Diffraction effect (originally titled: Reciprocal Failure) quick test MF

[leonardobarreto.com]

I have been reading a lot about the problem of reciprocal failure of digital cameras. Since this relates to actual size of sensels, I decided to do a quick test with my normal lens 80mm f:2.8 MAMIYA 645 AF

I just shot a newspaper at all available f.stops and processed a small crop in the center. I used one flash head on one side and removed all filters in the lens..

QUICK CONCLUSIONS

I think that f.16 and f.22 are perfectly usable apertures. Probably f.16 being the sweet spot and 2.8 to 5.6 noticeably soft.

What do you think?

 

[james sperry]

this test would have different results with different lenses, wouldn't it? ..... lol ........ maybe i should do this with all of my lenses to find my sweet spots.

thanks for the post!!

 

[leonardobarreto.com]

I think that it could be one other conclusion. I was about to test my other lenses. I have two view camera optics that I use with my Mamiya sometimes (with my SINAR and an adaptor) and want to test them, one is Fujifilm 180mm and the other 135mm Schneider...

Also want to see how my Mamiya 150mm does.

The first idea of this test was to try to find the difference in performance of different digital formats, like DX (I have a Nikon D300) as opposed to large size sensor.

The real world implication may be that a MF camera shot in studio (ISO 100) with tons of flash could be stopped down to f.16 - f.22 while the D300 would have to be shot at f.?

We would have to take in consideration that the normal for the D300 is 35mm while for the Mamiya is 80mm...

with implicit differences in Depth of Field etc.

this test would have different results with different lenses, wouldn't it? ..... lol ........ maybe i should do this with all of my lenses to find my sweet spots.

thanks for the post!!

 

[leonardobarreto.com]

I just shot the 180mm Fujinon with same target and f.16 - f.22 are the SWEET spot of that lens too.

----after f.22 things go down noticeably in resolution -----

I have to stop the experiment project to help my kid with his homework, but I think that it is good to know what to use for each lens to squeeze more resolution out of the captures..

I want to do a test with my Nikkor 85mm f.:1.8 to see what is its sweet spot and how it compares to the Fujinon and Mamiya...

By the way, there is no noticeable quality difference between the 80mm Mamiya and the 180mm Fujinon stopped at f.16

 

[Doug Kerr]

Hi, Leonardo,

I have been reading a lot about the problem of reciprocal failure of digital cameras.

I'm not sure what you mean. From the rest of your discussion, it seems as if you might be speaking of the effect of diffraction, which results in a type of blurring of the image, even for perfect focus, at small apertures.

There is a phenomenon called reciprocity failure. That means that the exposure result from an exposure of 1 second at f/22 would not be the same as for an exposure of 1/4 second at f/11 (which we ordinarily would expect).

But it doesn't seem as if you are talking about exposure matters.

Can you clarify what issue you are discussing?

Thanks.

 

[Daniel Buck]

yes, this isn't a reciprocity thing, I think this is a diffraction thing. However, I notice that your test object doesn't appear to be perfectly parallel to the camera (either that, or the lens isn't aligned correctly on the camera) as in the first few images the left side of the image looks to be a bit more in-focus than the right side. ?

On the note of diffraction, sometimes having an overall slightly softer image is better than having part of the image noticeably out of focus. Depends on the situation. A little contrast boost in an image that shows softening from diffraction can help it a bit

 

[leonardobarreto.com]

Doug

YES, you are correct and I was using a term that relates to film completely confused in my mind with diffraction (what if we said: DIFFRACTION FAILURE ?? the failure term would apply, no? anyway, that is not the point)

So, yes. I am exploring with my own hands -and eyes- the diffraction behavior of my PhaseOne/Mamiya as opposed to my D300.

And talking about that, I just shot a "roll" of images whit D300 and Nikon 50mm f.:1.8 lens at all apertures.

For what I can see the lens performs well at open apertures up to f. 11 and is underperforming already when stopped down at f.16

So it must be something like one f.stop difference from the Medium Format to DX.

I don't know if this is lens related or size-of-sensel related, but it makes sense that a larger sensor would perform better in the face of this phenomenon (Diffraction, that is)

thanks, leonardo
ps. It was a bit like "catch me if you can" after I realized I had made that mistake. Thank you for catching me...

Hi, Leonardo,



I'm not sure what you mean. From the rest of your discussion, it seems as if you might be speaking of the effect of diffraction, which results in a type of blurring of the image, even for perfect focus, at small apertures.

