PDA

View Full Version : Beware the Diffraction Limit



First Light
11-24-2006, 11:16 AM
Greetings,

What is the diffraction limit? Answer: The smallest aperture (largest f-stop) that you should use if you want sharp in-focus pictures at full resolution. Using a smaller aperture (larger f-stop) will cause the image to gradually soften as the light diffracts through the tiny iris opening in the lens. The diffraction limit varies from one camera model to the next, depending on the format size and resolution of the image sensor (digital cameras) or frame size and grain density (film cameras). The diffraction limit creates a barrier which limits the maximum usable resolution of a digital camera at small apertures because a higher resolution results in a lower diffraction limit. Eventually, the benefits of more megapixels at small apertures is negated by the blur created by diffraction. The diffraction limit can usually be ignored when your pictures will be output at a modest size and/or modest resolution.

Here are two samples which illustrate the effects of diffraction. Both shots were taken with an E-1 whose diffraction limit is f16 and a Zuiko Digital 50mm f2.0 macro lens with excellent sharpness. A stable tripod was used along with a remote shutter release and mirror lockup. The pictures were taken within seconds of each other and were exposed the same. The shots were both saved as "raw" images and were "developed" identically with no sharpening or post processing. The first shot was taken at f4.0 for 1/30th second using ISO 100. The lens is sharpest at this aperture setting. The second shot was taken at f22 for 1.0 second using ISO 100 and shows obvious diffraction blur.


3176
Click on the image to view it at full size.

These images are 100% crops of a larger image. The crops were taken near the center of the Field of View. They were converted to jpegs for display here and the least possible compression was used (quality setting "12" in Photoshop CS2). Note: The reason the paper in the photo is not a brighter white is because it is very thin and is printed on both sides. The thinness of the paper allows print from its back side and the next facing page to show through.

What do we see? First, diffraction softened the second image. Second, softening also reduced the contrast of the image.

Note: The above description and illustration were excerpted from a manual that I'm writing for a computer program.

roger h
11-24-2006, 11:20 AM
Fascinating. Great practical stuff. Thanks, FL!

Roger

Bojan Volcansek
11-24-2006, 11:42 AM
Great Practical Example!

Thank you very much FL

Yours Bojan

Jamie
11-24-2006, 12:41 PM
Thanks FL, great example. I think I may do some experiments myself...

luuluudave
11-25-2006, 07:06 AM
hmm, so what's the sharpest aperature setting for the ZD 14-54mm? I thought it would be around f/8....

llpoolej
11-25-2006, 07:16 AM
I have found the 14-54 sweet spot to be F8. Mine is very soft wide open. I really think I have a marginal copy of that lens though. At F8, it is very nice, at F 4, it is pretty soft.

My 50mm is just SHARP. My 50-200 is sharp wide open, but does show a hair of vignetting wide open, worse on the E-500 than the E-1. the 11-22 is super sharp also. Makes the 14-54 look like a dog

mxs
11-25-2006, 12:07 PM
I think my 14-54 goes to Oly soon (still in warranty). I just cannot use it wide open, especially from f 11 - 20 mm. I was doubting myself for the longest time, but yesterday I shot some landscape and the shots looked ... well soft, very soft indeed.

I always have luck with such things as I knew there are some bad copies floating around.

Regards

Marek

p.s. Did you consider it sending it back to Oly?


FL,

sorry for the OT reply. Can you write how far were you from the subject? I would just love to see you do this for 14-54 so I have finally something to compare my crappy samples with somebody who knows what his left and right hand is doing.

Thanks and regards

Marek

shrinkpictures
11-25-2006, 05:02 PM
FL - is there anywhere there is a list of the diffraction limit for all of our beloved ZD glass and 4/3 Camera Combo's?

I would like to see a chart, and some simple tests that we could all do so we could evaluate for ourselves and get a better understanding of the limits of our equipment.

acme
11-25-2006, 06:38 PM
FL - is there anywhere there is a list of the diffraction limit for all of our beloved ZD glass and 4/3 Camera Combo's?

I would like to see a chart, and some simple tests that we could all do so we could evaluate for ourselves and get a better understanding of the limits of our equipment.

