Huhu, true enough more storage and machine time wasted if you do not rely heavily on crop in your photography … I want believe that the Mpx race and marketing craze will die the same way the Ghz marketing scheme died when it comes to computers.
That said many more innovations already stated above are still valid reason to pick a camera in the sense that it will enable the photographer to shoot in a broader variety of situation and have a more consistent success rate !
My top example being :
Image stabilisation (especially in body for those of us using misc lenses)
More accurate metering (yeah autofocus but lighting as well)
Assist tools (zebra for over exposed zones and edge colouring style focus assist have been a blast for me )
small and rugged form factor, to always have your gear with you !
I hear you loud and clear on that one; I have an X-T10, and these features would certainly be a big plus. When my camera finally breaks, I might get one — and when I can afford it, that is; though the XT-50 will likely be out by the time that happens!
I believe you on that, I print some time but not that much … I tend to believe the general idea that in most cases big prints are often design to being seen from a little bit further then mitigating the need for high dpi/ppi …
but then again you may have a different opinion or have special use for large prints design to be seen up-close (giant where’s waldo ? called “Chaly” in french)
PS : 24mpx is close to 16"x20" @ 300dpi
not so high on today’s standard
And I can understand that and you know what you want ! And anyone wanting the same thing should choose it’s mpx count accordingly.
I was maybe too quick in my response and gave the impression to totally dismiss the idea of any high mpx sensor use case as in fact I just wanted to highlight the 2 widespread phenomenons that are :
mpx marketing
what in my opinion is a misconception that printing large automatically in any case means higher pixel count as in fact viewing distance should affect dpi evaluation.
In anyway I want to dismiss other members opinions
I have printed 1x1.5 meters with 16 MP. Upsample the image, add a little bit of noise, and it looks totally fine. Not even any fancy AI upscaling needed. You don’t notice any egregious softness, and there’s enough detail to satisfy close inspection.
Nowadays, I enjoy my 24 MP cameras to allow for a bit of crop. But beyond that, I really don’t have a need for more pixels.
Do you have a real use case for more pixels? I’d honestly be interested in that!
I know. Often I am too picky with it. And yes - there is beautiful art where extremely sharp image in the eye is not needed. But for some strange reason I look for it in my pictures and when I expect the focus should be there but it is missed (and this happen to be the best picture) well - for now I have to live with it. And continue dreaming.
More pixels can be useful when you can’t ‘crop with your feet’. Birders probably grok that; me, I can’t always put the train in a good composition for a few reasons, standing on a photo line, better positions aren’t safe, etc.
Coincidentally a new video came out yesterday addressing this exact subject (which is also something I’m interested in):
In short, he’s arguing that opening the lens up wide open (e.g. f/1.4 or f/1.8) and trying to shoot moving subjects (especially at longer focal lengths) is almost expecting the impossible from the AF system given the shallow DoF. I am going to take his advice and try shooting not quite so wide open and compensating with a higher ISO and see how it goes (up til now I’ve just shot at the fastest aperture my lens has to let the most light in but now I think I need to balance it better)
My 5D mark ii produces beautiful images with the right conditions (not to dark scene, not to fast subject).
Better AF and maintained dynamic range at higher iso are my main drive for looking for a new camera (still undecided what it will eventually be). But I had really like to have nice 10-bit video to top it off.
As for resolution. One part of me wants the A7S iii for its awesome video and I will probably not target displays with a higher resolution than 4K anyway which is pretty much the native resolution of the A7S iii. But another part of me wants the color of my image to be properly sampled and the Bayer/Xtrans layout actually samples color at half the resolution. That would mean that I need a sensor with twice the linear resolution compared to the target medium to sample the color at full display resolution! A 4K display needs a 8K sensor with that logic. So yeah one of those 48 Mpix cameras might not be that ridiculous after all and part of me wants a A7R III for that reason.
Interesting. In most cases I would not expect a downsize algo to adjust perceived colour much (although I have seen it happen), so what you’re essentially saying is that cameras with higher megapixels should have better colour, simply by being able to capture more of it at finer detail?
Has anyone here who has gone from a small to large MP sensor witnessed this?
Haven’t watched the video yet. Thank you for pointing it out. I’ve had some good shots 1.4 aperture at about 2-3 meters distance. But it becomes a challenge for me - of course no eye AF.
This guy however https://www.youtube.com/@TheSlantedLens
does quite a bit of comparison videos. And what I like is that he specifically tests AF performance. Maybe I have missed some points but from what I have seen the new cameras have this benefit.
Joking aside, I suffer from the same affliction. I use zone focusing quite a bit for street photography, but I also use auto focus when the light isn’t great — and I always end up with a shot where the wrong person’s face is in focus.
I sympathize with the view that for most people, a better camera is not the best path to becoming a better photographer, but at the same time the expectation that you can do everything in software is misleading.
Sure, post-processing allows you do to a lot. But if the subject moves, multiple shot techniques become increasingly difficult and tedious.
Also, a lot of features trickle down from flagship cameras (better AF, weather sealing, and IBIS are current examples) to enthusiast models. So for a lot of enthusiasts, “getting a better camera” means “getting a camera I can finally afford with features that were previously available in much more expensive cameras”.
That said, Thomas Eisl makes a lot of good points, sensor DR is not something I would get a new camera for these days. It has already trickled down to cheaper models to a large extent years ago. In another video, he makes a case for buying a second hand DSLR, which can also be reasonable for people learning photography on a budget. But MILCs below 1000EUR have become so competitive on other fronts that this really depends on the features one wants to have. For me, weather sealing, IBIS, and small body (around m43) are important enough.
The screen has a red, green, and blue led for each pixel, while the sensor has a red, green, or blue photodiode per pixel. The screen thus has a spatial resolution 3 times higher than the Bayer sensor.
The engineer in me had liked to oversample rather than undersample even though it won’t be visible. I promise I will explore this topic further and maybe even contribute a downscaling-based demosaicing algo if I get a high res camera in the future The effect should be visible for any resolution though as it’s a matter of your target resolution. I only have a 2K screen now so a 4K sensor is actually enough in that sense.
Yes I understand interpolation, but interpolation has to happen whether you have many megapixels or few, and so long as the algo is the same, the colour will be the same. (Theoretically, I can see the appeal of a downsizing demosaic algo. Didnt know you had that in mind. Would be interesting to compare results from that vs current method).
There are of course different pros and cons to higher resolution cameras, but I didn’t think a noticeable difference in colour was one of them.
From the back of my head: the reason why ARRI chose 2.8K horzontal resolution for the Alexa was that after debayer and downsample this was the resolution needed to deliver FullHD with 4:4:4 sampling (video world term meaning full color resolution as opposed to color-subsampling where chroma has half or quarter resolution while luma is still FullHD). So their calculations give a linear dimension increase by about \sqrt2 whereas brute force binning 4 bayer-pixels together would be a factor of 2 like @jandren suggests.
Also there are black and white resolution comparisons from Leica monochrome and bayer pattern Leicas (those are debayered first, turned into B&W and then compared to the monochrome pictures of the same object). The monochrome version is significantly outresolving the bayer pattern camera.
Omitting 2 of the 3 channels per pixels leads to a significant loss in details for each channel.
)
