gdanmitchell
Offline
• • • • • • •
Upload & Sell: Off
|
 p.5 #18 · I returned the X-T5 and kept the X-T4 | |
The above-mentioned video has a major problem. I'm surprised that anyone doing such a comparison would not know this.
I'm sorry, but this is not going to be short. But it is important, and it addresses a very common fallacy about using 100% crops to compare camera "sharpness." Anyone using 100% crops to compare sharpness needs to understand this.
- - -
MAIN POINT UP FRONT: Evaluating sharpness of sensors with differing MP by comparing 100% crops will always favor the lower MP sensor. It is a fundamentally flawed comparison in all cases.
- - -
I know it seems like "looking at 100% crops" should be the best, most accurate way to compare sharpness. Unfortunately, it turns out that it is pretty much the worst and least accurate way to evaluate comparative sharpness of different systems unless they use both use the same format and the same sensor pixel dimensions.
When comparing 100% crops from cameras with sensors that have different MP resolution and therefore different pixel dimensions, the 100% magnification crop from the higher MP camera is equivalent to looking more closely at its image than that of the lower MP sensor. (The problem and the confusion it creates has been around for a long time. Over a decade ago I wrote an article explaining it... back when we were comparing much lower resolutions: WHY YOUR 21MP FILE “LOOKS SOFTER” THAN YOUR 12MP FILE AT 100% MAGNIFICATION.)
Let's compare two cameras whose pixel dimensions just happen to match those of the XT4 and the XT5. Let's say that the lower MP camera has pixel dimensions of 6240 x 4160 and the higher MP camera has 7728 x 5152 pixels. (Yes, these numbers represent the XT4 and XT5/XH2.)
Let's look at 1000 x 1000 pixel 100% magnification crops from both cameras. Both crops will be exactly the same size on your screen. But: We should expect the 100% crop from the higher MP system to look softer.
Wait! What? Doesn't that mean that the higher MP camera is softer?
No. It does not.
It is at least as sharp and probably sharper.
In the first case the crop is approximately 15.6% of the image width. In the second the "same" 1000 x 1000 pixel 100% crop is only 13%% of the image width.
That's significant. Keep reading.
Let's say that you used exactly the same lens, same aperture, same shutter speed, same focus, same everything in both cases. If they were equally sharp - and in optical terms they would be identical — the higher MP 100% magnification crop will look slightly less sharp... not because the lens became less sharp or because the sensor is somehow less sharp, but because...
... you are looking more closely at a smaller section of the image.
In fact, you are looking at about a 20% wider section of the sensor in the case of the lower MP sensor, which is significant.(For comparisons, that is almost as big as the percentage width difference between a 36mm wide FF sensor and a 44mm wide miniMF sensor.)
Essentially, this is like looking at a piece of film with a magnifying lens and pronouncing it sharp... then picking up a more powerful magnification lens, looking at the same piece of film a bit closer and now stating that the film became less sharp. ;-)
Comparing sharpness on screens is tricky business, full of all kinds of traps. The 100% crop comparison is one of the most familiar. Another is resizing one of the two compared images to supposedly adjust for its different pixel dimensions. The problem here is that now you are comparing an "interpolated" image against an uninterpolated image. Guess which one gets degraded... and which one doesn't? Either way, the comparison is not valid.
How then to compare them? Not the way it is done in the video, that's for sure. (I don't mean to insult the gentleman doing the review. These things are hard, and the errors are pretty common.)
One option is to compare prints since typically printing at some given size will at least require both images to be interpolated. In addition, it is a real-world comparison. This isn't the only option, but it is the one in which it is easiest to overcome the problems with comparing on-screen.
Or one can just look at images from the target camera on their own and ask, "Is it sharp?"
By the way, this problem crops up in other ways, too. For example...
rbf_ wrote:
I would expect more diffraction from the 40mp sensor...
Nope.
At a given aperture on a APS-C camera there is exactly the same amount of diffraction whether you use s 6MP sensor, a 26MP sensor, a 40MP sensor, or a 400MP sensor. Diffraction is an optical phenomenon and it does not change with sensor resolution.
If you take a 26MP camera and a 40MP camera that are otherwise identical, fit them with exactly the same lens, and use precisely the same settings...
... they will produce exactly the same amount of diffraction blur at any aperture. If you make, for example, a 20" x 30" print from both, the effect of diffraction on each will be identical, regardless of the aperture used. There is no diffraction disadvantage for the higher MP system.
However (#1)..
... if you use an excellent lens, control for camera vibration, focus carefully, and so forth, you might get a bit more sharpness at a slightly larger ideal "diffraction-limited" aperture on the higher MP system.
However (#2)...
... if you compare 100% crops (see above!) you will be, again, looking more closely at a smaller section of the image from the higher MP sample, tricking the unwary into imagining that the diffraction is "larger" if you forget about the problem with 100% crop comparisons.
In the end, the higher MP system will always be at least as sharp as the lower MP system, and in ideal situations it may be a bit sharper. Increasing photo site density does NOT ever reduce the sharpness of the image.
Key point: If all else remains the same, INCREASING MP NEVER MAKES THE IMAGE SOFTER
Edited on Dec 24, 2022 at 12:24 PM · View previous versions
|
I don't need 10mm to 600mm for landscapes but then I am just a hobbyist. Existing GF 23mm, 20-35mm, 35-50mm, 45-100mm, 110mm, 250mm with 1.4xTC cover me well and then I add my Canon 24mm TSE, 50mm TSE and 90mm TSE when I want something different than all normal wide angle FF crowd.
