Jack Flesher wrote:
Exactly, because f2.8 is before the diffraction aperture for the 60mp sensor and f4 is about the beginning of diffraction. So you get the most squeeze from a lens over a high res sensor if it’s at its peak optical performance before f4.
The thing is it isn't, for my tastes anyway the diffraction aperture, where I want to stop stopping down. For landscapes I often want to go at least a stop beyond the diffraction aperture, which although stopping down that far typically weakens performance in the center it often continues to improve performance on the outer part of the frame. I like that more balanced performance in a lot of shots. In my view, one should not take "diffraction aperture," as meaning, "don't ever use apertures smaller than that. At times, I at least, like to stop down below a hypothetical aperture and I think on balance the lens performs better because of it.
Steve Spencer wrote:
The thing is it isn't, for my tastes anyway the diffraction aperture, where I want to stop stopping down. For landscapes I often want to go at least a stop beyond the diffraction aperture, which although stopping down that far typically weakens performance in the center it often continues to improve performance on the outer part of the frame. I like that more balanced performance in a lot of shots. In my view, one should not take "diffraction aperture," as meaning, "don't ever use apertures smaller than that. At times, I at least, like to stop down below a hypothetical aperture and I think on balance the lens performs better because of it....Show more →
That's all fine, but you are not "gaining" any performance from adding more pixels underneath a lens that is already being out-resolved by the current sensor when you exceed diffraction limits. You are stopping a given lens down to where the outer ⅓ performance of that particular lens improves and the sensor can record it, but at the same time the superior center performance area has pixels that are now being hampered due to diffraction. I agree it's usually a smart tradeoff and especially for landscapes; plus the added benefit of greater DoF from the smaller aperture usually also represents an overall resolution gain even though there will be a corresponding per-pixel detail loss in other areas. Any way you slice it, once you exceed the diffraction aperture for a given sensor, diffraction kicks in and optical resolution deteriorates. The deterioration may not be as significant as resolution gains in other parts of the image, but adding more pixels in this example lends nothing to further the image from a visible detail at the pixel level perspective.
PS: And we're currently at a point pixel-count wise where the human eye isn't going to see 99.99% of any of this we're discussing when viewing a full image at a normal viewing distance for whatever its output medium is.
Jul 03, 2026 at 02:04 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
Jack Flesher wrote:
That's all fine, but you are not "gaining" any performance from adding more pixels underneath a lens that is already being out-resolved by the current sensor. You are stopping a given lens down to where the outer ⅓ performance of that particular lens improves and the sensor can record it, but at the same time the superior center performance area has pixels that are now being hampered due to diffraction. I agree it's usually a smart tradeoff and especially for landscapes; plus the added benefit of greater DoF from the smaller aperture usually also represents an overall resolution gain even though there will be a corresponding per-pixel detail loss in other areas. Any way you slice it, once you exceed the diffraction aperture for a given sensor, diffraction kicks in and optical resolution deteriorates. The deterioration may not be as significant as resolution gains in other parts of the image, but adding more pixels in this example lends nothing to further the image from a visible detail at the pixel level perspective. ...Show more →
I think we mostly agree but on two points we disagree. I don't like describing a lens as being "out resolved," by the sensor. That is misleading. Even a coke bottle will be have slightly higher system MTF on a higher resolution sensor than a lower resolution sensor. It is just there are diminishing returns and the improvement for the higher resolution sensor will be so small it won't be noticeable. So, it isn't as if the lens is worse on the higher resolution sensor it just isn't perceptibly better.
Second, I don't like calling a given aperture the "diffraction" aperture. That too is misleading because it suggests it shouldn't be exceeded. Instead, that aperture is when diffraction begins to weaken performance, but in practice unless you are really pixel peeping at that aperture and focussing on the very best region for performance (typically the center), then you won't see any effect. Step back from a print or magnify the image less and there won't be any perceptible affect at all. The place where you can start to see effects at really high magnification is not the place to become concerned with diffraction, IMO. It is when the effect that is quite gradual affects how you view the image typically and that will vary from person to person and in how the image is presented and in how picky they are. For me it is not at the "diffraction" aperture that I get worried about using a lens because of diffraction. It is at a stop or two or sometimes three from that.
