Conclusion: most people may not notice the rolling shutter in the background in a portrait-mode horizontal pan SOOC, but there is always rolling shutter if the shutter rolls and the lens moves
If your subject went left to right and you panned it, and then it went right to left, and you panned it again, showing those images in sequence would reveal their difference in horizontal compression.
His choice of subject matter and panning speed is one where horizontal compression/expansion would be least noticeable. With a known face, or what is supposed to be a square or circle, the effect will be more visible. With 800mm and a 16ms rolling shutter FF, I get unequally spaced horizontal repeating lines in landscape mode just from the slight vertical sway of the lens during the roll, without panning. The faster one pans, the more consistent the speed, I think, so unevenly-spaced vertical lines are probably more common with slower pans, which are less consistent due to a lack of inertia and more of a direct control.
I already addressed what would happen panning in the same dimension as the roll: compression, or expansion. So, instead of:
o o o o
o o o o
o o o o
you might get:
o o o o
o o o o horizontally compressed
o o o o
or:
o o o o
o o o o horizontally expanded
o o o o
instead of:
o o o o
o o o o
o o o o
You can isolate diffraction, and look for diffraction as the thing you want to avoid most, and in that context, hiding the diffraction is easily done with larger pixels, which remove all fine detail, both wanted and unwanted and lowers the quality of sampling. What is the point, though? In the end, your final display size will determine how much diffraction could be visible if all capture was analog, spatially, and having more pixelation is not going to make the result be any more detailed, and you always have multiple ways of increasing the size of the circle of confusion at the point of display. The same poor logic occurs where people use bigger pixels to hide fine noise, or camera or subject motion blur, or imprecise AF, or aberration.
Looking quickly through the manual, I see that flash sync works up to 1/2000. I assume a leaf shutter in the lens, then; is that the "mechanical shutter"? That's very nice for flash fill in the bright outdoors (and for dealing with strobing artificial lights). None of my cameras with leaf shutters in their fixed or OEM lenses have a sensor as large as DX.
Had the AA filter situation been reversed, things would be quite different. I actually assumed that was how it was, as many MFRs don't drop AA filters on FF sensors unless they have more than 30MP, and I assumed that only Canon was using AA filters on APS-C sensors with 20 or more MP.
An AA filter, of course, will not allow full luminance contrast to occur between neighbor pixels, so when you have a dark background underneath the brighter hairs, that you see with this fly, the one-pixel-wide black lines are dark grey. That is the problem with comparing pixel densities that are fairly close, especially with AA variability, there will often be subject matter that has scale that is seemingly benefited by one pixel density over the other.
Had Donald used an AA-less 20MP m43, instead of the AA-filtered 6300, then the black lines would have been a bit darker between the hairs, and it would take minimal USM to blacken them if they weren't already showing narrow black crevices, if that's what you want to see.
So, it is frustrating to be an advocate of MUCH HIGHER pixel densities than we currently have, and watch people comparing pixel densities that are only 1.2:1 different in a linear dimension, where there are so many larger wildcards (such as conversion styles, and AA filters and other glass on the sensors, and microlenses) that can obliterate much of the benefit of the 1:1.2 smaller pixel pitch. Change that to a 1:3 ratio, and the higher density will pull way ahead, as any AA filter will have a tiny absolute PSF in microns, on the denser sensor, and it may not even have the filter.
I don't think focus stacking, with all the extra wildcards it presents, is the best thing for demonstrating pixel density ramifications in general. We saw early on that the software was choking on the smaller, AA-filtered pixels, probably because the software was written by someone who liked to input under-sampled images, and the software relied on high neighbor pixel contrast to see what it was doing, or decide what is "in focus". There's no question in my mind, that despite higher density being better on average, AOTBE, for maximum IQ (NOT PQ!), there are certain functional things where higher density is more of a challenge when neighbor-pixel contrast is relied upon, such as the software Don used, which needed more specific control to stack properly with the softer pixels, and we also have very-low-light AF sensitivity being somewhat proportional to pixel size, such as with Canon R bodies, creating a situation where it is easy to get focus and more detail with smaller pixels in good light, but you need to use faster, shorter lenses that put no extra pixels on-subject, to maintain focus ability in very low light.
If it is truly easier to align images that are under-sampled, or easier to determine which frame is most-focused for a given point in the scene, then software could conceivably meet its own demands by taking pixelated versions of the images, determining everything that needs to be determined, and then use that information on the full-res images, but software that doesn't rely on high neighbor pixel contrast, which might be a little slower, might be a better solution, just like AF in low light may be slower with smaller pixels, but there is a potential payoff, if or when you do get accurate focus.
Despite years of debate, you seem to have learned nothing. Diffraction (and aberration, too), effect a "diminishing returns", but not "negative returns". You need a certain pixel density to get the most out of an optical projection, but more density will get a little more, and the only "problem" from excess density is larger files.
