The White Paper

Chapter 15 Diffraction and How to Avoid It

What Is Diffraction?

Diffraction is about waves and how they behave. It applies not just to cameras but to all manner of waves, including light, radio, sound, even water. Diffraction is about how waves propagate - how they move through space from their source to their destination, and what happens after they pass through an aperture (such as the iris of your camera) or pass by an object.

We encounter many examples of diffraction in daily life without giving it a thought. For example, you're hiding behind a tree. Someone is calling to you. You hear them, even though the tree is directly between you and the source of the sound. Yet the sound waves manage somehow to reach your ears. How? Diffraction.

The bright "circle" around the moon you see sometimes at night. That's diffraction.

Punch a hole in a barrel of water. How does the water come out? In a perfect cylinder the same size as the hole? Or does it spread out after passing through the opening? The smaller the hole, the wider the water spreads. That's diffraction.

Ever make or use a "pinhole" camera? Poke a tiny hole in one end of a box, point it at some subject or scene. An image that's hundreds of times the size of the pinhole fills the opposite side of the box.

That's precisely what happens when light passes through your camera's aperture. The smaller the opening (the higher the f/stop), the more the light bends and spreads after passing through.

Why Is Diffraction A Problem?

There's a bunch of problems associated with diffraction. Paramount among them are: loss of sharpness, muddiness and the appearance of interference patterns (moiré). In other words, diffraction can have a strong negative impact on image quality.

Check out the illustration below. The light comes through a large aperture (f/2.8, f/3.5, f/4) fairly well. There is some spread and bending, but not enough to cause any serious image quality issues.

Figure 15-1 Light Coming Through Large Aperture

Now let's take a look at what happens to the light when it comes through the smallest aperture you can select on the H-Series cameras: f/8.

Figure 1-2 Light Coming Through Small Aperture

There are several things to note in the above illustration - none of them beneficial. The light coming through the tiny aperture gets dramatically bent and spread as a result of diffraction.When the light cones spread like they do above, the footprint at the sensor is much larger than it would be without diffraction.

Footprint, angle and spread. What impact do they have on your images?

1) Luminance Issues When cones of light overlap, their values may be additive. If that happens, you may end up with bright light where there should be only moderate light or shadows interrupted by stripes or puddles of light that shouldn't be there. Dark lines can be noticeably larger than they should be and edges may look coarse and heavy.

2) Cancellation Two waves that are out-of-synch can cause cancellation, wherein the information in one cone of light completely neutralizes the information in the other. The net result would be shadows where you don't expect to see them (note the dark areas in the illustration above).

3) Interference Two cones of light can combine information when they overlap, resulting in interference patterns as illustrated in fig. 2. Note the strange diagonal and curved patterns in the box in the lower right hand corner. That's caused by lines overlapping at different angles of incidence to create what is called moiré (named after the fabric whose shiny weave shimmers in constantly-changing patterns as the fabric moves).

Aside from the unsightly impact on the content of the image (strange effects in repeating patterns, such as windows in a building or bricks in a wall), moiré has a secondary effect. It fools the eye into seeing iridescent colors where there are none. As a result, you can get serious color distortion.

4) Convergence One other side-effect of interference is "muddiness", wherein the detail from separate cones of light are so close together that it becomes impossible to distinguish their edges. You can see this effect when you look at a lens resolution test. The closer the lines get to one another, the more difficult it is to distinguish individual lines. First, you get a moiré effect, then you get mud (amorphous gray region), then the lines converge into a solid. There is simply too many details, too close together for the eye to separate one from another. The visual effect looks like blurry patches of indeterminate color.

Figure 15-3 Convergence: Moiré and Mud

5) Loss of Resolution This is the most serious issue, and often is referred to as "softness". It's not really softness, it's lack of detail. Some blur can be fixed. Lack of detail cannot. It is missing information, information that cannot be recovered.

Diffraction's toll on resolution is caused by the size of the light cones hitting the photo sites on your sensor.

Ideally, each cone of light should illuminate a single pixel. That is, however, an ideal, not a realistic expectation. Most zoom lenses are just not that good at all focal lengths, there are filters in front of your sensors, microlenses, Bayer interpolation and a host of other potentially damaging processes between the light and the memory card.

When the cone of light covers more than one pixel, you lose resolution.