There is a phenomenon called reciprocity failure. That means that the exposure result from an exposure of 1 second at f/22 would not be the same as for an exposure of 1/4 second at f/11 (which we ordinarily would expect).

But it doesn't seem as if you are talking about exposure matters.

Can you clarify what issue you are discussing?

Thanks.

 

[Bart_van_der_Wolf]

I have been reading a lot about the problem of reciprocal failure of digital cameras. Since this relates to actual size of sensels, I decided to do a quick test with my normal lens 80mm f:2.8 MAMIYA 645 AF

I just shot a newspaper at all available f.stops and processed a small crop in the center. I used one flash head on one side and removed all filters in the lens..

Hi Leonardo,

Diffraction patterns are determined by the aperture number, and the shape of the iris. In my experience with DSLRs with an Anti-Aliasing filter and a Bayer CFA, I can notice degradation of the per pixel sharpness when the diameter of the diffraction pattern exceeds 1.5x the sensel pitch. The diameter of the diffraction pattern is commonly taken to be 2.44 x Wavelength x Aperture# , so e.g. (for green light) 2.44 x 0.555 x 8.0 = 10.8336 micron)

The difficulty with shooting a flat surface is that any defocus will add to the unsharpness that's caused by diffraction. So to get a good sense of the degradation, one has to either make sure that the object and sensor plane are absolutely parallel and have the best possible focus (focus bracketing), or shoot a flat surface at an angle (which will guarantee that there is a small zone in absolutely optimal focus, even with field curvature).

So were you shooting film, or did you use the Mamiya 645 AF with a digital back?

QUICK CONCLUSIONS

I think that f.16 and f.22 are perfectly usable apertures. Probably f.16 being the sweet spot and 2.8 to 5.6 noticeably soft.

What do you think?

I assume we're looking at 100% zoom crops? In that case it depends on the output size wether the unsharpness is still acceptable. If you need the absolute best per pixel quality, e.g. because you need to enlarge a lot, and the scene DOF requirements allow to use the optimal aperture, it looks like f/11 (or maybe f/16, depending on where one looks) is best.

Bart

 

[leonardobarreto.com]

Bart_van_der_Wolf:

Obviously your knowledge is in inverse proportion of mine regarding this problem (Diffraction that is) so I could ask:

Should there be one stop difference in sharpness from DX to MF?

I was only joking when I said I shot a "roll", it was more like a series of captures with my digital back (PhaseOne P25), I just couldn't resit.

I think that the flat surface in my tests where not perfectly parallel to the filmplane, and I should do some more tests with 3d targets. It should be interesting to compare effects of a normal lens in DX format as opposed to MF regarding Depth of Field and optimum aperture.

Hi Leonardo,

Diffraction patterns are determined by the aperture number, and the shape of the iris. In my experience with DSLRs with an Anti-Aliasing filter and a Bayer CFA, I can notice degradation of the per pixel sharpness when the diameter of the diffraction pattern exceeds 1.5x the sensel pitch. The diameter of the diffraction pattern is commonly taken to be 2.44 x Wavelength x Aperture# , so e.g. (for green light) 2.44 x 0.555 x 8.0 = 10.8336 micron)

The difficulty with shooting a flat surface is that any defocus will add to the unsharpness that's caused by diffraction. So to get a good sense of the degradation, one has to either make sure that the object and sensor plane are absolutely parallel and have the best possible focus (focus bracketing), or shoot a flat surface at an angle (which will guarantee that there is a small zone in absolutely optimal focus, even with field curvature).

So were you shooting film, or did you use the Mamiya 645 AF with a digital back?



I assume we're looking at 100% zoom crops? In that case it depends on the output size wether the unsharpness is still acceptable. If you need the absolute best per pixel quality, e.g. because you need to enlarge a lot, and the scene DOF requirements allow to use the optimal aperture, it looks like f/11 (or maybe f/16, depending on where one looks) is best.

Bart

 

[Ken Tanaka]

There always seems to be a buzz on photo fora concerning diffraction distortion. But the fact is that it's highly unlikely that anyone will encounter such a phenomenon in any obvious way with a 35mm, or even a medium format, camera. I only know of two photographers who have ever encountered it in any demonstrable way, both with large format cameras and lenses stopped down to pinholes. Most of what is normally paraded as diffraction distortion is actually simple, and far more common, chromatic aberration.

What your images show, if anything, is that sharpness may be falling-off at f/22. That''s common, particularly for close subjects. It's not diffraction distortion.

You've too much free time, Leonardo! ;-)

 

[leonardobarreto.com]

Ken:

How come we used to shoot f. 64 when using film to get the most out of a lens resolving capacity.

