At this point in our conversation a reasonable person would ask which provides greater detail with an APS-C body: a professional grade full frame lens that loses detail with the smaller image circle or an APS-C lens with perhaps lower optical resolving power but which the sensor can fully utilize?
When we measure a lens’ optical quality usually it is connected to a bench and recorded without a camera attached. This has meant we could understand a lens’ sharpness in the abstract, but it doesn’t take image circle into the equation and doesn't consider the camera you're actually using to shoot the picture. Manufacturer’s (and other testers) will put out charts on a lens’ sharpness. You can easily find these MTF (or modulation transfer function) charts, but they don’t take our basic question into account. For a long time, I didn’t think anyone was measuring this question.
But a company called DXO Mark does. They have very readable tests of most lenses and they test them on most cameras. Then DXO created their own unit of measurement for interpreting sharpness data in a way that relates to our exact question. It has not been adopted by the rest of the industry yet, but since no one else has another (let alone better) system, I think it’s worth your attention.
It is called the Perceptual Megapixel and it works like this: we shoot the lens on a particular camera at a variety of focal lengths and record the detail of the subsequent images. Then we average these together and describe the result as a particular camera resolution. The idea is that we are comparing the recorded detail against an imaginary lens with zero loss of sensor resolving power.
Since all lenses lose some detail this number will always be lower than the full resolution of the camera, which ideally would be paired with an optically perfect lens. But those just don’t exist.
The upside of this system is a very readable number that tells us what a lens is like to actually use on your particular camera. The downside is the number averages all focal lengths and apertures together, so it tells us nothing about the sharpest part or aperture of a lens. But that isn’t our question at the moment. We want to know what is better: an APS-C lens on an APS-C camera, or an optically superior full frame lens on that same camera? Let’s look at the data.
By the way, to reproduce my results I’m simply searching DXO Mark’s website for camera and lens combinations. Please, examine other lens and camera combinations to build off what I’m writing here and expand your understanding thereof.
I first took a look at Canon cameras. So let’s put some macro lenses on the 7dii, a 20MP APS-C camera. First, let’s put the Tamron 60mm macro on. This is a $500 lens and it’s giving us 9 P-MP (or 9 million perceived pixels). I know that seems like an enormous drop, but now let’s put the $900 100mm L macro on the same camera. We record 11 P-MP. Is $400 worth two million more pixels?
Let’s do a similar test with Nikon gear. The 40mm is a small macro lens that costs under $300. On a 24MP D7100 it produces 15 P-MP. Now let’s put the full frame $900 105mm on the same camera. It’s giving us 12 P-MP. That’s right, we’re recording less information with a lens that costs over three times as much. Put the same lens on a 24MP D750, which is a full frame camera, and we record 19MP. There is always some loss with a lens, but we give up, on average, 45% of a full frame lens when mounting it on an APS-C camera.
That said, as photographers we need to keep other factors in play when looking at lenses. Image circle and optics can’t be everything. What about how large it is? What about it’s brightest aperture? What about your budget? What about the way its focal length renders background? All these things matter. I don’t think people make themselves better photographers (or happier people, for that matter) by merely chasing optical quality. I do believe, though, that lenses are a long term investment, and making the best decision possible involves having a plan for your photographic future. Whatever that may mean to you.