Figure 15-4 Lost Resolution From Oversize Light Cone

Consider the illustration above. On the left, one cone of light feeds its information to one single pixel. On the right, with the cone spread wide (by diffraction), each cone of light exposes two pixels. To make matters worse, they overlap on the middle pixel!

In this case, your output image would have the normal number of pixels, but the input, the image hitting the sensor, would have only 1/4 of the information it should have.

If the diffraction is bad enough, you could end up with an 8 mp image that has the detail of a 2 mp image! Not good.

How Serious Is Diffraction Effect in the H-Series Cameras?

Sony's H-Series cameras certainly suffer from some diffraction. It shows up mostly as a loss of sharpness in the final image. But it's far from crippling.

Just before the initial release of the H2 and H5, there were animated discussions online where "experts" claimed that these cameras would be an unmitigated disaster because their sensors were so small, and their photo sites so incredibly tiny that diffraction would ruin its images at all but the very lowest, widest aperture (f/2.8) The discussion was even more heated when the H7 and H9 came out with even more pixels squeezed into the same real estate as the earlier cameras.

Fortunately, they were wrong. Well, they were not wrong about having diffraction, that's just a mathematical formula and not in dispute. But wrong in terms of how bad the diffraction would be and how much impact it would have on image quality.

Figure 15-5 Portion of Lens Chart Test with DSC-H5

Figure 15-6 100% Crop From Test In Fig. 15-5

The tests above (fig. 5 and fig. 6) were shot with Sony's DSC-H5. The target used was printed out on photo paper using an ink jet printer, so some of the fuzziness and artifacts were a result of creating the chart. But you can clearly see the difference between f/8 and f/3.7 on any H-Series Camera.

The f/8 detail is a bit blockier and a bit fuzzier, which may lead to the perception of a certain amount of softness - a little less apparent detail. This is caused by diffraction. To be honest, it's not bad. I expected it to be considerably worse.

Fig. 6 above is a 100% crop (one pixel of the image equals 1 pixel on your monitor). If you were to print at that resolution, fig. 6 would be approximately 30" X 42" and viewed with your nose pressed up against the print. Here's a picture of the original full-size test sheet (8X10), just to provide a little perspective on what you're seeing in the preceding illustrations:

Fig 15-7 The Full Test Chart

Why were the so-called experts so wrong? Because they failed to take into account the multiplicity of factors that determine whether or not you get sharp pictures: the sharpness of the lens, the anti-aliasing filter on the sensor, the quality and design of the microlenses, the in-camera processing and the camera's depth of field at various focal lengths.

How Much Diffraction Is There In H-Series Cameras?

You can quantify diffraction relatively easily, but not the effect that it has on images. You have to do that with your eyes. And it's totally subjective. How much diffraction is bad? How much diffraction is too much? There's no such thing as "perfect focus" or "perfect sharpness". There's only "acceptable focus" and "acceptable sharpness". You are the only judge of what's "acceptable" to you.

Here's a comparison test I shot using the DSC-H5 and a brick wall. For a variety of reasons, brick walls have been the traditional test target for diffraction testing. It makes some sense, brick has lots of edges and lots of texture.

This series of images is displayed as a 100% crop, one photo pixel for each monitor pixel. All pictures are exactly as they came from the camera. No resizing, no sharpening, no post-processing.

Figure 15-8a 100% Crop at f/3.5

Figure 15-8b 100% Crop at f/5.6

Figure 15-8c 100% Crop at f/8

At 100%, you can easily see that the f/8 image shows noticeably less detail than the f/3.5 image.There's less texture in the brick, much less texture in the mortar. If you look carefully, you'll note that the color, contrast and saturation also seem to be diminishing as the aperture gets smaller.

Now, let's look at another crop of the same image. This time at a reasonable zoom. Nobody looks at 100% crop pictures. Nobody prints images at monitor resolution. Nobody stands nose-to-print evaluating photos.

The following images are resized to a more realistic viewing resolution - about 1/3 of the 100% crops in the previous figures. And they pose the only important question about diffraction: how sharp is sharp enough?

Figure 15-9 33% Crop Diffraction Comparison

As expected, the f/3.5 shot has the most detail and sharpness, each version getting a bit duller and softer down to f/8. But, at this more typical viewing resolution, the differences between the three f/stops is far less noticeable than it was in the full-resolution crops.