In my test the resolution gets progressively worst after f. 16

I think that this is a very visible behavior, and well worth the short time I spent target shooting today.

As a matter of fact, It only opened my appetite for more tests....

 

[Ken Tanaka]

In the absence of particulars I don't know why you did that.

But if you're enjoying the "experiments" please don't let me tarnish your fun.

 

[Bart_van_der_Wolf]

Ken:

How come we used to shoot f. 64 when using film to get the most out of a lens resolving capacity.

When the output magnification is small, then the diffraction or DOF induced unsharpness is too small to be resolved by eye. Output magnification has to be a part of the 'usability' equation.

To resolve something like 5 line-pairs/milimetres, one needs to resolve 0.1 mm detail in output. That would mean something like an 8x magnification of a high resolution film, where the diffraction is minimized. Higher amounts of diffraction will restrict either the output resolution, or the output size with the same quality level. Large format film requires only modest magnification, so diffraction can be pushed further without becoming visible in output.

Bart

 

[Bart_van_der_Wolf]

There always seems to be a buzz on photo fora concerning diffraction distortion. But the fact is that it's highly unlikely that anyone will encounter such a phenomenon in any obvious way with a 35mm, or even a medium format, camera.

Hi Ken,

Sorry, but I have to contradict you on this one. The 'obvious' part of your comment is obviously depending on the output magnification. However, I have already demonstrated the phenomenon with the following example (an oblique close-up of the surface of a banknote):

The camera used (Canon 1Ds3) has its optimum diffraction/sensel ratio at f/7.1 for green light and one can indeed see that the f/8 shot starts to lose resolution at the pixel level, but gains on (lower resolution) DOF. The same principle applies to other sensors, and the sensel pitch is the dominant criterion.

The choice of how much loss of resolution due to diffraction is still tolerable, obviously depends on the total image (e.g. required DOF) and its final use (output maginification). There are also special (deconvolution) sharpening techniques that may be able to recover some of the losses.

Bart

 

[Thierry Hagenauer]

May be interesting to read, if you are interested in diffraction:

http://luminous-landscape.com/forum/index.php?showtopic=31975&pid=259618&st=0&#entry259618

Best regards,
Thierry

 

[leonardobarreto.com]

Thanks, that is a very elegant and clear explanation of diffraction. Glad you had the time to produce it... : )

Hi Ken,

Sorry, but I have to contradict you on this one. The 'obvious' part of your comment is obviously depending on the output magnification. However, I have already demonstrated the phenomenon with the following example (an oblique close-up of the surface of a banknote):

The camera used (Canon 1Ds3) has its optimum diffraction/sensel ratio at f/7.1 for green light and one can indeed see that the f/8 shot starts to lose resolution at the pixel level, but gains on (lower resolution) DOF. The same principle applies to other sensors, and the sensel pitch is the dominant criterion.

The choice of how much loss of resolution due to diffraction is still tolerable, obviously depends on the total image (e.g. required DOF) and its final use (output maginification). There are also special (deconvolution) sharpening techniques that may be able to recover some of the losses.

Bart

 

[John Sheehy]

The camera used (Canon 1Ds3) has its optimum diffraction/sensel ratio at f/7.1 for green light and one can indeed see that the f/8 shot starts to lose resolution at the pixel level, but gains on (lower resolution) DOF. The same principle applies to other sensors, and the sensel pitch is the dominant criterion.

The choice of how much loss of resolution due to diffraction is still tolerable, obviously depends on the total image (e.g. required DOF) and its final use (output maginification). There are also special (deconvolution) sharpening techniques that may be able to recover some of the losses.

That's pretty soft for f/16. I suspect that there is more than diffraction at work in your example. When I use my 50D with 4.7u pixels at f/22 with my Tamron 90mm macro, I can easily pull up contrast near the nyquist from the subject with USM. I can't pull up *anything* but the diffusion error dithering for the indexed GIF near the nyquist in the f/16 frame of your sample, with a lower f-stop and much larger pixels.

 

[Ken Tanaka]

Bart,
That's a pretty cool illustration. I'll take your word on your observation. But that's not really what I see at all. Of course at this tiny size and coarse resolution it's remarkable that we can see any subtlety on any images at all!

I, too have a 1Ds3. When you get a moment could you outline the experimental procedure so that I can give it a whirl here? I think that's the only way I'll be able to see what you see. (Perhaps a more common subject?)

