Diffraction blur with the R5 and four prime lenses

Hi,

I just did a test of sharpness and the impact of diffraction blur, which I wanted to share here.

I used four lenses: EF 35 F1.4L (I), RF 50 F1.2L, RF 85 F1.2L and EF 135 F2.0L. All photos were done with an R5, from a tripod, with a cable trigger, electronic shutter, ISO 100 and were then scaled up to 300 %.

It is interesting to see how each lens has its own sweet spot when it comes to sharpness, depending on how well other optical aberrations are corrected at the wider apertures. For the RF 50 F1.2L and the RF 85 F1.2L it seems the sweet spot is already somewhere at F2.8, because other aberrations are well under control at that point, and diffraction blur starts to weigh in. For the other two lenses the sweet spot seems to be at F5.6 and F8.0, because till then the other aberrations outweigh the diffraction blur.

But have a look yourself, here are my results:

And here the whole scene for each lens, so you can see how the test setup looked like:

Seems like in the real world, or in real world use, you can get away with F11 using all of those lenses except maybe the 35mm.

RogerZoul wrote:

Seems like in the real world, or in real world use, you can get away with F11 using all of those lenses except maybe the 35mm.

Additionally, these were from the center of the image, which benefits the least from stopping down. I would expect that corners would peak at a somewhat higher f-number.

As a practical matter on most lenses, it does seem that there is only a slight fall-off due to diffraction between f8 and f11 but a more dramatic falloff going to f16 and beyond (you hit the knee of the diffraction curve). Certainly, as you stop down from wide open, the diffraction is getting worse, but this is usually offset by the improvements in other ways by stopping down. With top-notch lenses, you can see the image soften slightly in the center within one stop of wide open, but it is not dramatic as it gets between f11 and f16. With cheap zooms, you may have to go beyond f8 to peak the sharpness.

Karl_Guttag wrote:

RogerZoul wrote:

Seems like in the real world, or in real world use, you can get away with F11 using all of those lenses except maybe the 35mm.

Additionally, these were from the center of the image, which benefits the least from stopping down. I would expect that corners would peak at a somewhat higher f-number.

As a practical matter on most lenses, it does seem that there is only a slight fall-off due to diffraction between f8 and f11 but a more dramatic falloff going to f16 and beyond (you hit the knee of the diffraction curve). Certainly, as you stop down from wide open, the diffraction is getting worse, but this is usually offset by the improvements in other ways by stopping down. With top-notch lenses, you can see the image soften slightly in the center within one stop of wide open, but it is not dramatic as it gets between f11 and f16. With cheap zooms, you may have to go beyond f8 to peak the sharpness.

Good info, thanks!

Karl_Guttag wrote:

RogerZoul wrote:

Seems like in the real world, or in real world use, you can get away with F11 using all of those lenses except maybe the 35mm.

Additionally, these were from the center of the image, which benefits the least from stopping down. I would expect that corners would peak at a somewhat higher f-number.

As a practical matter on most lenses, it does seem that there is only a slight fall-off due to diffraction between f8 and f11 but a more dramatic falloff going to f16 and beyond (you hit the knee of the diffraction curve). Certainly, as you stop down from wide open, the diffraction is getting worse, but this is usually offset by the improvements in other ways by stopping down. With top-notch lenses, you can see the image soften slightly in the center within one stop of wide open,

That's new information for me. So diffraction also plays a role before the so called "diffraction limited aperture", not only beyond it?

but it is not dramatic as it gets between f11 and f16. With cheap zooms, you may have to go beyond f8 to peak the sharpness.

thunder storm wrote:

Karl_Guttag wrote:

RogerZoul wrote:

Seems like in the real world, or in real world use, you can get away with F11 using all of those lenses except maybe the 35mm.

Additionally, these were from the center of the image, which benefits the least from stopping down. I would expect that corners would peak at a somewhat higher f-number.

As a practical matter on most lenses, it does seem that there is only a slight fall-off due to diffraction between f8 and f11 but a more dramatic falloff going to f16 and beyond (you hit the knee of the diffraction curve). Certainly, as you stop down from wide open, the diffraction is getting worse, but this is usually offset by the improvements in other ways by stopping down. With top-notch lenses, you can see the image soften slightly in the center within one stop of wide open,

That's new information for me. So diffraction also plays a role before the so called "diffraction limited aperture", not only beyond it?

