High-Res Camera Sensors: Worth It?
It’s assumed that when you double your camera’s megapixels that you get all of that new resolution, right? Not quite. Usually, not even close.
Hasselblad X1D-50C: 50 megapixels
I did a little test using a Nikon D610 (24 MP) and a Nikon D850 (45.7 MP). I didn't have any Hasselblads handy. The pixel count on the tested cameras is thus: D610 = 4016 X 6068 pixels; the D850 = 5520 X 8280 pixels. The linear change is 5520 / 4016 = 1.37 (37% increase in “linear” resolution).
You’d typically expect that whatever lens you use, it would now get about 37% more resolution (as opposed to expecting nearly double the resolution going from 24 to about 46 MP).You’d typically be dead wrong.
My testing has shown that the limiting factor in resolution is more the lens than the camera. This might not be a big deal if you’re buying a typical DSLR or a mirrorless camera, but I think it’s a huge deal if you’re shelling out about $17,000 for a medium format camera to get those extra pixels.
I understand that there are other factors, such as “color bit depth”, but in actual fact the color bit depth isn’t that much different in going from FX-sized DSLR technology to medium format. Similarly, the dynamic range being captured isn’t very different, either.
There are a couple of web sites that evaluate camera sensors, and they bear out what I’m talking about. At DXO, for instance, I saw the following: Hassleblad X1D-50C is 26.2 bits color bit depth versus D850 26.4 bits. Hasselblad X1D-50C dynamic range is 14.8 EV versus D850 14.8 EV. Hasselblad resolution: 50MP versus the Nikon 45.7 MP.
Now, what’s the price difference? About $17,000 versus $3,000. Wow. I’d be slightly concerned if I were Hassleblad these days. By the way, the autofocus on the D850 smokes the Hassleblad. I didn’t test the Hasselblad; I’d rather buy a car.
But I digress. Getting back to resolution gains, I decided to take a look at a lens with a pretty decent reputation: the Nikkor 85mm f/1.4 AF-S “pro” lens. How much do you gain in resolution by switching to a camera with nearly double the megapixels? Let’s take a look.
Nikkor 85mm at f/1.4 on Nikon D610
Peak resolution is about 36 lp/mm with the D610.
Nikkor 85mm at f/1.4 on Nikon D850
Ouch. You can barely tell the difference between the D610 results and the D850 results. What in the heck happened? The lens itself is kind of “treading water” at f/1.4, and more camera sensor resolution doesn’t get you anything extra.
Next, let’s try stopping down that lens, to see if that helps the situation:
Nikkor 85mm at f/2.8 on Nikon D610
Nikon D610 gets about 47 lp/mm at f/2.8.
Nikkor 85mm at f/2.8 on Nikon D850
Within experimental error, the D850 resolution is no better than the D610 resolution in the f/2.8 shots. The overall resolution gets better when you stop down, as expected, but the lens resolution is still maxed out on the D610; the D850 can’t improve it.
Sigma 70-200 at 70mm f/2.8 on Nikon D610
Nikon D610 MTF50 results using the Sigma 70-200 at 70mm and f/2.8 is a better example for resolution comparison. The resolution range is from about 20 lp/mm to 51 lp/mm.
Sigma 70-200 at 70mm f/2.8 on Nikon D850
Shifting over to the Nikon D850 shows a resolution range on the Sigma 70-200 at the same 70mm and f/2.8 from about 20 through 62 lp/mm. That’s roughly a 22% resolution gain (or 62/51 = 1.22) by using the higher resolution sensor.
We’re still not up to a 37% resolution gain, but I think we’re once again up against a lens resolution limit. Stopping the lens down further, let’s see what we get.
Sigma 70-200 at 70mm f/4.0 on Nikon D610
Sigma 70-200 at 70mm f/4.0 on Nikon D850
The D610 center resolution is about 60 lp/mm. The D850 center resolution is around 71 lp/mm. That’s 71/60 or roughly a 16% increase over the D610.
You can tell by looking at the two-dimensional resolution results that providing “the lens resolution number” is pretty much a fool’s errand. Resolution is all over the map in different parts of the sensor, and sagittal versus meridional directions are hugely different as well. That’s why I use words like “roughly” and “about”. That’s also why I always show these somewhat messy two-dimensional plots.
This testing shows why lens manufacturers have their work cut out for them. New camera sensors are now hungry for better lenses. It also shows that you’re wasting your time and money if you think that a new camera is going to make that old lens really excel.
The only conclusion that should be drawn from this testing is that the combination of a good lens on a high-resolution sensor will net better resolution than a good lens on a lower-resolution sensor. How much better depends upon many factors; generally speaking, the improvement will be a bit underwhelming.
The actual mathematics behind this phenomenon goes like this:
System_MTF = Camera_MTF x Lens_MTF
The “MTF”, or modulation transfer function, is a measure of resolution-versus-contrast that ranges from 0 to 1.0, where 1.0 would be perfect. This math shows that even a great sensor combined with a poor lens won’t give great results, because the lens drags down the “system”. The same is true for a great lens on a poor sensor. The weakest link in a chain spoils the whole chain.
I’m not even considering things like diffraction (by stopping down the lens aperture too far) or poor photographic technique. There’s a whole laundry list of ways to ruin your picture resolution.
It’s a good thing that newer cameras offer more features like faster focus, bigger shot buffers, more frames per second, reduced sensor noise, and the ability to basically see in the dark. Increased sensor resolution isn’t going to win them many more fans, unless photographers enter the very expensive avenue of buying new, higher-resolution lenses.