hania wrote in post #16075134
Sorry, may have confused you all - it was with an authorised Canon repair Centre (local ).
It had a repair in March (both the contacts on my 7D and on the 100-400 needed replacing - failed at end of workshop in Poland).
Zoom ring also replaced and all calibrated (so service docket said).
Took it to Irish Sea Bird workshop June - found it was rubbish under f8 and not much better above .(checked by tutor).
Fortunately I also had my 70-200 2.8 with me - so I managed.
Once home I took it back again as still under 6 month warranty : they said the focus module had failed so they replaced it.
Did set of 100% crops at all f-stops and sent them to tutor - who said that the lens may be within Canon';s tolerances, but is very soft!
I have sold it now
Ah, OK. Do you have those crops handy still? I would be curious to see how they compare to my lens' behavior. The 100-400 is certainly not known to be Canon's sharpest lens anyhow, though.
This is drifting a bit off-topic (after reviewing this, maybe I should start a new thread elsewhere, but I'm sure it has been written many times over...), but I was poking about the interwebz yesterday, and came across a variety of articles about lens theoretical resolution capabilities vs sensor resolving power. Now, this is no excuse for the 100-400 to look far worse than the 70-200, but does seem to indicate we are nearing some physical limitations in modern lenses and sensors.
A rule-of-thumb for resolving power is something like:
For a lens:
700/(f-stop) = lines pairs per mm
So any lens, at f/7.1 (f-stop 7.1), resolving power is just under 100 line pairs / mm (lp/mm).
And for a sensor, the line pairs / mm is just half the lineal pixel density (# pixels wide / physical width, divided by 2 for "line pairs").
For a T3i or 7D APS-C sensor, quoting numbers from a canonrumors forum post by jrista (a lot of other info is well presented in that thread at CR, but I did read similar stuff elsewhere, so it isn't my only source!):
115.97 lp/mm (3456 lines/14.9mm sensor height = 231.94 l/mm, divide by two to get lp/mm
Thus, any lens at or beyond f/8 is being outperformed by the sensor of any of the newer ~20MP APS-C cameras.
Wide open, say f/2.8, a theoretical lens should resolve to 250 lp/mm, which is well beyond pixel density of APS-C sensors. However, other limitations to lens design can start to be noticed, which I guess is why good lenses peak in resolving power at about f/4 or so.
Now, the rule of thumb I used, 700/(f-stop), may be making a lot of assumptions. For example, it assumes a contrast ratio MTF of 50% is needed to resolve extremely fine line pairs on a digital sensor, whereas human vision is reportedly about 9%. If human vision is used as a guide, you can use a rule of thumb of about 1500 / f-stop. Which means you need a sensor of about 90MP to start out-resolving a lens at f/7.1.
So maybe improving sensor performance for lower contrast is the next frontier in performance improvement.
I can't recall in my browsing if there is any opportunity to cheat the theoretical MTF limits. If not, then the only other opportunity to improve lens resolving power is improving lens performance at lower f-stops, but sometimes you need/want higher f-stops to get a useful depth of field.
Sigh, darned physics!