I got into this same argument for saying the dreaded "lens out resolves sensor". This, I believe, makes sense. "Sensor out resolving lens" can be made to make sense, too, but it's just definitions. Note: I am NOT claiming that higher resolution sensors will not improve some lens's image. Just that there is an upper bound, and you can meaningfully discuss when you are close enough to "good".
To be clear, I completely agree with the "use whatever lens whose rendering you like with any camera". But that doesn't mean we can't TALK about resolution.
Here's what I wrote in another forum in response to an appeal to Roger's (excellent) article. In particular, the (and I'm paraphrasing) "anyone who says this out-resolves that is jejune". I was being (for me) quite snarky. No insults intended!:
That's the problem with appeal to authority. Quoting is not the same as understanding. While it's true that "out-resolving" is undefined, that's simply because Roger doesn't define it. He defines resolving for each component as MTF = 0.3. Other definitions are possible (as he admits), but he picks one. Well, we can pick one, too.
MTF is a function of frequency (and distance from the axis, but we'll talk about resolution on axis). Roger picks an MTF of 0.3 as his definition of "resolved". That means "at what frequency (lines/mm, say) does the MTF of the system drop below 0.3. The sensor has a maximum frequency given by its pixel pitch (Nyquist). It has a lower frequency f_sensor, where its MTF = 0.3. So here's a definition. A lens out-resolves a sensor if its MTF value at the frequency f_sensor is greater than 0.3. That means that SOME sensor could cause that lens to resolve at that frequency, just not the sensor in question. We could use the Nyquist frequency of the sensor instead. It's a choice.
Similarly, if the lens MTF at f_sensor is less than 0.3, then no higher density sensor could cause the system to resolve that frequency, and so we can say that the sensor already out-resolves the lens.
In practice, though, the lens out-resolves the sensor if no possible increase in lens quality would improve the detail "much", where "much" is some parameter chosen by the observer. Just because a parameter hasn't been chosen does not make the definition silly. Similarly, if an increase in sensor resolution improves the detail by "close" to the theoretical maximum, we could also say that the lens out-resolve the sensor. Again, "close" needs specification.
These are all possible (perhaps even reasonable) models for determining which part of a system out-resolves another. Someone has to make up new models. They were all new, once.
Or we can skip the thinking part, quote Roger or Jim Kasson, or some other expert, and call it a day.
Matt