But what's that fourth example? It's the same shot as the one directly above it. It's an f/8 shot at maximum diffraction. Yet, it appears to be the crispest of all four shots and, to my eye, the richest.

How is that possible? Post-processing. I applied a little Smart Sharpening (Photoshop CS3) and just a touch of levels (darkened mid tones a tiny bit, increased highlights a tiny bit). The detail seems to pop right off the page. The other examples look flat by comparison.

Look at the mortar detail. Quite surprisingly, the tone adjustments I did seem to bring out more fine detail than any of the other images, even at their lower, less-diffracted f/stops.

What's the point? That there's more to the sharpness of a photo than the original acuity of the image produced by the camera.

Just to prove the point, here's another copy of the f/8 shot in fig. 8c. Remember how soft that image was? Well here's the same shot after 30 seconds of post-processing. I'm not adding anything that isn't there, just bringing out what is:

Figure 15-10 f/8 100% Crop, Post-Processed

Conclusion: How Sharp Is Sharp Enough?

For me, f/8 is far from the disaster predicted before the release of the H5. Do the H-Series cameras suffer some diffraction at the higher f/stops? Sure. Is it crippling? Hardly.

Sony did a number of things right with the H2 - H9. They used a perfectly-balanced anti-aliasing filter on the front of the sensor. It kills moiré without producing the usual softness typical of aggressive A-A filters. Moiré is simply not an issue in these cameras at any f/stop.

The tiny sensor and high-crop lens produce plenty of depth of field. They trade off some of the loss in acuity due to diffraction against better focus across a wide portion of the scene.

The beautiful color saturation and tone curves of the H-Series cameras add to the sense of three-dimensionality that makes their images "pop" so well.

How sharp is sharp enough? Sharp enough that a picture looks good hanging on the wall and viewed at a comfortable distance. Can you create an image like that with the H-Series cameras, even at the highest f/stops? Yes.

Does diffraction matter? Yes. Does it seriously limit these cameras? No.

My Recommendations

I generally like to take the sharpest images I can with the most detail possible. If I start with sharp, I can tone it down. You can't add detail.

So I generally avoid f/8. I limit myself to f/7.1, and shoot as much as possible from f/2.8 to f/4.

However, I'm not afraid to use the higher f/stops if I need them. If f/8 gives me the depth of field I need to get sharp focus over a huge range, I'll use it without a second thought. If I need f/8 to slow down water on a bright day, I'll use it.

I have no compunctions about post-processing my images to bring out the best of what's in the image, and you shouldn't either.

All cameras have limitations. DSLRs suffer from the same problem, but usually at f/16 to f/32. On the other hand, most DSLRs don't produce particularly sharp photos wide open. The H-Series cameras do.

On those occasions when sharpness is critical, use the lower apertures. For other images, use the f/stop that's best for the composition and the purpose of the shot. Remember, no one is going to look at your shots at 100% crop.