The narrower the aperture, the larger the circle of confusion even a perfect lens can project. A perfect lens would perform at its best wide open, but of course few lenses reach this ideal.

Your sensor is also not a perfect device, having some discrete sensel pitch. The "sweet spot" for current fourthirds sensors is f-8 (E-XXX) or f-11(E-1). Faster than this and the extra accuity is wasted (either directly, as through the anti-alias filter, or as a consequence of the limits of sensor resolution, or because aberrations in the lens become more apparent). Slower than this, and you very quickly will find fine detail has been mushed up by the fundamental effect of diffraction.

There is one more hazy aspect to this calculation, and that is the magical software between the bayer filter and resultant pixels in your image. It has to take samples from four adjacent sensels and interpolate this data into the final result making easy analysis (such as in the previous paragraph) problematic. Sharpness corresponding with the true resolution of the sensor is only approached through clever programming, so you may feel free to use a stop beyond what circles of confusion versus sensel pitch say you should.

First Light
11-25-2006, 08:03 PM
...FL, sorry for the OT reply. Can you write how far were you from the subject? I would just love to see you do this for 14-54 so I have finally something to compare my crappy samples with somebody who knows what his left and right hand is doing. ...
You're welcome. The left hand, right hand thing is funny---my wife would get a good laugh over it because I'm ambidextrous and get my left and right confused all the time.:D As for your question, I don't remember exactly how far I was from the book when I made the test shots. I think I was probably between 6 and 12 feet (1.8 - 3.7 m) away. As for measuring a ZD 14-54mm f2.8-3.5 zoom, I plan to measure my zooms some day but it's a lot more work than measuring a prime and my ZD 50mm f2.0 and 150mm f2.0 tests seemed to take "forever".


FL - is there anywhere there is a list of the diffraction limit for all of our beloved ZD glass and 4/3 Camera Combo's? ...
First, when we talk about the "diffraction limit" of a camera, we're usually talking only about the camera body. The diffraction limit is calculated from the format size (physical dimensions of the image area of the sensor) and resolution. It defines the point where further decreases in the aperture size will probably not contribute to an increase in depth of field because diffraction will begin to blur the entire image (both the in focus and out of focus parts).

Here are the diffraction limits of all the 4/3rds cameras. These limits are the same no matter which lens you use. The values are rounded up to the closest 1/3 f-stop.

E-1 diffraction limit = f16.
E-300 diffraction limit = f13.
E-500 diffraction limit = f13.
E-330 diffraction limit = f13.
E-400 diffraction limit = f11.
DMC-L1 diffraction limit = f13.As you can see, the higher the megapixels, the more diffraction will be a problem. But there is a huge qualifier to this: the limit applies when you are using all of your pixels. If your output doesn't uses all of the pixels, then the effective diffraction limit will be pushed farther away. For example, if your typical output is a 4 x 6 inch paper print, you won't be using the full resolution of your image and you won't notice the problem until it becomes severe. The diffraction limit is important when you:

Crop or enlarge a lot.
Output large posters, maps or fine art prints and people view them at relatively close distances.
Use your images in high-resolution media that pan and zoom in on your images.
View your images at 100% on a video display.When we talk about lenses, the discussion must grow to encompass lens aberrations because they also cause the image to blur. The sharpness of most lenses is limited at the wide-aperture end by aberrations and limited at the small-aperture end by diffraction.

The best thing to do is to measure the sharpness of a lens at many different f-stops to determine how it behaves. Here are the measurements that I made of my ZD 50mm f2.0 and 150mm f2.0. Both were tested on my E-1. You'll get higher readings on a camera with more megapixels but the shape of the curves should be similar. What you see in the ZD 150mm f2.0 chart is very rare: few lenses hold their maximum sharpness with a wide open aperture. It is an impressive lens!


3188


3189
Click on a graph to see it full size.

As you can see above, the sharpness of both lenses begins to soften long before the diffraction limit of f16. Why? Because the diffraction limit isn't the point where diffraction begins to soften the image. Rather, it is the point where diffraction becomes great enough to be visible---or so the theory goes. To be accurate, diffraction is aways present. But isn't strong enough to have a visible effect until the aperture is very small.

ziggy
11-25-2006, 10:50 PM
I was reading on another fora, (a local Australian one) and a shot with a nikor 105 macro took my interest. The shot said f45 , but in checking the lens specifications , the max for this lens is f32.