 

[leonardobarreto.com]

Ken

I'm glad to see that you will find time to do tests on you own -and not the easy ones I do but Bart's gourmet kind-.

I send images to my stock agency that are viewed at 100% and rejected for softness and other crimes and misdemeanors, so the cleaner the files look at that magnification the better.

This may also be relevant for people deciding to go MF or 35mm -and also DX- since they may have different points of max resolving capacities.

Please post any results of tests that you do...

 

[Bart_van_der_Wolf]

I, too have a 1Ds3. When you get a moment could you outline the experimental procedure so that I can give it a whirl here? I think that's the only way I'll be able to see what you see. (Perhaps a more common subject?)

Hi Ken,

The example was from a thread in the Canon lens forum. I used a banknote and shot it at an angle of, if I recall correctly, some 40 degrees (close to 45 degrees anyway). After (manual, live view) focusing so that the plane of focus ran through the center of the image, all I did was change aperture and exposure time between shots. The crops were aligned and layered in Photoshop, and converted into a GIF animation with Photoshop CS3.

The example is taken at a larger than true size magnification, hence the narrow DOF. However, it's what I also see in other, more distant, shots.

Once the aperture on a 1Ds3 gets narrower than f/7.1 the resolution suffers from diffraction. The diameter of the diffraction pattern for green light at f/7.1 (2.44 x 0.555 x 7.1 = 9.615 micron) equals 1.5x the 6.4 micron sensel pitch. I've seen the same for the 1Ds2, although at a slightly narrower aperture (f/8), due to the 7.2 micon sensel pitch. I've also seen it on a Powershot G3 (3.2 micron sensel pitch) at f5.6 it suffers from diffraction (at the pixel level) compared to wider apertures.

It is also reflected in the system (not lens only) MTF curves at various apertures, the sharpest lenses generally peak at the same apertures at where the diffraction pattern diameter reaches the 1.5x sensel pitch threshold.

Bart

 

[Bart_van_der_Wolf]

That's pretty soft for f/16. I suspect that there is more than diffraction at work in your example.

It's a macro shot. DOF is measured in fractions of a millimetre.

When I use my 50D with 4.7u pixels at f/22 with my Tamron 90mm macro, I can easily pull up contrast near the nyquist from the subject with USM. I can't pull up *anything* but the diffusion error dithering for the indexed GIF near the nyquist in the f/16 frame of your sample, with a lower f-stop and much larger pixels.

As I stated earlier, part of the diffraction losses can be recovered with (deconvolution) sharpening, but that's another subject. Diffraction is what it is, and although it doesn't change with sensel pitch, the pitch does affect the per pixel quality due to finer or coarser sampling of the diffraction pattern. It can affect e.g. the technical acceptance of images at a stock agency.

Bart

 

[leonardobarreto.com]

It means that camera upgrades, for example, from Canon 1DsM2 to 1DsMk3 or P25 (9 x 9 micron) to P45 (6.8 x 6.8) the forces of diffraction may be compansating megapixel gains.

Or that when you upgrade to a more pixes but samller size sensel, you have to change your shooting workflow to set f.7 instead of f.11

 

[Thierry Hagenauer]

Have a look HOW the pixel pitch DOES affect the effect of diffraction on the sharpness, here:

http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

There is a nice illustration.

Thierry

 

[leonardobarreto.com]

Thierry,

Thank you, this is better than a Wikipedia article for diffraction...

Don't know if I did it correctly, but found that from a P45 to a P25 optimal f.stops goes like f.8 and f.13 or something like one and a half f.stop difference. Probably with the newer backs that have 60mp's there should be even more than that.

What about, for example, the difference in diffraction factor from a Nikon D3X and a digital back with 9 x 9 micron ...

 

[John Sheehy]

Have a look HOW the pixel pitch DOES affect the effect of diffraction on the sharpness, here:

http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

There is a nice illustration.

It seems, however, that many people who look at this site overlook the statement near the end which says that higher pixel densities do not necessarily mean less resolution/detail/sharpness for the image. Their language is a bit too soft. I think it is safe to say that higher pixel density always increases resolution, even in so-called "diffraction-limited" situations, by at least an infinitesimal amount, as long as you keep all things equal by having the same strength AA filter, as measured in pixels.

It is a logical non-sequitor that the fact that higher pixel densities *show* diffraction faster as you climb the f-stops means that higher pixel densities have more diffraction, or that diffraction presents a hard limit in that range, for resolution. The only reason that lower pixel densities seem to hide diffraction is because lower pixel densities themselves limit resolution more than diffraction does.