	Chapter 15 Diffraction and How to Avoid It

	What Is Diffraction? Diffraction is about waves and how they behave. It applies not just to cameras but to all manner of waves, including light, radio, sound, even water. Diffraction is about how waves propagate - how they move through space from their source to their destination, and what happens after they pass through an aperture (such as the iris of your camera) or pass by an object. We encounter many examples of diffraction in daily life without giving it a thought. For example, you're hiding behind a tree. Someone is calling to you. You hear them, even though the tree is directly between you and the source of the sound. Yet the sound waves manage somehow to reach your ears. How? Diffraction. The bright "circle" around the moon you see sometimes at night. That's diffraction. Punch a hole in a barrel of water. How does the water come out? In a perfect cylinder the same size as the hole? Or does it spread out after passing through the opening? The smaller the hole, the wider the water spreads. That's diffraction. Ever make or use a "pinhole" camera? Poke a tiny hole in one end of a box, point it at some subject or scene. An image that's hundreds of times the size of the pinhole fills the opposite side of the box. That's precisely what happens when light passes through your camera's aperture. The smaller the opening (the higher the f/stop), the more the light bends and spreads after passing through. Why Is Diffraction A Problem? There's a bunch of problems associated with diffraction. Paramount among them are: loss of sharpness, muddiness and the appearance of interference patterns (moiré). In other words, diffraction can have a strong negative impact on image quality. Check out the illustration below. The light comes through a large aperture (f/2.8, f/3.5, f/4) fairly well. There is some spread and bending, but not enough to cause any serious image quality issues. Figure 15-1 Light Coming Through Large Aperture Now let's take a look at what happens to the light when it comes through the smallest aperture you can select on the H-Series cameras: f/8. Figure 1-2 Light Coming Through Small Aperture There are several things to note in the above illustration - none of them beneficial. The light coming through the tiny aperture gets dramatically bent and spread as a result of diffraction.When the light cones spread like they do above, the footprint at the sensor is much larger than it would be without diffraction. Footprint, angle and spread. What impact do they have on your images? 1) *Luminance Issues* When cones of light overlap, their values may be additive. If that happens, you may end up with bright light where there should be only moderate light or shadows interrupted by stripes or puddles of light that shouldn't be there. Dark lines can be noticeably larger than they should be and edges may look coarse and heavy. 2) *Cancellation* Two waves that are out-of-synch can cause cancellation, wherein the information in one cone of light completely neutralizes the information in the other. The net result would be shadows where you don't expect to see them (note the dark areas in the illustration above). 3) *Interference* Two cones of light can combine information when they overlap, resulting in interference patterns as illustrated in fig. 2. Note the strange diagonal and curved patterns in the box in the lower right hand corner. That's caused by lines overlapping at different angles of incidence to create what is called moiré (named after the fabric whose shiny weave shimmers in constantly-changing patterns as the fabric moves). Aside from the unsightly impact on the content of the image (strange effects in repeating patterns, such as windows in a building or bricks in a wall), moiré has a secondary effect. It fools the eye into seeing iridescent colors where there are none. As a result, you can get serious color distortion. 4) Convergence One other side-effect of interference is "muddiness", wherein the detail from separate cones of light are so close together that it becomes impossible to distinguish their edges. You can see this effect when you look at a lens resolution test. The closer the lines get to one another, the more difficult it is to distinguish individual lines. First, you get a moiré effect, then you get mud (amorphous gray region), then the lines converge into a solid. There is simply too many details, too close together for the eye to separate one from another. The visual effect looks like blurry patches of indeterminate color. Figure 15-3 Convergence: Moiré and Mud 5) *Loss of Resolution* This is the most serious issue, and often is referred to as "softness". It's not really softness, it's lack of detail. Some blur can be fixed. Lack of detail cannot. It is missing information, information that cannot be recovered. Diffraction's toll on resolution is caused by the size of the light cones hitting the photo sites on your sensor. Ideally, each cone of light should illuminate a single pixel. That is, however, an ideal, not a realistic expectation. Most zoom lenses are just not that good at all focal lengths, there are filters in front of your sensors, microlenses, Bayer interpolation and a host of other potentially damaging processes between the light and the memory card. When the cone of light covers more than one pixel, you lose resolution. Figure 15-4 Lost Resolution From Oversize Light Cone Consider the illustration above. On the left, one cone of light feeds its information to one single pixel. On the right, with the cone spread wide (by diffraction), each cone of light exposes two pixels. To make matters worse, they overlap on the middle pixel! In this case, your output image would have the normal number of pixels, but the input, the image hitting the sensor, would have only 1/4 of the information it should have. If the diffraction is bad enough, you could end up with an 8 mp image that