Lens Diffraction - When is it an issue with Medium Format Lenses?

[theguywitha645d]

The software is ImageJ, free for anyone who wants to use it. Google and name and you should find it easily

This is not a test target and so you need to dig a little further. The plot of the image is far more complex, which is why just using resolving power to make some sort of grid to define "equivalent MP" does not really work--you are not going to find convenient line pairs and line pairs in and of themselves do not describe an image (it is just a quantification of system that allows comparison, but does not have enough detail to express image formation or perception). Since the 645D has a pixel pitch of 6um, the image does actually provide clear separation of detail between 1-2 pixels or 6um-12um. You can see that in the image and if you analyze the plot, it is there as well--as in other areas of the image. You will also see it is defining edges very well. Just taking peaks on the plot will not work.

This is also a low contrast target which would lower the resolving power of the system. Which suggests that the Rayleigh criterion/limit, which you use in your post, is far too crude to model real systems. Actually, the Rayleigh limit is really just a useful learning tool, but it is out of date.

[John R Smith]

I've tested all of my lenses with my MFD back at distances that are relevant to my subject matter and my output. I recommend that anyone who is concerned about diffraction does the same.

Well spoken. 'Tis the only way to do it. And once you have done it, internalise the results and stop worrying about it.

I did some testing recently off a tripod using a subject with mixed distances from about 30 feet to infinity, to check a couple of Zeiss 60mm lenses which I was servicing. Folks, don't you worry too much about f22. In this case it was f4 which was scary - I was horrified how the image fell apart when those lenses were wide-open. The corners with these old Distagons are not that great at the best of times, but . . .

John

[theguywitha645d]

I've tested all of my lenses with my MFD back at distances that are relevant to my subject matter and my output.

That's just crazy talk!

[bjanes]

The software is ImageJ, free for anyone who wants to use it. Google and name and you should find it easily

Thanks. I do use ImageJ extensively, but did not recognize the output in your graph.

This is not a test target and so you need to dig a little further. The plot of the image is far more complex, which is why just using resolving power to make some sort of grid to define "equivalent MP" does not really work--you are not going to find convenient line pairs and line pairs in and of themselves do not describe an image (it is just a quantification of system that allows comparison, but does not have enough detail to express image formation or perception). Since the 645D has a pixel pitch of 6um, the image does actually provide clear separation of detail between 1-2 pixels or 6um-12um. You can see that in the image and if you analyze the plot, it is there as well--as in other areas of the image. You will also see it is defining edges very well. Just taking peaks on the plot will not work.

Your test method seems quite convoluted and I doubt that it is validated. Why don't you use a validated method? Some choices are:

ISO 12233 Chart: allows visual determination of resolution by inspection and the slanted edge can be used with Imatest and other software. However, one must calibrate for the magnification factor if using the visual method. With the slanted edge target, the magnification is not critical.

Koren Resolution Chart: Can be used visually and with ImageJ, where one measures peak to peak. The magnification factor must be accounted for.

Van Der Wolf modified Siemens star: Target can be used at any distance and the slanted edge can be used with Imatest.

This is also a low contrast target which would lower the resolving power of the system. Which suggests that the Rayleigh criterion/limit, which you use in your post, is far too crude to model real systems. Actually, the Rayleigh limit is really just a useful learning tool, but it is out of date.

The Rayleigh  criterion is hardly out of date and, along with the Dawes criterion, is a fundamental concept in optics. The old USAF charts measure resolution near the Rayleigh limit and this is the result you get when using this method. As I stated, a MTF of 9% is too low to be used in terrestrial photography. The Dawes limit is where MTF is zero. A MTF of 50% corresponds best with perceived sharpness. What MTF did you get with your method. Resolution without reference to contrast is not that helpful.

Regards,

Bill

[DeeJay]

Depends from scene to scene. Lighting to lighting. Lens to lens.

I see it a more on my P65 with V Series lenses more so than the H Series which I don't really see it.

It's always cleaned up with some sharpening though and hasn't ever really been a problem.

It starts to appear at f16.

It's never been an issue when I used the V with a P45. Probably more exaggerated with the IQ180.

[theguywitha645d]

Your test method seems quite convoluted and I doubt that it is validated. Why don't you use a validated method? Regards,

Bill

They are my tests for myself--you had challenged us to show results. Also, test charts do not provide the information I need. These test give very useful data. And the final part of the test is always real output.

Bill, lets see some of your tests on diffraction.