 

[Doug Kerr]

Hi, John,

It seems, however, that many people who look at this site overlook the statement near the end which says that higher pixel densities do not necessarily mean less resolution/detail/sharpness for the image. Their language is a bit too soft. I think it is safe to say that higher pixel density always increases resolution, even in so-called "diffraction-limited" situations, by at least an infinitesimal amount, as long as you keep all things equal by having the same strength AA filter, as measured in pixels.

It is a logical non-sequitor that the fact that higher pixel densities *show* diffraction faster as you climb the f-stops means that higher pixel densities have more diffraction, or that diffraction presents a hard limit in that range, for resolution. The only reason that lower pixel densities seem to hide diffraction is because lower pixel densities themselves limit resolution more than diffraction does.

Well said.

Best regards,

Doug

 

[David A. Goldfarb]

In the banknote example, the effect of diffraction is exaggerated, because the effective aperture is considerably smaller than the f:number set on the lens, isn't it? I don't recall offhand, but you remember these things better than I do, Bart. Is the effective aperture relevant for calculating the maximum resolution limited by diffraction? It seems like it should, and if it is, then the effect of diffraction at a given f:stop should be less in a landscape than in a macro shot.

 

[Bart_van_der_Wolf]

It seems, however, that many people who look at this site overlook the statement near the end which says that higher pixel densities do not necessarily mean less resolution/detail/sharpness for the image. Their language is a bit too soft. I think it is safe to say that higher pixel density always increases resolution, even in so-called "diffraction-limited" situations, by at least an infinitesimal amount, as long as you keep all things equal by having the same strength AA filter, as measured in pixels.

It is a logical non-sequitor that the fact that higher pixel densities *show* diffraction faster as you climb the f-stops means that higher pixel densities have more diffraction, or that diffraction presents a hard limit in that range, for resolution. The only reason that lower pixel densities seem to hide diffraction is because lower pixel densities themselves limit resolution more than diffraction does.

I fully agree. I'l repeat what I've said earlier, diffraction at a given aperture is what it is. A smaller sensel pitch only samples that diffraction pattern more accurately. The drawback of that for huge enlargements is that diffraction becomes visible at wider apertures, so for a real increase in resolution it's best to avoid such losses (if the DOF requirements allow).

Bart

 

[Bart_van_der_Wolf]

In the banknote example, the effect of diffraction is exaggerated, because the effective aperture is considerably smaller than the f:number set on the lens, isn't it? I don't recall offhand, but you remember these things better than I do, Bart. Is the effective aperture relevant for calculating the maximum resolution limited by diffraction? It seems like it should, and if it is, then the effect of diffraction at a given f:stop should be less in a landscape than in a macro shot.

The effective aperture has to do with only using a smaller area of the projected image, the available photons are spread thinner due to the magnification. The formula is (M+1)^2, where M stands for the magnification factor. So e.g. at 1:1 macro (M=1), one has to quadruple the exposure when compared to infinity focus.

Diffraction however has to do with the wavefronts created by the edge of the aperture blades, they are not magnified by the optics like the in-focus subject matter is. Yes, the rays have to travel further with macro photography, but we are talking about wavelength effects with a very high frequency, so they'd require significant magnification to notice any effects, even in macro photography.

I'll digg up some aperture bracketing images to demonstrate the principle at close to infinity focus.

Bart

 

[Thierry Hagenauer]

hi Bart,

in any case, when I was shooting on a 4x5", with film, the diffraction effect was MUCH amplified with the reproduction scale (magnification). At 1:1, e.g., one could visibly see diffraction with f45+, when at infinity, such a closed f-stop would not have any influence. And of course, at higher reproduction scales, e.g. 2:1 this was much worse.

How much the reproduction scale plays on a digital sensor like we are used to have them currently, I couldn't say for not having tested it, but certainly it does from a certain scale, respectively from a certain lens to image plane length.

Kind regards,
Thierry

The effective aperture has to do with only using a smaller area of the projected image, the available photons are spread thinner due to the magnification. The formula is (M+1)^2, where M stands for the magnification factor. So e.g. at 1:1 macro (M=1), one has to quadruple the exposure when compared to infinity focus.

Diffraction however has to do with the wavefronts created by the edge of the aperture blades, they are not magnified by the optics like the in-focus subject matter is. Yes, the rays have to travel further with macro photography, but we are talking about wavelength effects with a very high frequency, so they'd require significant magnification to notice any effects, even in macro photography.

I'll digg up some aperture bracketing images to demonstrate the principle at close to infinity focus.

Bart