I didnt quite get this so I went looking and found the reason was in the lens design, and at close range the circle of confusion is differently calculated , using symetrical vs assymetrical design. I have read the wiki article on this and still it doesnt make a lot of sense.

My question is probably more along the lines of is the apparent fstop similar in zuikos ? say at f16 - does the (apparent?)aperture decrease(f number increase) the closer one focuses? And does this have an positive or negative effect?

Thanks
Bruce

acme
11-25-2006, 11:02 PM
Here's a sequence of images I took to hammer out the issue. They're all 100% crops with plenty of pixel sized detail, and run through f-5.6, f-8, f-11, f-16, f-22, and f-32.

To "blink" compare multiple images, open them in separate tabs/windows in your browser and then flip back and forth between them.

The f-5.6 image is actually the most interesting because it is noticably sharper than the image at f-8, despite out resolving the sensor. Look at the bottom of the three white dust motes toward the centre of the image.

You can also see that some combination of limited DOF and aliasing is causing artifacts resembling posterization in the fly's eye. Diffraction has nullified its effect by f-16. Without the quite strong anti-aliasing filter Olympus uses, these artifacts would have been severe.

Between f-8 and f-11, some fine detail that was present in the front eye has been lost, while the additional DOF has revealed coarser detail in the rear eye. Beyond f-16 things are just getting more and more mushy. No surprise at all that 4/3 lenses have dropped f-32.

One final reinforcement about what FirstLight was talking about. Assuming light is no object, deciding whether to trade DOF for detail depends on what resolution you intend to reproduce the images at. If you look at the thumbnails, probably the photo taken at f-22 was the best match. In the 100% versions, probably it's the one at f-11.

First Light
11-25-2006, 11:29 PM
Hi Bruce,

Yes, the f-stop value does change with some lenses but the description puts the emphasis on the wrong thing. Here's my understanding:

Some lenses (mostly zooms, I think) adjust the focal length in order to focus the lens when the focus distance is less than infinity. For example, when you zoom the ZD 14-54mm lens all the way to a 54 mm focal length, you only get a true 54 mm focal length when the lens is focused at or near infinity. As you adjust the focus closer, the focal length gradually shrinks. I don't know what it really is at close distances, but I've noticed the change in focal length when I was testing focus screens many months ago.

Remember that the aperture is the diameter of the iris opening in the lens. Yet we don't set the aperture of our lenses by selecting its diameter. What we set is its "f-number" or "f-stop". Understanding what the f-number is will explain why the f-number changes with focus distance for some lenses and yet the aperture, itself, (its diameter) does not change. Here's the definition:


f-number = focal length / aperture diameter

Do you see what is going on now? The only reason the f-number would change when you change the focus distance of a lens is because the lens is adjusting the focal length---NOT the aperture diameter. And, since the focal length is used to calculate the f-number, it gives the false impression that the aperture diameter changes when it really doesn't.

The bottom line is this: The f-number is really a ratio. It is the ratio of the lens focal length to aperture diameter. If either the focal length or aperture diameter changes, so will the f-number. We have to dig deeper to find out which.

acme
11-25-2006, 11:37 PM
Some lenses (mostly zooms, I think) adjust the focal length in order to focus the lens when the focus distance is less than infinity.

It's also a common trick for macros that reach 1:1. You need a great long helicoid to do it at a longish focal length (say, 100mm), but not so bad if you've dropped down to 60mm by the time you hit 1:1. The same lens to imager distance effects a higher magnification at a shorter focal length.

ziggy
11-26-2006, 03:58 AM
Wow , thank you both, a clear and consise answer

mxs
11-26-2006, 06:27 AM
Hi FL,

and thanks for your reply. I think I finally understand this topic a little better. Interesting how diffraction kicks in much sooner than I thought, we just don't see it most of the time unless we use the "right" aperture.