I never said that he was wrong about getting better results when he used the FF sensor. I said he was wrong if he claims that pixel density has anything to do with it. He also did not make clear the context of his comparison. It had to be drawn out of him. You can't have a meaningful discussion about sensor vs sensor, "AOTBE", without including ALL of the relevant parameters, and making sure that they are photographically equal, rather than taking someone's word for it. 90% of what I see in forums that is supposed to be an equitable comparison, is not.
What if the FF had 2x the pixel density of the 20MP m43, and 8x as many pixels? Would his claim of the pixels in the m43 being too small make any sense, now that the FF's pixels are even smaller? Would he not get at least slightly better quality, with less possible moiré in the repeating eye patterns, and less jaggy bristles? It is subject size and/or subject distance from the lens, with a fixed pupil size, that are actually the base photographic metrics when the same objective is used.
Only to an automaton who views everything at 100%, and judges by visceral reaction to what they see, regardless of how many pixels there are, which would form the desired composition.
This is what a horizontal white fiber, a little out of focus, actually would look like in an analog capture, with enough magnification, where every registered photon was colored according to its wavelength. The more you pixelate it (simulating larger pixels), the whiter and smoother it gets, but the more it lies about what is really there, as it loses information:
That is reality. It is not damage done by a camera, other than the lack of 100% quantum efficiency; it is what light becomes, when captured. You may not feel the need to express this level of detail, but this is always a better source image to start with than one that is already pixelated to be whiter and smoother, especially in an image with lots of fine detail, because pixelating to a pixel-sharp, smoother-variation version in-camera, and then resampling yet again, can cause extra aliasing, which you wouldn't get going from full pixel res to the target display res.
Did you not start out talking about the effect of pixel size? That is what I remember, so correct me if I'm wrong.
If your claim is about pixels, then you have proven nothing, because you have the confounders of sensor size and pixel count, and the fact that you seem to be using all of it with the FF, by getting closer to the subject, based on the fact that you say that you need more focus-stacked shots with the FF, because of a shallower DOF. With the same insect head size, DOF will be shallower with a fixed aperture if you're closer. None of this has anything to do with pixels, but with what happens when you have a prime with only one available aperture and different sensor sizes, when trying to frame the same way with the different sensor sizes
If pixel size had anything to do with this, then my 3.1MP APS-C should be better than my 32.5MP APS-C, AOTBE, and nothing can be further from the truth. Even if I give the old 3.1MP 2x the exposure time to make up for quantum efficiency improvements over about 15 years, the pixelation of the 3.1MP, the much larger absolute point spread function of its AA filter, and the color aliasing will be far more detrimental to fine detail with the 3.1MP than the fact that any change of diffraction is easier to see, independent of other sources of "confusion", with 32.5MP.
Half a stop or 4 stops of what? I don't think any camera tries to create an ISO setting where there will be less than 2.2 stops of headroom above metered middle grey. So what is the "half"? Half a stop more than 2.2 stops of headroom? "4" with the same definition would be tremendous headroom. You can't do that with any quality in the extra levels stuffed in, in a JPEG, because then whites would be dark grey. You can do it in a raw because where white lies within a raw is open to interpretation when converting to sRGB. If it is done by stuffing a couple of stops into just a few top sRGB values like 251 to 255, then if you try to pull those highlights down, you will have visible posterization, which often happens when you try to increase contrast in bright clouds in JPEGs,
Yet, there are lots of people there who complain about technical discussions.
The greatest opposition to technical discussions I have found was at the late photography.on.the.net. One member said that he upgraded from from his EOS 7D2 to the EOS R, and couldn't use the high ISOs on the 7D2 like he can on the R, and I had done some comparisons of noise and found that these two cameras had about the same visible read noise per unit of sensor. So, I asked him if he was talking about when he had to crop or when he used the whole frames at the same ISO, and he got very defensive, and lots of people started treating me as a trouble-maker, and the member soon disappeared, and I got blamed for it.
Another time, someone said, "My 600/4 is 900/4 on my DX camera" and I said that the reality was that it is a crop from 600/4, or "like" 900/6 on a FF", but the 900 and f/4 should never go together, and a mod deleted my post, and the reason given was that it was only John Sheehy's reality. It was like POTN existed to be a haven for fantasy and magical thinking, protected from empirical reality.
That will depend on the sensor, and how deep the shadows go. If it is a flatly-lit scene, and the light color is close to the native color balance of the sensor, then you may see no visible noise difference. If the light is deep shade or halogen, and there are significant shadow areas, then you may see more noise if the sensor uses different analog gain at the two ISO settings, especially with cameras with banding noise in base-ISO shadows..
That's not the best analogy; that would be like using millimeters vs inches for focal length. A closer one would be "wind chill factor". If you say that it is 20C, but the actual temperature is 25C (when wind makes it "feel" like 20C), then the next person can say, "Wow, if it is 20C, then it feels even colder with this wind". That is very similar to when someone fails to mention FF equivalence when they use millimeters of focal length to describe an angle of view. Focal lengths are NOT angles of view. Unfortunately, actual angles never became the common way to refer to angles of view, and that is partly why equivalence is invoked.