"Diffraction limit" is basically when you can't resolve two lines due to diffraction. Diffraction is an analog effect and not all or nothing. Whether your images are limited by diffraction depends on many factors, including pixel size. As pixels get smaller (and thus can resolve more lines per mm) diffraction has a more significant effect.

but it is not dramatic as it gets between f11 and f16. With cheap zooms, you may have to go beyond f8 to peak the sharpness.

The "generic" curve below from a Canon white paper shows that "focus aberrations" improve dramatically as the camera stops down while diffraction gradually worsens. If you combine the red and black curves, you get the net.

Then you have to consider that this curve is for one spot on the image; you will find the red curve for the outside part of the image will be much worse than for the center. As you stop down, the outsides will improve as the center will start to soften. So there is a bit of a balancing act as to which aperture would give the best overall sharpest image.

The bottom line is that a top-quality lens in the center with a high-resolution camera will start softening due to diffraction almost immediately.

From: https://teltec.de/out/media/canon-eos-r-white-paper.pdf

Thank you, Karl!

Karl_Guttag wrote:

"Diffraction limit" is basically when you can't resolve two lines due to diffraction. Diffraction is an analog effect and not all or nothing. Whether your images are limited by diffraction depends on many factors, including pixel size. As pixels get smaller (and thus can resolve more lines per mm) diffraction has a more significant effect.

The bottom line is that a top-quality lens in the center with a high-resolution camera will start softening due to diffraction almost immediately.

From: https://teltec.de/out/media/canon-eos-r-white-paper.pdf

Thanks Karl -- I have attached a comparison I prepared of current and potential (shown in salmon pink) Camera Body Sensor Resolutions and Diffraction "limits". The term "limit" is not a hard stop, rather it is the aperture beyond which (ie more closed down) the impact of diffraction becomes increasingly visible reducing the effective resolution of images taken.

The Pixel Wars in 35mm, APS-C , Micro 4/3rd and SmartPhones are just "stupid".

I did a lot of looking and a little analysis of what I found. I can thoroughly recommend the following websites: Digital Picture , Cambridge in Colour, PhotoPills.com and by far the most detail and useful Scantips.com.

The 2nd argument against ever more MP and higher and higher resolution is all about the adverse impact of diffraction, which becomes increasingly visible with ever finer and finer the pixel pitch (and high and higher resolution per mm2) at lower and lower aperture values. The first is genuinely "who needs it" - sure fast card makers and computer and SSD manufacturers love it. Rather like 8.1k 12-bit RAW videos -- why?

The data is clear the effects of diffraction may start to be seen when shooting Sony AR7 IV with its 60.2 MP sensor at f/5.7 and will certainly be apparent when closed down to f/8.5 and beyond. A 90mp sensor on a A7R V or Z8 or R5? and whatever follows would be even more impacted - diffraction may be seen at f/4.4-4.5 and definately be visible at f/6.6-6.8 (the sensors are very different sizes).

AND the same is seen on the 32.3MP APSC cropped sensor Canon R7 (f/4.7 - f/7.3). This means that unless one is shooting close to wide open, with all that this means for the depth of focus, the resolution of the images taken with these "ultra" high resolution sensors (with increasingly tiny pixel pitch) will be diffraction limited. This is before considering the resolving capabilities of the lens.

Whereas the same calculation for a ~45mp R5 or Z9 are f/6.6 and f/9.8 and the ~24.3mp R3 or Z6II are f/9 and f/13.4. AND so on.

It is no surprise that higher resolution 35mm sized, Dx/APS-C sized and smaller sensors are adversely impacted by earlier at relatively wide apertures and this adversely impacts "resolution" more and more as apertures are closed down. While it is true that the early effects of diffraction (ie at the values I indicated above) are most probably only visible in bench tests using very fine targets and analytic tools -- gone are the days one uses a lens at f/16 or even f/32 on a high res sensor to shoot still life. Now we must routinely use automatic focus stacking at the optimal point for the lens/system one is using -- which is almost always wider open than the values shown above (in in my table).

AND where cameras offer HiRes mode (or pixel shift shooting) extraordinary resolutions and results can be easily achieved with a 35mm system. Of course should one transition to the GFX100 or IQ4 then extraordinary is not enough of a descriptor for what one can do at the highest end of shooting.

