Me too. I have found that often when doing something new (I remember the first time replacing brake pads on a car about a decade ago…) I assume there is only one right way to do it, and that the reasons are too complex for me to understand.
Now just a little older and sadder and wiser I seem to have realised that there often isn’t a 'silver bullet approach ’ and a cautiously experimental approach can be the basis for learning almost anything.
Course, in another decade I may find a new opinion
P.S. I also sometimes envy people who happily approach a new project with naivity - it can be easy to scorn but again, can actually be a massive help, as long as it’s not self destructive.
The more experiments I’ve done, the more original sources I’ve read, the more I learned that “common knowledge” is way too often false.
I find great joy in experimenting and evaluating. I suppose that’s my education as an engineer and scientist poking through.
There seem to be two most effective ways of learning for me: experimenting, and teaching. Sadly, I don’t teach any more. Thankfully, that leaves me with more time for experimenting.
Well, and watching YouTube. Watching proficient people being productive is just too satisfying.
I kind of missed his point about lenses having a resolving power limit. He says that he disproves this myth, but I’m not entirely sure what the main conclusion is. I think he’s saying that the lens sharpness scales with the megapixel count, so more megapixels will always show more details and therefore make a lens look sharper.
And DxO proves this because their sharpness metric goes up with higher mexapixel camera bodies. Is this what he’s saying?
There is a somewhat widespread sentiment (eg dpreview forums) that a given lens has a given resolution, which can be quantified as a number of megapixels. This is, perhaps inadvertently, reinforced by DXOMark’s “perceptual megapixel” score, which does vary with the body, but a lot of the time they test with a single body so people don’t notice this.
The video says that this is not true, and provides examples. I think that’s all it does. A simple, but important point. (Disclaimer: I like his videos, even when they don’t provide new information to me.)
Yes, I’ve come across this kind of comment a lot as well. But I’m not sure he fully answered the other statement I also read along the lines of “you need lens x to fully resolve that high megapixel sensor”. If I’m not mistaken, he doesn’t address that directly. I think that statement is true, although needs clarification. It just means that you need an ultra-sharp lens to show the increased detail at its best (sharpest), but increased detail will always be increased detail, no matter the lens.
When Fuji came out with the 40MP APS-C sensor, they made several statements about which lenses could fully resolve the 40MP, which made many people think that the other lenses would look worse on that sensor. They eventually clarified to say that those lenses would still look better on the new sensor but only the high-end lenses would show the sensor at its best.
And this is probably only applicable to wide open apertures as well. Stop down any prime lens from the last 50 years I am sure most of them will resolve similarly at f8 or so.
I still think it is the wrong mental model for practical purposes. You can think of a lens as introducing some kind of blur even when perfectly focused, but the problem is this is a complex function of its optics, not something you can map into a pixel count. So while a 100 Mp sensor would probably be overkill for a 1990s kit lens, there is an interim range where you are still recording useful information.
I approach this in the context of modern raw processors like Darktable and Rawtherapee: in the interim range, the more pixels you have, the more capture sharpening has to work with, so you still benefit. And more pixels also help with CA removal etc.
As a simplification, all modern lenses are sharp in the center, stopped down a little. Even more pointedly, all modern lenses are more than sharp enough for almost all displays and print sizes.
I recently compared my “bad” Fuji 16-80 to my “good” 50/2. Objectively, these show significant differences in sharpness when you look closely. However, their centers stopped down are both perfectly sharp even at 100%; and zoomed out to A2-equivalent size, even the corners are indistinguishable.
We are a ridiculous breed. Looking at pixel level details is a complete waste of time.
Yes, I’m sure you’re right. I guess the terminology is a little misleading. When people say things like “fully resolve”, it implies that there is a threshold when something passes from not resolved to resolved.
But while lens optics will have a given sharpness/contrast (at a certain f number), it’s nothing to do with pixels, as you say.
I think what people really mean by “fully resolve” is simply that the extra details of a high MP sensor are adequately sharp at the 100% zoom level, which is probably not that important in practical terms. Would you agree?
To a certain extent, yes, but this is how I think about it.
I think that higher pixel densities always result in more information you will be able to extract from the image, using computational methods. Some may say that they are not using computational methods for this purpose, but that is a misconception unless they have a Foveon or a monochrome sensor — demosaicing is a computational method.