Steve Spencer wrote:
I think we mostly agree but on two points we disagree. I don't like describing a lens as being "out resolved," by the sensor. That is misleading. Even a coke bottle will be have slightly higher system MTF on a higher resolution sensor than a lower resolution sensor. It is just there are diminishing returns and the improvement for the higher resolution sensor will be so small it won't be noticeable. So, it isn't as if the lens is worse on the higher resolution sensor it just isn't perceptibly better.
Second, I don't like calling a given aperture the "diffraction" aperture. That too is misleading because it suggests it shouldn't be exceeded. Instead, that aperture is when diffraction begins to weaken performance, but in practice unless you are really pixel peeping at that aperture and focussing on the very best region for performance (typically the center), then you won't see any effect. Step back from a print or magnify the image less and there won't be any perceptible affect at all. The place where you can start to see effects at really high magnification is not the place to become concerned with diffraction, IMO. It is when the effect that is quite gradual affects how you view the image typically and that will vary from person to person and in how the image is presented and in how picky they are. For me it is not at the "diffraction" aperture that I get worried about using a lens because of diffraction. It is at a stop or two or sometimes three from that....Show more →
I suggest you do some research on Rayleigh and even Dawes (more forgiving) limits. No optical system can exceed these without interpolation based on deconvolution or interferometry...
Unfortunately the Bayer filter and processing throw in a monkey wrench to the IQ. I'd be happy with fewer than 67Mpixels in a Foveon type sensor. Getting only 12.5M reds from the 50MP sensor is a rip-off. I'm hoping for 100MP in my lifetime. Even some cellphones have 200MP cameras.
I used the A7RVI and the new 100-400 for an airshow/flyover this morning, along with the A1II+400-800. The EVF of the A7RVI seemed very choppy, not something I expect to see from a stacked sensor. I was using electronic shutter. I vaguely remember someone talked the chopping EVE due to certain settings?
Douglas L wrote:
I used the A7RVI and the new 100-400 for an airshow/flyover this morning, along with the A1II+400-800. The EVF of the A7RVI seemed very choppy, not something I expect to see from a stacked sensor. I was using electronic shutter. I vaguely remember someone talked the chopping EVE due to certain settings?
Douglas L wrote:
I used the A7RVI and the new 100-400 for an airshow/flyover this morning, along with the A1II+400-800. The EVF of the A7RVI seemed very choppy, not something I expect to see from a stacked sensor. I was using electronic shutter. I vaguely remember someone talked the chopping EVE due to certain settings?
Here is an early A7Riv comparison to the older A7Riii. Fred wanted to find out if resolution above 42Mpix resulted in greater image resolution. Based on the discussion above, 42Mpix should have been able to fully resolve, or mostly resolve the lens resolution.
bwcolor wrote:
So, real world improvements in IBIS, greater dynamic range and lower noise are future improvements that might benefit the final image vs. continuing to offer higher and higher Mpixel sensors when resolution is limited by lens optics.
I don’t think itis quite that simple, and there are several reasons.
First, there are three possible relationships between lens resolution and sensor resolution: they are identical (basically impossible), sensor is better than lens, lens is better than sensor. Of the three, the ideal I think (if resolution is a concern) is to have the sensor output outresolve the best lens at its best setting. So the norm would be to aspire to sensors that have more than enough resolution to work with any lens you throw it it and at any aperture.
Second, there are other potential advantages to higher sensor resolutions. For one, the “grain size” of any noise is smaller with smaller pixels. Until we are no longer concerned about noise… that’s a plus.
Third, while the effect is probably subtle, gradients can be slightly smoother with higher MP sensors.