Now back to my 14-54 tests for a second. Reading your shooting distance and seeing the crops posted makes me thing that either 14-54 doesn't measure up to DZ 50 at all or my DZ 50 is a dog even at f 50mm wide open. Or my newspaper print wasn't that good ... :-)


Regards

Marek

p.s. Congrats on the hands .... :-)

mxs
11-26-2006, 06:34 AM
I guess this is one of the Foveon advantages. No Bayer interpolation messing around. I am not sure whether they use AA filter or not though.

Regards

Marek

TrapperJohn
11-26-2006, 10:45 AM
What if one were to use a neutral density filter to stop down, instead of the aperture? Since that does not reduce the iris, would it also not forestall diffraction limits? Or am I reading this wrong?

acme
11-26-2006, 12:53 PM
What if one were to use a neutral density filter to stop down, instead of the aperture? Since that does not reduce the iris, would it also not forestall diffraction limits? Or am I reading this wrong?

You are completely correct. A high quality ND filter will allow you to maintain the resolution of a wide aperture while the light reaching the sensor is the same as if you'd stopped down.

If you think about it, this is how solar filters on telescopes work too. They maintain the resolution of the big intrument, while cutting down the light to something equivalent to a pinhole.

Bill_Turner
11-26-2006, 03:03 PM
Arguably one of the best threads I've had the pleasure of following.
I'd suggest a "sticky" in the Keepers area.
Thanks for your work and excellent, unbiased information.

TrapperJohn
11-27-2006, 07:45 AM
Interesting... perhaps somewhere down the road an electronic iris can be developed - possibly LCD based - that reduces light without reducing aperture.

This does explain why I was seeing less than perfect detail in some flower shots I did a couple of months ago. I was using the macro flash with the normally very sharp 50 Macro at around F18 to really black out the background. May have to chase down a couple of ND filters and go back and try that one again.




You are completely correct. A high quality ND filter will allow you to maintain the resolution of a wide aperture while the light reaching the sensor is the same as if you'd stopped down.

If you think about it, this is how solar filters on telescopes work too. They maintain the resolution of the big intrument, while cutting down the light to something equivalent to a pinhole.

RFMan
11-28-2006, 08:51 AM
Interesting... perhaps somewhere down the road an electronic iris can be developed - possibly LCD based - that reduces light without reducing aperture.

The technology to do this already exists actually. There are electrostatic films that can be built into a glass assembly that controls its opacity depending on voltage applied. I've seen it used in some very high end automobiles to provide "tint on demand" windows.

Cost is probably still rather high, but it seems that high cost is rarely a barrier for technology in the photo world. My guess is that the films haven't reached the optical quility necessary for photographic use.

It should be noted that this would be the equivalent of a variable ND filter. DOF would remain the same through the range, because aperture needs to be physically reduced to change DOF.

Granted, an electronic iris could be made that mimics the actual blades by being opaque on the edge and clear in the centre. Perhaps if the transition from opaque to transparent is graduated instad of the hard edge of today's lenses, diffraction effects would be minimized.

The more I type about this, the more I think it could be feasible...

Luc

TrapperJohn
11-28-2006, 04:43 PM
Ah, I had forgotten about the relationship between physical aperture and DOF. Easy to think of it as just light coming in.

No reason an LCD couldn't be designed to close down a circle, up to the point of diffraction limits, and then just darken the remaining circle gradually. The LCD could tailor the aperture to the size of the sensor much more precisely than blades. Wonder if that would produce any image improvement?

acme
11-28-2006, 07:15 PM
The LCD could tailor the aperture to the size of the sensor much more precisely than blades. Wonder if that would produce any image improvement?


Depends on the situation. Sometimes it's desirable to trade sharpness for more DOF.

If you want to experiment with variable ND filter, you can cross two polarizers. The outer one at least must be the linear type. They'll cool things substantially, possibly pass a lot of IR, and won't be as good optically as a quality ND filter, but if they happened to be in your bag when you needed one...

Come to think of it, this is just how an LCD works.

jimcb
11-29-2006, 04:59 AM
It should be noted that this would be the equivalent of a variable ND filter. DOF would remain the same through the range, because aperture needs to be physically reduced to change DOF.

You can have the equivalent of a variable ND filter by using two linear polarizing filters which you can rotate independently from one another.

I remember an airliner in the 1970's using this technique for windows: you could adjust the darkness by turning a knob, which caused one filter to rotate.

Cheers!