ULTIMATELY -- please don't just get sucked into the pixel wars -- buying ever more and more resolution -- you simply become prey for those that sell you fast and faster computers, larger and larger SSDs and yes better and better glass.

How about just buying what you need to do your job or want to enjoy. For some of us that is 45mp 20fps/30fps and even 11mp at 120fps (once or twice in 7 months) -- which is perfect and my camera used rarely at those settings. Most often I shoot single-shot or 3-5fps. For others (including most news and sports shooters) 19-24mp JPG is what they take, sell and squirt to their photo-desks within seconds and they do more than just fine. Most tourists can barely handle 5-11mp and this is already far more than they need. They certainly do not have the desire to buy or carry ultra-fast lenses with very high resolving capability and transmission losses. So what is to gain from putting a 18-300 travel lens on a 60mp sensor -- only the salesperson paid commision does well from that deal.

Both a large DOF and a high sensor resolution are about sharpness. If sharpness from a high sensor resolution isn't that important that's fine. However, why would sharpness from falling into a larger DOF be that important at the same time?

ajm057 wrote:

Thanks for the analytical information and good sources. I'm snipping various comments you made in this response to keep things manageable.

The Pixel Wars in 35mm, APS-C , Micro 4/3rd and SmartPhones are just "stupid".

To a degree, this is true. But based on my tests (and I think your information), we are not at the limit yet with the R5 at 45mp. I have compared RP and R5 test charts with the same lenses, and the result is that you get the linear resolution improvement of the sensors.

Based on your information, we are probably right on the edge with the R7. I'm still waiting to see a really good study. I had an R7 backordered but canceled it based on factors other than resolution.

I did a lot of looking and a little analysis of what I found. I can thoroughly recommend the following websites: Digital Picture , Cambridge in Colour, PhotoPills.com and by far the most detail and useful Scantips.com.

The 2nd argument against ever more MP and higher and higher resolution is all about the adverse impact of diffraction, which becomes increasingly visible with ever finer and finer the pixel pitch (and high and higher resolution per mm2) at lower and lower aperture values. The first is genuinely "who needs it" - sure fast card makers and computer and SSD manufacturers love it. Rather like 8.1k 12-bit RAW videos -- why?

Hey, cards and data and even computing today are next to free. Around 2001, I bought a 340 Megabyte IBM Microdrive CompactFlash (CF) Type II for about $400 (~$650 in 2022 dollars and over $1/megabyte). That drive would hold about 15 cRAW R5 pictures.

My only problem is when I fill up a 256GB flash drive (thanks to the 20 frames/second) it is a pain to back up.

The data is clear the effects of diffraction may start to be seen when shooting Sony AR7 IV with its 60.2 MP sensor at f/5.7 and will certainly be apparent when closed down to f/8.5 and beyond. A 90mp sensor on a A7R V or Z8 or R5? and whatever follows would be even more impacted - diffraction may be seen at f/4.4-4.5 and definately be visible at f/6.6-6.8 (the sensors are very different sizes).

AND the same is seen on the 32.3MP APSC cropped sensor Canon R7 (f/4.7 - f/7.3). This means that unless one is shooting close to wide open, with all that this means for the depth of focus, the resolution of the images taken with these "ultra" high resolution sensors (with increasingly tiny pixel pitch) will be diffraction limited. This is before considering the resolving capabilities of the lens.

I will agree there is essentially little headroom left for smaller pixels. Maybe they can get "clever" with some tricks, but there is not a lot of room left before physics limits everything.

Based on what I see from my R5 test charts, I'm not sure that the "net resolution" won't be OK at f8.

I have looked at a bunch of RF lenses over a wide range of focal lengths and from their widest apertures to at least f16 on my R5 (16f2.8, 85f2, 50f1.8, 15-35f2.8, 24-70f2.8, 70-200f2.8, 100-500, 24-240, 100f2.8). And I fairly consistently see:

• All lenses (even the RF24-240 at all focal lengths) are sharp (or nearly so in the case of the RF24-2450) to the limit of the R5's resolution, wide open in the center.
• The IS consumer prime lenses (I only have the 85f2, so also based on other people's postings) are as sharp as the L-zooms at similar focal lengths in the corners.
• The "consumer" non-IS (RF16 and RF50f1.8) are soft and have low contrast in the corners. The RF24--240 is soft in the corners but has a pretty good contrast.
• Even looking at test charts (that make it much easier to see issues) magnified by 200%, I only see almost no reduction in net sharpness (of course, improvements with most lenses) from wide open to f8 and then only a small difference between f8 and f11. Lenses that are very sharp wide open don't lose a barely visible amount of sharpness through f8.
• Between f11 and f16, I consistently see a noticeable (maybe 25%) fall off in sharpness with the various lenses at various focal lengths.

