Web Analytics RiceHigh's Pentax Blog: When the Sensor Outresolves the Lens.. (How Meaningful is a 24MP APS-C Sensor?)

Thursday, August 25, 2011

When the Sensor Outresolves the Lens.. (How Meaningful is a 24MP APS-C Sensor?)

As we all have known, Sony is making a new 24MP APS-C sensor and now markets their new NEX-7 and A77 with this new sensor.

I have been wondering since the first rumour was leaked, do the lenses and optics keep up with the (much) increased high resolution of the sensor? I have got no answer about this not until the MTF charts and figures of the new E-mount Zeiss lens are out. Look at these:-


(Source: http://sonyalphanex.blogspot.com/2011/08/sony-nex-carl-zeiss-24mm-sonnar-e-mount.html)

So, at a moderate MTF figure of 50%, this US$1,000 luxury lens in APS-C resolves (only) 40 lp/mm (line pairs per millimetre), for most of the image borders (7mm and outwards). Translating into the 24mm length for the horizontal resolution of the APS-C sensor, it means that the horizontal resolution of the lens is 960 line pairs in total, at about an averaged MTF of 50%.

We all know that a 24MP 3:2 sensor is roughly about 6,000 x 4,000 in pixels. But don't forget that it is a Bayer sensor, of which the physical layout and colour filtering looks like this:-

G R G R
B G B G
G R G R


The above is for 12 *physical* pixels. To have two perfectly resolved vertical line pair (i.e., a black line adjacent to a white line) with perfect alignment (such that there is no ambiguity in the received colour information), four pixels in a line which run horizontally are required.

Hence, now divide 6,000 pixels by 4 which yields 1,500 line pairs as resolved by the sensor with a perfect MTF, i.e., 100% (theoretically, putting aside the Dynamic Range and S/N limits of the sensor and image processing). Now, it's time to recall the MTF50 line pair resolution figure of the Zeiss lens. It resolves only at about 960 line pairs in 24mm for an APS-C sensor!

The Conclusion? The sensor out-resolves the lens, and by much! Do note also that the above best optical resolution figure is obtained at f/8 stopped down, the situation could be made even worse with the smaller pixel pitch size of the 24MP APS-C sensor of which the diffraction limit occurs at only about f/5.6.

After all, I just wonder why Sony/Zeiss did not make a lens with higher resolution since the image circle has been shrunk. It is noted that the resolution and MTF figures of this expensive "much desired" Zeiss 24mm APS-C lens is not much better than a typical Pentax prime, e.g., the FA28/2.8, check these Photodo figures to compare (unfortunately, the original MTF charts were deleted).

Ironically, the matching sensor resolution in APS-C for the current optics is simply a 12MP sensor in 4,000 x 3,000. It is because 4,000 / 4 is 1,000 which is exactly what a possibly best prime could deliver in term of the maximum resolution in APS-C! Not even to say the noise performance of such a lower pixel count sensor undoubtedly would be better - exactly the same case for the 12MP K-r Vs 16MP K-5 as long as noise is concerned (and that the *actual* image resolution of the K-5 is not exceeding the K-r by that much).

At the end of the day, sometimes I just wonder: marketing always overrules engineering, that is a sad thing after all.

Last but not least, if you're still not satisfied with my theoretical calculations of the above, anyone can verify if what I've said is true or not by looking at more of those *real-world* tests and samples, here is a first very useful comparison and shootout made by the IR, see:-

http://www.imaging-resource.com/PRODS/AA77/AA77A.HTM

Just scroll down to the IQ section and judge on the sample crops by yourself. I am sure that you can decide!


Read Also:-

When the Diffraction Limit is Reached..

K-m Olive Vs 5D Pixel Peeping