My position is that as long as the camera is fast enough for the photography I do and the cost doesn’t balloon to ridiculous extremes, it generally makes sense to go ahead and increase sensor resolutions in the highest resolution cameras.
One final thought. Back when we saw cameras achieve 36MP (and before that at lower levels) there we people making essentially the same arguments about the uselessness of increasing sensor resolutions. Yet today, the high speed cameras have resolutions as high or higher…and they are buying them! ;-)
Jul 03, 2026 at 08:12 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
Jack Flesher wrote:
I suggest you do some research on Rayleigh and even Dawes (more forgiving) limits. No optical system can exceed these without interpolation based on deconvolution or interferometry...
Thanks for the suggestions. I do know about such limits and have a rough understanding of the physics behind them. The physics doesn't bother me. It is the terminology that does. Specifically, I do not like the term limit even though I understand why they chose it. I would much prefer the term threshold, as it would imply that diffraction begins to become an issue. So my quibble is about the terminology as it is applied to amateur photographers, not as it is used by physicists. That is just me and my preferences about how I would like to talk about these issues. You of course are free to have a totally different take.
Just for edification, the term “limit” has a very specific mathematical meaning. Which is why it’s used when discussing things like diffraction which is in fact derived and evaluated by a mathematical formula. Some of us here were formally trained in mathematics, and thus we will continue to use the terminology we were taught. FWIW.
Jul 03, 2026 at 08:24 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
Jack Flesher wrote:
Just for edification, the term “limit” has a very specific mathematical meaning. Which is why it’s used when discussing things like diffraction which is in fact derived and evaluated by a mathematical formula. Some of us here were formally trained in mathematics, and thus we will continue to use the terminology we were taught. FWIW.
Yes, I know that, but I also spend a fair bit of time thinking about and engaging in science communication. It is my view that terms that make sense in a specific technical sense in a scientific or mathematical field can and often are poor choices to use in less technical settings. I would further argue and I think I have been that the term limit in this context is a rather poor choice for a non-technical audience. You are of course free to disagree and use whatever term in whatever way you want, but I am also free to not like that use of the term. We ought to be able to agree to disagree on this one.
Jack Flesher wrote:
That's all fine, but you are not "gaining" any performance from adding more pixels underneath a lens that is already being out-resolved by the current sensor when you exceed diffraction limits. You are stopping a given lens down to where the outer ⅓ performance of that particular lens improves and the sensor can record it, but at the same time the superior center performance area has pixels that are now being hampered due to diffraction. I agree it's usually a smart tradeoff and especially for landscapes; plus the added benefit of greater DoF from the smaller aperture usually also represents an overall resolution gain even though there will be a corresponding per-pixel detail loss in other areas. Any way you slice it, once you exceed the diffraction aperture for a given sensor, diffraction kicks in and optical resolution deteriorates. The deterioration may not be as significant as resolution gains in other parts of the image, but adding more pixels in this example lends nothing to further the image from a visible detail at the pixel level perspective.
PS: And we're currently at a point pixel-count wise where the human eye isn't going to see 99.99% of any of this we're discussing when viewing a full image at a normal viewing distance for whatever its output medium is. ...Show more →
First, regarding "adding more pixels in this example lends nothing to further the image from a visible detail at the pixel level perspective" - I don't see why we need to be concerned with the "pixel level". I think the practical level of interest is that of the circle of confusion that, typically, can be expected to contain a number (e.g. ten) pixels. Thus, what happens at the pixel level, e.g. diffraction, is not important in practice.
The second consideration is that there are two ways of increasing the number of pixels: one is by fitting more pixels per sensor of a fixed size, and the other is by increasing the sensor size while keeping the pixel density. This second case is distinct in that it is easier to resolve detail in an image projected on a larger sensor, even when the lens in front of the sensor is optically is not better than a reference FF lens.