These observations are on test charts, and it would be even harder to tell the differences with real-world photos. I concluded that with the R5 with any lens, I could go to f11 without a resolution penalty. Beyond f11 with the R5 is when the trade-off of DoF versus resolution kicks in. Based on your information, it may be one stop less for the R7.

ULTIMATELY -- please don't just get sucked into the pixel wars -- buying ever more and more resolution -- you simply become prey for those that sell you fast and faster computers, larger and larger SSDs and yes better and better glass.

An ILC camera is a jack of all trades device. My two most used lenses are my RF100-500 and my RF15-35f2.8. I'm doing radically different things with the two lenses shooting mostly moving things with the RF100-500 and pretty still things with the wide angle lens. I like having all the pixels with the R5 so I can crop.

As I wrote before, the storage devices are essentially free.

The quality of the lenses are amazing compared to where they were years ago. Between the ability to deal with noise and the quality of the lenses plus IS and IBIS, I rarely want a flash or a tripod.

The lenses are crazy expensive but adjusted for inflation, not any more so than years ago. Adjusted for inflation, my R5 cost about the same as my 3.1MP APS-C D30 did in 2000.

The big problem ILC cameras have is that smartphone cameras are "free" (come with all phones) and take pretty good pictures. Certainly more than good enough for posting online. Unless you are willing to spend north of $2,000, it gets hard to beat a smartphone.

I think a huge issue for the camera companies is where they can go from here. Smartphones keep improving and are introducing somewhat telephoto lenses. They are also doing much more with computational photography.

How about just buying what you need to do your job or want to enjoy. For some of us that is 45mp 20fps/30fps and even 11mp at 120fps (once or twice in 7 months) -- which is perfect and my camera used rarely at those settings. Most often I shoot single-shot or 3-5fps. For others (including most news and sports shooters) 19-24mp JPG is what they take, sell and squirt to their photo-desks within seconds and they do more than just fine. Most tourists can barely handle 5-11mp and this is already far more than they need. They certainly do not have the desire to buy or carry ultra-fast lenses with very high resolving capability and transmission losses. So what is to gain from putting a 18-300 travel lens on a 60mp sensor -- only the salesperson paid commision does well from that deal.

Unless a lot of people don't heed this advice, the camera companies are going to stop making cameras. Except for a few "blogging cameras," cameras with fixed lenses are pretty much dead, and I think the low end of ILC is very much in danger. You need a pretty good camera and lens combination to demonstrate much of an advantage of a smartphone. The tourist that can't handle 5-11mp is already gone and using smartphones where 10 to 15 years ago they might have a Rebel DLSR camera.

Unfortunately, in the current climate with smartphones, a "low-cost" ILC is a bit of an anachronism. You either move up to better cameras and lenses or stick with a smartphone. Sure there will be some, but not enough for camera companies to develop new cameras and lenses. The "dollars times the volume" is too small.

I really like my RF24-240 on both the RP and R5. It is a fun lens if I am only going to take one lens with me. These days, I mostly put it on my RP when I have the 100-500 on the R5. Still, even the 24-240 is very sharp in the center on the R5 (but suffers greatly in the corners).

Karl_Guttag wrote:

Additionally, these were from the center of the image, which benefits the least from stopping down. I would expect that corners would peak at a somewhat higher f-number.

I have also done a series of the corners, see below. Images are not taken from the extreme corners but roughly at 2/3 of the line between the center and the corner. By and large it seems to me that the aperture of highest sharpness remains the same as in the center, the RF glass is already very sharp at F2.8, while the older EF glass peaks at F5.6 to F8.0. The biggest difference is seen wide open, where the quality of the EF 35 F1.4L (I) is noticeable worse than in its center and the EF 135 F2.0L also shows some falloff, albeit to a lesser extend. Compared to that, the newer RF glass already wide open delivers very good results in the corners.

