolution in lines per millimeter of the Nikon with 50mm f/1.4G lens was 97.8412 lines per millimetre.
When factored into sensor size, the Sinar gained a very, very slight overall lead against the Nikon thanks to its larger sensor (48mm x 36mm), with a horizontal resolution of 3536.73 lines and a vertical resolution of 2652.522 lines compared with the Nikon D800's horizontal resolution of 3512.502 lines and vertical resolution of 2348.1888 lines.
Though both cameras have different aspect ratios, it can be seen that, despite the Sinar's 3 million fewer pixels, it achieves a negligibly higher horizontal resolution, and a very slightly higher vertical resolution.
To put this in the context of maximum printable size at an arbitrarily standard viewing distance at which the eye could resolve 5 lines per millimeter, this is the difference between making a 70.73cm x 53.51cm print with the Sinar and a 70.25cm x 46.96cm print with the Nikon D800.
To put this in perspective, last year I conducted the same test with a Nikon D7000 and found that the maximum print size (again assuming the same standard viewing distance) would be 42.79cm x 28.29cm...
Final Thoughts
My primary motivation in shooting this comparison was to determine if the Nikon D800 can really keep up with lower end medium format digital cameras, as many have claimed and argued. Interestingly, it seems that, yes, the Nikon D800 can in fact hold its own when compared with a 33 megapixel medium format back.
It also seems that, with the 36 megapixel D800, 35mm cameras may in fact have reached a point where lens resolution is now the limiting factor.
The fact that the Sinar, with its 3 fewer megapixels, was able to achieve very slightly higher resolution than the D800 is indicative that packing more megapixels into a 35mm sensor beyond this point may be fruitless with diffraction limited lenses, especially considering that the 50mm f/1.4 G is one of Nikon's sharpest lenses between f/4 and f/8.
It will be interesting to see whether there is any advantage gained by packing even more megapixels into a 35mm sensor, as Canon is rumoured to be doing... My suspicion is that 36 megapixels may be approaching the threshold of potential resolution for 35mm sensors with current lens technology and traditional glass optics (metamaterials, anyone...?)...
It would have been interesting to also include a D800E in this comparison, as everything that I have read so far has indicated that it is able to achieve very slightly higher resolution than the D800; without an AA filter, perhaps 35mm resolution might be able to be pushed slightly further still...
Ultimately, I would find it very difficult to justify purchasing a medium format camera with 40 megapixels or less at this point. Admittedly, to achieve these results with the D800, a "medium format" approach must be taken; perfect focus must be achieved, and studio lighting also gives an advantage.
Still, this ultimately confirms what I had previously been uncertain about; when used correctly, this camera is in fact a direct competitor to the low end medium format digital backs. I imagine that Hasselblad would not be very happy with Nikon at this point...
On the flip side, however, it is clear that there absolutely is still an advantage to be gained by shooting medium format digital when the number of pixels is greater than about 40 megapixels, and this is not going to change as long as lens diffraction is a factor within imaging systems.
If you would like to download the original raw files to compare for yourself, they are available here.
If you have questions or comments about the way I conducted this comparison, please leave a (sensible) comment, and I will respond when I can.
Win a Nikon D800 or Canon 5D Mark III
ReplyDeleteEnter in the Draw here:
http://bit.ly/17cfRxE
Do it fast before the Deadline Hits