It’s just that diminishing returns kick in: at some point it becomes tricky to distinguish the smaller sensor resolution image from the larger sensor resolution one after the latter has been resampled to the same size, assuming that both went through the same kind of computational pipeline. There is always a benefit to larger resolution, it’s just that at some point it stops mattering for practical purposes, and there are trade-offs (sensor readout speed, file size, etc).
Lens and sensor resolution are multiplicative in the frequency domain. Thus improving one always improves the other.
But there are definite limits still. The color filter array limits resolution to 2/3 of Nyquist. Moiré folds above-Nyquist detail into below-Nyquist beating and noise. We fool ourselves if we expect every pixel to contain unique new information. It’s either filtered by the antialiasing filter, or obscured by Moiré noise.
And that’s before even considering the limited resolution and shot noise of visible light.
But as has been said before: the demands we put on our gear are ridiculous. There exists no print or screen that can reasonably (!!!) present more than 12ish megapixels to a human observer (of the entire photo). So all these considerations really don’t matter much, as anything pixel level is just not visible anyway.
On another forum someone showed a stunning 100 MP portrait of a young lady. You could inspect the depth and width of each individual pore of her skin. It was astounding. Truly, I stood in awe of the achievement. But I have no idea what purpose it might serve. A building-scale billboard that still holds up to close scrutiny? A dermatology study? The opportunities are endless!
I have seen similar images of cityscapes where you can zoom in and see what’s happening in all the windows of the skyscrapers. It was fascinating, even if a little voyeuristic. I think it was the YouTuber Omar Gonzalez that recently said that pixel-peeping is silly and not necessary. But we still like to do it for our own enjoyment. And I can see what he means. There is something very satisfying about looking at a scene and then zooming in to see the small details, even if it isn’t a good way to judge gear. It still provides a geeky pleasure.
Beyond that, there are some practical purposes for wanting higher megapixels and sharp lenses.
Here’s an extreme case:
With this image, my lens couldn’t reach the bird adequately. Cropping in to the framing I would want, I go from 26MP (right-hand image) to about 1MP (resulting image is 868x1070 pixels).
The resulting image is not good and not display-worthy other than perhaps on a smartphone. To achieve this kind of crop at a resolution of 6MP, I would need a 150MP medium-format body.
Now, obviously this is a silly scenario, and you shouldn’t be cropping in that much. But it shows how high a resolution you actually need for large crops. So if you are the type of photographer than wants the extra flexibility of cropping, then a higher MP body and sharp lenses make sense (as long as it isn’t just a crutch for bad technique).
Also makes sense for photographing art for archival purposes. I think those are mostly done with GFX or bigger nowadays. It makes sense to go for higher resolutions both for later reproduction and also for art historians, artists, etc, to be able to inspect details and techniques which they would otherwise do anyway using the eye up close or loupes.
Yes, several years back I went through the laborious process of scanning hundreds of negatives with my DSLR to digitize them. It actually took several years to complete the project, but it was well worth it.
However, even though I used a good macro lens, my camera at the time was only 12MP and had average IQ by today’s standards. At some point I may scan a roll with my 40MP camera or even use the 80MP high-res mode of my OM-5 to see if it’s worth redoing some of my favourite photos, but I doubt it will be, unless I want to print out any of them quite large. The original photos were still restricted by the technology of the day, and I never owned high-quality cameras or lenses back then.
I like this guy: https://www.youtube.com/channel/UCvYK4mDvRB_hMGSYaxgQGCg
Most youtubers in general are really annoying to me which is why I don’t use the site much anymore, but this guy seems to just be happy to talk about art and try making it on camera, very nice stuff.
I’ve zero taste or knowledge, and seem immune to acquire either. After dropping out of the local technical university’s program that I seemingly only ever had an eye on but nothing else, one of the few things I remember to this day is that sometimes it’s necessary to have exposure to “Stallgeruch” (stable odor), as some things can only be learned by osmosis during extended exposure.
The problem is, the difference between stable odor and bullshit barely exists.
So, is this YouTube channel providing suitable exposure to stable odor, or just full of bullshit? (My opinion: I like most videos, though I don’t think I learn anything. I watch them during my final beer of the night, usually.)
I seem to have realised that there often isn’t a 'silver bullet approach ’ and a cautiously experimental approach can be the basis for learning almost anything.