[bjanes]

They are my tests for myself--you had challenged us to show results. Also, test charts do not provide the information I need. These test give very useful data. And the final part of the test is always real output.

Bill, lets see some of your tests on diffraction.

Certainly. I used Bart's target with my Nikon D3 with the 60 mm f/2.8 AFS Micro-Nikkor. I rendered the images with ACR 6.5 with the default settings, which include a slight amount of capture sharpening.

Here is the result for f4.0 with Bart's method and with Imatest. The blur circle is 98 pixlels giving a resolution of 55 cy/mm or 0.47 cy/pixel, near the Nyquist limit of 0.5 cy/px. Bart's chart shows aliasing particularly well, but I think that the resolution results are for a relatively low MTF, perhaps near Rayleigh. Maybe Bart can comment if he is following this thread. Bart's chart also allows visual inspection of the test images; the low contrast at f/32 is evident at a quick glance, more so than the reduced resolution. The Imatest results are self explanatory. The resolution per picture height is not valid since I used cropped images, but the other figures are valid. Imatest allows calculation of the MTF at various resolutions. The MTF at 50% contrast is considered most important for image quality.The effects of diffraction would be more marked with the D3x or other higher resolution camera. At larger apertures, the lens out resolves the sensor with the D3.





And for f/32. The blur circle is 109.5 mm, corresponding to a resolution of 41 cy/mm or 0.42 cy/pixel. Note that the aliasing is completely eliminated.
Also note that Imatest shows the diffraction limit for f/32, which is closely approximated by the curve.





It would be most interesting if you would shoot Bart's chart at optimal aperture and the minimum aperture and post the results.

Regards,

Bill

[BJL]

...  f/22 ... The image is resolving features 12 micros or a less on the image plane.

This fits a pattern I have noticed in various field observations: the diffraction effects coming in on a length scale in microns of about half the f-stop. Fitting roughly with the wavelength of visible light being about half a micron. This seems to fit fairly well with the theoretical figures based on 50% MTF, or am I remembering that incorrectly?

Of course the effect is incremental, not a brick wall, and some more fastidious observers draw the line one stop earlier (like f/8 with 5.6 micron pixel spacing, and some willing to go about a stop further, especially with high pixel count MF images expected to be printed at high PPI.

[Bart_van_der_Wolf]

Certainly. I used Bart's target with my Nikon D3 with the 60 mm f/2.8 AFS Micro-Nikkor. I rendered the images with ACR 6.5 with the default settings, which include a slight amount of capture sharpening.

Here is the result for f4.0 with Bart's method and with Imatest. The blur circle is 98 pixlels giving a resolution of 55 cy/mm or 0.47 cy/pixel, near the Nyquist limit of 0.5 cy/px. Bart's chart shows aliasing particularly well, but I think that the resolution results are for a relatively low MTF, perhaps near Rayleigh. Maybe Bart can comment if he is following this thread.

Hi Bill,

Yes, the limiting resolution (which approaches Nyquist) is at a relatively low MTF. It is closer to the human visual resolution limit, at approx. 10% response on the MTF, than to the 50% MTF metric (which gives a more overall perception of sharpness and contrast, and is closer to the maximum constrast sensitivity of the human eye). The Rayleigh criterion is more relevant for separating 2 point light sources, such as stars, and works out to something like a 20% contrast at that limit for a diffraction pattern.

To place that in perspective, a subject with 10:1 contrast at the 10% MTF point will exhibit a 1% resulting contrast, and that's at the perception threshold of detection (without sharpening) for human vision. The unaltered test chart, when printed on glossy paper will have something like a 100:1 contrast, and therefore plenty to produce aliasing even at very low MTF responses.

And for f/32. The blur circle is 109.5 mm, corresponding to a resolution of 41 cy/mm or 0.42 cy/pixel. Note that the aliasing is completely eliminated. Also note that Imatest shows the diffraction limit for f/32, which is closely approximated by the curve.

Sharpening will raise the MTF curve somewhat above the diffraction limited MTF, as long as there is sufficient signal left for sharpening. Low subject contrast combined with diffraction will result in loss of resolution, unrestorable even for deconvolution sharpening.

Cheers,
Bart

[bjanes]

Yes, the limiting resolution (which approaches Nyquist) is at a relatively low MTF. It is closer to the human visual resolution limit, at approx. 10% response on the MTF, than to the 50% MTF metric (which gives a more overall perception of sharpness and contrast, and is closer to the maximum constrast sensitivity of the human eye). The Rayleigh criterion is more relevant for separating 2 point light sources, such as stars, and works out to something like a 20% contrast at that limit for a diffraction pattern.

