Saturday, December 28, 2013

Sensor Sizes – How much difference in image quality between full frame, APS-C and four thirds?

This comes up now and again on the various forums, most notably on the micro-4/3 forum on DPReview, no doubt because that forum suffers more sensor-envy than any other. This comes from a lack of understanding of how much difference really exists in physical area – the single most important factor where image quality is concerned.

This happens to be a law of physics, and until it is repealed, doubling the area of a sensor will generally improve the quality of the result by one stop, at least where noise is concerned. This, in turn, tends to be associate with a similar improvement in dynamic range, since dynamic range is itself inexplicably tied to noise floor. Of course, I am over-simplifying by quite a lot, but then I think at this level anyway :-)

The point is that sensor size matters. And here are the sensors sizes compared. The actual width of the field of view is full frame, with the lines inside being Nikon APS-C, then Canon APS-C, then 4/3.

sensor_sizes
 
Now that is kind of interesting, as the difference from full frame is rather dramatic but the difference between the APS-C and 4/3 sensors is rather small. In fact, the majority of the sensor size difference is in the width.

Why is this? Well, the traditional “35mm” film camera shot a 2:3 width to height ratio as that is simply how the film SLR was designed. We all got used to it and so we carried it forward into digital for dSLRs. This is a terrific ratio for landscapes, since it gives a sense of width, but it is rather unpleasant for portraits, since a 4:5 ratio is far more comfortable for a head and shoulders portrait, thus rendering the 8x10 as a favourite portrait size.

And so we should really compared 4:5 cropped sensor sizes to get a gist for how they will differ in real life (portraits, weddings, etc.) Is there a lot of difference in noise?

sensor_sizes_8x10_cropped
 
Now that is rather a bit closer in size. The full frame is still much larger, but the three smaller sensors are clustered quite closely together in relative terms.

Perhaps it is time therefore for some hard numbers.

image
 
And there we have it. On a typical portrait, the difference in noise between a cropped Canon APS-C and a 4/3 sensor is going to be about a third of a stop. This is often less than the difference in behavior between meters, thus rendering the noise difference under the control of the photographer. For example, it is possible to set +1/3 stops for all images shot on the 4/3 sensor, rendering the total amount of light in the image identical. Thus, the noise is as well. (Again, slightly over-simplified but you get my drift.)

So this is why my various “battle of the sensor sizes” articles always tend to show a big gulf from full frame to APS-C and then essentially no gulf to 4/3. In fact, 4/3 tends to have superior technology to the Canon and Sony NEX cameras, thus rendering image processed from raw of higher general quality with micro 4/3 over these others. Nikon tends to be pretty even with m4/3.

Now, I am not saying that you can get equally good images from m4/3 in all circumstances. In fact, dSLRs retain some significant advantages – AF speed (diminishing, but still there), OVF (always a huge advantage for certain types of photography), sheer performance of the capture subsystem, ergonomics for many people, and so on. But if you think you will get better images just because the sensor looks bigger, then think again.