This thread has taken a lot of side roads, and we may be long past the usefulness of it, but I shall continue down the dirt path anyway with these thoughts
It's good in an earlier post you brought up some of the psychological aspects of things. I think this occasionally plays into the "clinical" look description of a lens, but it's not always so simple.
My current thinking (which means it's subject to evolve as I discover more, test more, or simply realize more) is this:
- People are generally defensive about their purchases. If we're talking about modern lenses that are in the fast prime territory, none are cheap and not every forum member is living with infinite discretionary spending ability. So this works both ways - those who dropped a lot on, say, F mount glass for their D850 are likely to defend it to the hills ("I don't see the Z mount lenses being so much better and I think they are clinical") (which is ironic when some of the lenses they just bought were considered - you guessed it - "clinical" compared to older AF-D or AIS glass), and then of course those who invested a truckload of rupee in the latest Z glass aren't really keen on hearing it's no good.
- But biases run deeper than that. Biases can often limit or constrict learning. Since you brought up audio in your earlier post, specifically vinyl, we'll head there briefly once again. I was "in the biz" when CD first came out, and lord it sounded like bloody hell. A vast majority of the engineering types were "but it's perfect - the math says so" or something to that effect. Or "The problem is in the mastering - CD is so much more accurate, what you're hearing is the mastering" - which has some truth to it. But then people started *questioning* the status quo and dug deeper, and we (over time) find out a lot about what we weren't measuring, and what other aspects of audio were causing some of the harshness - and to be quite clear - it's not just one thing - it was several, from the early brick wall filters, the impact of jitter, insufficient sample rate versus the filters we had in the converters at the time, some problems with the early Sony based storage medium, to mastering that was optimized for analog and so forth. But the point was that minds were opened and people learned and things improved. Today, I'm a long retired ex-audiophile - I have no dog in this hunt at all, and honestly could care less - a good recording is a good recording, whether it be put down on 1/2" tape in 1965 or 24/192 digital from yesterday. I could care less - I enjoy some vinyl, I have no problem with well done digital. Good = Good. So I want to get into this concept of open mind/questioning/learning/exploration in the next section of this post to tie things together...
So now let's move back to lenses. A couple of immutable rules that aren't subject to debate:
- It is physically impossible to design a perfect lens.
- Thus all lenses involve tradeoffs, of course of varying degrees of magnitude
So if we say a modern lens is superior objectively so, it actually gets quite interesting, because we have to define what defines "objectively so", and that's where the "art" of lens design being a marriage of science and art come into play. Forgive me as we're about to get deep into the weeds here.
MTF50 measurements are not sufficient to fully categorize a lenses performance envelope; far from it. Really the best they show is a very rough proxy for a generic term of "sharpness".
Optical bench style MTF in the form we see from a manufacturer - a few resolution frequencies plotted against contrast across the frame, is a lot better, but still is *inadequate* for proving a lenses superiority in an absolute, 100% sense.
Optical bench style MTF we would see from the design software - resolution frequency plotted against contrast (but each graph would be at varying field positions) is far more useful, but is rarely found.
And here's where it gets really deep. A manufacturer can decide to design a lens that is as perfect as they can make it from an optical bench MTF perspective. And it may not render that well. Why? The method of measurement - while quite important, doesn't tell us everything. I'm sorry if we're going to get deeper here, but years ago, guys like Jun Hirakawa of Pentax, and Nikon designers including Haruo Sato noted that if you design lens with perfect MTF (and the MTF remember is measuring on a two dimensional plane), you may run into a rendering where the sense of dimension of images in the photograph are flattened slightly. (This is a very, very rough description, I don't feel like writing all night). So these guys - and likely others whom I don't know about - designed for a combination of good MTF performance as well as good rendering performance. As lens design evolved, or rather, as the tools to evaluate lens design evolved, we have things like Nikons "Optia", which allows a designer to simulate the *rendering* of a design - so they can look at what, say, a 1960's Leica or a modern era Zeiss Apo Sonnar, whatever, render like, and this is in addition to their design software that tells them the "objective" MTF data. Nikon is not the only player - I believe Zeiss has this capability, and I'm sure Sigma/Sony/Canon/Tamron do as well. The odds the chinese third party guys have it - not so much, but I can't say for sure. The point here is that it's entirely possible that Nikon *could* design a lens that would "ace" the MTF (optical bench style), where the guys who run designs through Zemax and then talk about the results, would be impressed, but they chose *not* too! Why is that? Because there is more than just MTF on a subject plane.
So this means that even for a very well designed, very high end lens, that it's entirely possible for lens A to render a bit one way, and lens B to render a bit the other. How on earth could you really classify which one is "perfect" - particularly when the very best measurement methodology we have - optical bench MTF - even as much as that gives us - still may not tell the whole story?
Let's take that a bit further. If you're designing a portrait lens, you're likely to trade off, in portrait distance, some ultra high frequency resolution in order to get some OOF transfer qualities and bokeh. This makes sense *for this use case* because someone shooting a portrait, hand held, will never be able to *achieve* the ultra high frequencies that they designer willingly gave up in the portrait distance range, but that same someone would prefer the OOF transfer and bokeh.
But then, how do you quantify bokeh as a grade? How do you quantify OOF transfer?
You can't, so you can't always say, with 100% certainty, one lens is absolutely without question objectively better than the other.
(BTW, OOF transfer can be measured with through-focus MTF, where we look at MTF performance as we from in front of / at / behind the focus point)
So while I am in general in agreement with those who prefer lenses that are well designed, don't impart a lot of "mistakes" onto the sensor so that I have a very nice image to work with in post, I also realize there is always going to be a subjective aspect to performance, and we can't get around that. Thus, it's not possible to ever say lens A is absolutely superior to lens B objectively. You might be able to talk about one lens being generally sharper than another in a general sense, and we do, but for the overall equation, it's tough to say lens A is universally better than lens B from a measurement perspective. It's not like we're a company making precision widgets and we meet .01 picometer accuracy of diameter and our competitor only can do .02
Sorry for the long winded post that likely lost a *lot* of readers....