Here are the images:

And here the overview of the complete test scene:

Thanks,

In looking at your images, I see a lot of correlation with what I have found with the RF-L zoom lenses (reported here https://www.dpreview.com/forums/post/66394314). Namely, the sharpness will peak near f2.8 or f4 and then remain pretty flat through f5.6. It might dip almost imperceptible at f8, drop off very slightly at f11 and then drop significantly at f16.

I took your two sets of images for the 50f1.2 and 85f1.2 (putting the corners and centers adjacent to each other) and then ran a "find edges" to help identify the sharpest part of the background texture (based on a trick taught by Roger Cicala at Lensrental: https://www.lensrentals.com/blog/2016/09/fun-with-field-of-focus-part-1/).

The find edges operation causes textures to be darker where they are sharper. I marked below where it looked to me like the peak sharpness. Note it is very flat from f2 through f5.6 for both lenses. As expected, the mid-corners peaked at a high f-number. Similarly, for worse lenses, the peak sharpness tends to be at higher f-numbers, but then sharpness starts falling off slightly at f11 and then more severely by f16.

A bit of a surprise for me in the whole process of checking many lenses I have tested is the consistency with which the diffraction is a function of f-number (a ratio) and not the absolute dimension of the aperture regardless of the focal length or lens quality. I consider f11 to be kind of the sweat spot for improving DoF without significantly hurting the center resolution. This would also suggest that the RF600f11 and RF800f11 are not significantly hurt by diffraction due to the higher f-number.

theninth wrote:

Karl_Guttag wrote:

Additionally, these were from the center of the image, which benefits the least from stopping down. I would expect that corners would peak at a somewhat higher f-number.

I have also done a series of the corners, see below. Images are not taken from the extreme corners but roughly at 2/3 of the line between the center and the corner. By and large it seems to me that the aperture of highest sharpness remains the same as in the center, the RF glass is already very sharp at F2.8, while the older EF glass peaks at F5.6 to F8.0. The biggest difference is seen wide open, where the quality of the EF 35 F1.4L (I) is noticeable worse than in its center and the EF 135 F2.0L also shows some falloff, albeit to a lesser extend. Compared to that, the newer RF glass already wide open delivers very good results in the corners.

Here are the images:

And here the overview of the complete test scene:

Sittatunga wrote:

Karl_Guttag wrote:

...

A bit of a surprise for me in the whole process of checking many lenses I have tested is the consistency with which the diffraction is a function of f-number (a ratio) and not the absolute dimension of the aperture regardless of the focal length or lens quality.

That's what the mathematics predict. The angular size of the Airy Disc for a circular aperture is 1.22λ/d in radians, where λ is the wavelength and d is the diameter of the aperture. For a lens focal length f, focussed at infinity the size of the Airy Disk will be f(1.22λ/d). But (from the definition of the f-number) at any f-number n, d =f/n, so the size of the Airy Disk on the sensor will be f(1.22λ/(f/n). The fs cancel out, leaving it as 1.22λn, a constant multiple of the f-number.

Thanks for breaking down the math. Nice that math agrees with the experimental results. The experimental results show at what f-number with an R5's sensor the diffraction becomes significant.

I had read for years that a perfect lens would be the sharpest wide open due to diffraction. I was surprised that the diffraction losses were barely visible even with a 45MP sensor until f8. Also, f11 is only slightly worse than f8, and it was not until f16 that the diffraction became significant.

Hi -- yes for me the 68mp full-frame sensor or 33-35 APSC/DX would as far I I would be willing to go. The big next step for the 35mm sensor is 16-bit colour depth and some of the other tools / capabilities that I have access to when shooting with a Phase One IQ4 back. The technology must be close I  35mm and yes this will cost a little more, but nowhere near the price of Phase One or Hassi.

Karl_Guttag wrote:
I took your two sets of images for the 50f1.2 and 85f1.2 (putting the corners and centers adjacent to each other) and then ran a "find edges" to help identify the sharpest part of the background texture

Interesting, thanks!

The find edges operation causes textures to be darker where they are sharper. I marked below where it looked to me like the peak sharpness. Note it is very flat from f2 through f5.6 for both lenses.

Yes, may perception of the peak was a bit different but it is a really close call, the results are really quite good even wide-open, great when stopped down only a little bit and remain so for a few stops.