Thus, while I tend to agree with you on the "diminishing returns" of adding more pixels to a FF sensor, I believe that the larger than FF camera systems that have 100MP and more (with their native system lenses) do outresolve FF camera systems with best FF lenses. Deep cropping with such larger systems also works better than cropping with high-MP FF systems.
ruthenium wrote:
First, regarding "adding more pixels in this example lends nothing to further the image from a visible detail at the pixel level perspective" - I don't see why we need to be concerned with the "pixel level". I think the practical level of interest is that of the circle of confusion that, typically, can be expected to contain a number (e.g. ten) pixels. Thus, what happens at the pixel level, e.g. diffraction, is not important in practice.
The second consideration is that there are two ways of increasing the number of pixels: one is by fitting more pixels per sensor of a fixed size, and the other is by increasing the sensor size while keeping the pixel density. This second case is distinct in that it is easier to resolve detail in an image projected on a larger sensor, even when the lens in front of the sensor is optically is not better than a reference FF lens.
Thus, while I tend to agree with you on the "diminishing returns" of adding more pixels to a FF sensor, I believe that the larger than FF camera systems that have 100MP and more (with their native system lenses) do outresolve FF camera systems with best FF lenses. Deep cropping with such larger systems also works better than cropping with high-MP FF systems....Show more →
We agree 👍 Sensor area improves output resolution all else equal and assuming the lens in question covers the larger sensor.
Jack Flesher wrote:
Just for edification, the term “limit” has a very specific mathematical meaning. Which is why it’s used when discussing things like diffraction which is in fact derived and evaluated by a mathematical formula. Some of us here were formally trained in mathematics, and thus we will continue to use the terminology we were taught. FWIW.
Well, if you do that… I’m going to use musical terminology, including electronic music terminology! (My academic field was music theory and composition, specializing in electronic music.)*
*(“Electronic music” has a different meaning today, for those of you imagining me with a degree in… electronica. ;-)
gdanmitchell wrote:
I don’t think itis quite that simple, and there are several reasons.
First, there are three possible relationships between lens resolution and sensor resolution: they are identical (basically impossible), sensor is better than lens, lens is better than sensor. Of the three, the ideal I think (if resolution is a concern) is to have the sensor output outresolve the best lens at its best setting. So the norm would be to aspire to sensors that have more than enough resolution to work with any lens you throw it it and at any aperture.
Second, there are other potential advantages to higher sensor resolutions. For one, the “grain size” of any noise is smaller with smaller pixels. Until we are no longer concerned about noise… that’s a plus.
Third, while the effect is probably subtle, gradients can be slightly smoother with higher MP sensors.
My position is that as long as the camera is fast enough for the photography I do and the cost doesn’t balloon to ridiculous extremes, it generally makes sense to go ahead and increase sensor resolutions in the highest resolution cameras.
One final thought. Back when we saw cameras achieve 36MP (and before that at lower levels) there we people making essentially the same arguments about the uselessness of increasing sensor resolutions. Yet today, the high speed cameras have resolutions as high or higher…and they are buying them! ;-)...Show more →
The higher resolution sensor with a high resolving lens, did outperform the lower resolution sensor. As mentioned elsewhere, smoother transitions also result from higher resolution sensors, assuming the lens is highly resolving.
Jack Flesher wrote:
I suggest you do some research on Rayleigh and even Dawes (more forgiving) limits. No optical system can exceed these without interpolation based on deconvolution or interferometry...
If you use the Rayleigh Criteria for a diffraction limited f/4 lens, what is the pixel pitch such that a single pixel spans the distance from the peak of the Point Spread Function to its first zero? For a FF sensor, how many pixels is that?
This discussion is a bit more interesting because of the need to consider the Sparrow Criteria, the Bayer array, and what is Critically sampled? You can find a nice presentation of this starting here: https://blog.kasson.com/the-last-word/camera-vs-sensor-resolution/
Why did Sony leave out medium raw? I use it all the time on my A7R5. I love how I can shoot 26mp images and STILL get 26mp when I use crop mode in mRaw. Anyone know why they left this useful tool out on the new camera?