To place that in perspective, a subject with 10:1 contrast at the 10% MTF point will exhibit a 1% resulting contrast, and that's at the perception threshold of detection (without sharpening) for human vision. The unaltered test chart, when printed on glossy paper will have something like a 100:1 contrast, and therefore plenty to produce aliasing even at very low MTF responses.

I repeated my experiment at f/22, but omitted capture sharpening. The resolution calculated from your target was 57 lp/mm and the MTF 10 calculated by Imatest was 59 lp/mm, confirming the approx. 10% figure you quoted. It is interesting to note that the MTF10 at f/22 is at Nyquist, which would imply that deconvolution sharpening could restore some of the lost detail imposed by shooting at this small aperture.





I am somewhat confused by your statement that MTF at Rayleigh is around 20%, since I had always thought that it was aroung 10% as stated by Roger Clark. However, on checking in Wikipedia, they do mention a figure of 20%. How does one resolve these conflicting values?

Regards,

Bill

[Bart_van_der_Wolf]

I repeated my experiment at f/22, but omitted capture sharpening. The resolution calculated from your target was 57 lp/mm and the MTF 10 calculated by Imatest was 59 lp/mm, confirming the approx. 10% figure you quoted. It is interesting to note that the MTF10 at f/22 is at Nyquist, which would imply that deconvolution sharpening could restore some of the lost detail imposed by shooting at this small aperture.

Hi Bill,

That's correct. Unfortunately we'll not be able to recover all detail all the way up to Nyquist in your example because the actual image contrast gets too low (in case of an adequate Low-pass filter). Image contrast multiplied by MTF response approaches zero. This is also where a low glare lens helps to maintain contrast in the microdetail. I expect the image contrast of a MF system without low-pass filter to reach Nyquist without much attenuation other than that from an area sample (and produce aliasing).

I am somewhat confused by your statement that MTF at Rayleigh is around 20%, since I had always thought that it was aroung 10% as stated by Roger Clark. However, on checking in Wikipedia, they do mention a figure of 20%. How does one resolve these conflicting values?

Frankly, I traditionally assumed Roger's 9% number to be correct, but when checking Wikipedia found a higher contrast being quoted. To verify, I made a simulation in ImageJ of two (f/16, 564 nm) diffraction patterns I already had, offset by the radius to the first zero, and I find something like a bit over 26% contrast (depending on sensel size (!) and alignment with the sensel grid). This simulation is without the effect of a low-pas filter.

Cheers,
Bart

P.S.  The sensel's aperture size (or microlens) will quickly reduce the above oversampled (1 micron) rendering of the neighboring diffraction patterns (Rayleigh criterion) into a complete blur with zero contrast. I'm preparing a demonstration of it, but it takes the computer a while to calculate the base images.

[Alan Klein]

When I shot 67 on film I normally used the f/8 markers for depth of field when shooting at f/16.

Eric:  Basically you're doing the same thing I described I do except you're using a two stop margin rather than my one stop.  I checked that link you provided on Hyper-focal.  It appears a method for calculating for digital cameras and lenses rather than for anything for my film RB67.  But thanks for the link.  It'll be useful with my DSLR (as soon as I get one!)

[ErikKaffehr]

Alain,

A few observations:

1) When I started shooting 67 I discovered that I had issues with sharpness, and realized that those were caused by the "optimistic" DoF scales. I didn't buy MF to get same sharpness as from 135, but to have better images.

2) I sort of revisited the issue of stopping down. The philosophy I have right now is essentially to focus on what is important and stop down as much as needed. But I try to avoid going to f/22. I found out that of my around 58000 images about 300 are shot at f/22, many of those are tests. But after reconsidering my images I'd say that stopping down to f/22 is more OK than I expected. It's still  a waste of resolution, but resolution may less important than what I (we) may believe.

3) Sharpening matters a lot. Cannot replace the real stuff but matters a lot.

I hope you are going to enjoy digital once you have it. For many of us digital is a revelation but for some it's just a steep learning curve.

Best regards
Erik

Eric:  Basically you're doing the same thing I described I do except you're using a two stop margin rather than my one stop.  I checked that link you provided on Hyper-focal.  It appears a method for calculating for digital cameras and lenses rather than for anything for my film RB67.  But thanks for the link.  It'll be useful with my DSLR (as soon as I get one!)