First of all, thanks for your long post, much appreciated.
At f/4 the lens could resolve about 173lp/mm, in which case the maximum useful sensor resolution at APS-C would be about 40mp, or 7716x5155 pixels.
How you get to that approximation will be obvious to me when I read through that link you posted, I suppose? Looking forward to some time to do that.
Yes...in the Luminous Landscape article, Table 1 lists the diffraction-limited (maximum physically possible) resolutions at all primary aperture stops for any "perfect" lens. Most lenses are not perfect, often by quite a degree, so the resolutions listed in that table are theoretical, but rarely achieved in reality. There are some rather obscure lenses, I think by Zeiss, that have (claimed) to achieve around 400 lp/mm, I've never seen one myself or used one, however that would have to be around an f/1.4 aperture...it would be physically impossible to achieve that resolution at any lower aperture. One would assume, however, that lower apertures were as nearly perfect as could be if f/1.4 reached 400 lp/mm (physical maximum at that aperture would be 494 lp/mm according to Table 1.)
The EF 50mm f/1.4 lens has an MTF chart that indicates the lens has high contrast but acceptable resolution. Even at its ideal aperture, which appears to be f/4, this lens is not going to resolve as much detail (lp/mm) as a perfect lens. Its about 50% resolution at worst, 86% at best, or on average 68%. This indicates that an 18.4mp APS-C sensor would be the highest resolution it could reasonably resolve to before the sensor starts consistently outresolving the lens at all apertures.
Ok, now the underlying math here seems to be interesting. Will get back to it.
From the three charts, the finest resolving sensor is in the 50D, maxed out it could (best case) deliver 106lp/mm. The MTF figure for the lens at f4 on the 50D is 2598. Can I, without any reservations, say that this figure equals 87lp/mm? (87lp/mm in turn equals about 10MP resolution on an APS-C.) If so: We now don't know what caused the gap between the sensors maximum 106lp/mm and the actual 87lp/mm, but couldn't it just be that the lens is already maxed out right there?!
You are indeed correct, resolution at 2598 lines over the 14.9mm height of the 50D sensor would be 87lp/mm. Its plausible that the lens maxes out at that point, however I own the 50/1.4 and a 7D, and I am pretty sure it is capable of resolving more detail than that at f/4. I'm not sure I've ever taken a shot that did not look a little soft on my 7D when pixel-peeping, but I have not spent a lot of time fine-tuning microfocus adjustment for that lens on that body, so I can't say definitively I've seen the most the lens has to offer. Given how good most of the reviews about that lens are, I would assume its capable of well more than 87 lp/mm @ f/4.
Back to your post: 68% of 173lp/mm is 117lp/mm, which is like 18.5MP on APS-C, check. But: Best case you say, the 50/1.4 resolves 86% of a perfect lens. that would be 173lp/mm * 0.86 = 149lp/mm. Where does this come from?
Canon MTF charts are somewhat complex. They plot several sets of information, including contrast and sharpness for both maximum aperture and f/8, from center to corner of the lens. The vertical scale is the degree of perfection of the replicated image..from 0 to 1.0...or effectively 0% to 100%. If, at the center of the lens, wide-open sharpness is about 0.5, that would mean the lens reproduces the original image with about half the sharpness it could if it was perfect. At least, thats my understanding of Canon MTF charts...and not every manufacturer uses the same exact speficications when generating their own MTF's...so correlating Canon's MTF information with other lens brands can be difficult. Anyway...the 50mm f/1.4 lens' MTF chart can be seen on the Canon page for that lens here: http://www.usa.canon.com/cusa/consumer/products/cameras/ef_lens_lineup/ef_50mm_f_1_4_usm
Thin lines represent sharpness, thick lines represent contrast (or put another way, thin lines represent high frequency replication, thick lines represent low frequency replication...the two are useful for measuring the two key aspects of resolution...sharpness and contrast.) I'm drawing a blank on color and dashed vs. solid...one represents aperture (wide open vs. f/8), the other represents sagittal vs. meridonial, or the angle offset of the lines used to measure resolution (positive vs. negative 45°). So, from a resolution standpoint, wide open the lens can offer at best 50% the maximum sharpness that could be extracted from a scene with a perfect lens, and at f/8 it can offer about 86%. The average of those two is 68%. Its a rough number that I guess I was using to demonstrate the lens at around f/4, or "more frequently used apertures". As I mentioned before...lenses don't offer just one resolution, and you really need to pick the aperture your interested in to compare the lenses resolution at that aperture
to the sensor resolution.
If you used the lens at f/8, theoretically according to Canon's MTF, you should be able to achieve about 149 lp/mm. Canon uses specialized devices and methods to measure lens resolution that are capable of far higher resolution than we can achieve with a Camera sensor. So, if their measurements are true, then the 50mm f/1.4 should outresolve any sensor on the market today...which would mean we could never really verify outselves...with a DSLR...whether the lens is actually capable of that or not. Most tests seem to indicate that the 50/1.4 tops out around f/4-f/5.6, but there could be a variety of other factors involved affecting those results (as seems obvious with the 50D test you included in your post.)