has the detail of a 2 mp image! Not good. How Serious Is Diffraction Effect in the H-Series Cameras? Sony's H-Series cameras certainly suffer from some diffraction. It shows up mostly as a loss of sharpness in the final image. But it's far from crippling. Just before the initial release of the H2 and H5, there were animated discussions online where "experts" claimed that these cameras would be an unmitigated disaster because their sensors were so small, and their photo sites so incredibly tiny that diffraction would ruin its images at all but the very lowest, widest aperture (f/2.8) The discussion was even more heated when the H7 and H9 came out with even more pixels squeezed into the same real estate as the earlier cameras. Fortunately, they were wrong. Well, they were not wrong about having diffraction, that's just a mathematical formula and not in dispute. But wrong in terms of how bad the diffraction would be and how much impact it would have on image quality. Figure 15-5 Portion of Lens Chart Test with DSC-H5 Figure 15-6 100% Crop From Test In Fig. 15-5 The tests above (fig. 5 and fig. 6) were shot with Sony's DSC-H5. The target used was printed out on photo paper using an ink jet printer, so some of the fuzziness and artifacts were a result of creating the chart. But you can clearly see the difference between f/8 and f/3.7 on any H-Series Camera. The f/8 detail is a bit blockier and a bit fuzzier, which may lead to the perception of a certain amount of softness - a little less apparent detail. This is caused by diffraction. To be honest, it's not bad. I expected it to be considerably worse. Fig. 6 above is a 100% crop (one pixel of the image equals 1 pixel on your monitor). If you were to print at that resolution, fig. 6 would be approximately 30" X 42" and viewed with your nose pressed up against the print. Here's a picture of the original full-size test sheet (8X10), just to provide a little perspective on what you're seeing in the preceding illustrations: Fig 15-7 The Full Test Chart Why were the so-called experts so wrong? Because they failed to take into account the multiplicity of factors that determine whether or not you get sharp pictures: the sharpness of the lens, the anti-aliasing filter on the sensor, the quality and design of the microlenses, the in-camera processing and the camera's depth of field at various focal lengths. How Much Diffraction Is There In H-Series Cameras? You can quantify diffraction relatively easily, but not the effect that it has on images. You have to do that with your eyes. And it's totally subjective. How much diffraction is bad? How much diffraction is too much? There's no such thing as "perfect focus" or "perfect sharpness". There's only "acceptable focus" and "acceptable sharpness". You are the only judge of what's "acceptable" to you. Here's a comparison test I shot using the DSC-H5 and a brick wall. For a variety of reasons, brick walls have been the traditional test target for diffraction testing. It makes some sense, brick has lots of edges and lots of texture. This series of images is displayed as a 100% crop, one photo pixel for each monitor pixel. All pictures are exactly as they came from the camera. No resizing, no sharpening, no post-processing. Figure 15-8a 100% Crop at f/3.5 Figure 15-8b 100% Crop at f/5.6 Figure 15-8c 100% Crop at f/8 At 100%, you can easily see that the f/8 image shows noticeably less detail than the f/3.5 image.There's less texture in the brick, much less texture in the mortar. If you look carefully, you'll note that the color, contrast and saturation also seem to be diminishing as the aperture gets smaller. Now, let's look at another crop of the same image. This time at a reasonable zoom. Nobody looks at 100% crop pictures. Nobody prints images at monitor resolution. Nobody stands nose-to-print evaluating photos. The following images are resized to a more realistic viewing resolution - about 1/3 of the 100% crops in the previous figures. And they pose the only important question about diffraction: how sharp is sharp enough? Figure 15-9 33% Crop Diffraction Comparison As expected, the f/3.5 shot has the most detail and sharpness, each version getting a bit duller and softer down to f/8. But, at this more typical viewing resolution, the differences between the three f/stops is far less noticeable than it was in the full-resolution crops. But what's that fourth example? It's the same shot as the one directly above it. It's an f/8 shot at maximum diffraction. Yet, it appears to be the crispest of all four shots and, to my eye, the richest. How is that possible? Post-processing. I applied a little Smart Sharpening (Photoshop CS3) and just a touch of levels (darkened mid tones a tiny bit, increased highlights a tiny bit). The detail seems to pop right off the page. The other examples look flat by comparison. Look at the mortar detail. Quite surprisingly, the tone adjustments I did seem to bring out more fine detail than any of the other images, even at their lower, less-diffracted f/stops. What's the point? That there's more to the sharpness of a photo than the original acuity of the image produced by the camera. Just to prove the point, here's another copy of the f/8 shot in fig. 8c. Remember how soft that image was? Well here's the same shot after 30 seconds of post-processing. I'm not adding anything that isn't there, just bringing out what is: Figure 15-10 f/8 100% Crop, Post-Processed Conclusion: How Sharp Is Sharp Enough? For me, f/8 is far from the disaster predicted before the release of the H5. Do the H-Series cameras suffer some diffraction at the higher f/stops? Sure. Is it crippling? Hardly. Sony did a number of things right with the H2 - H9. They used a perfectly-balanced anti-aliasing filter on the front of the sensor. It kills moiré without producing the usual softness typical of aggressive A-A filters. Moiré is simply not an issue in these cameras at any f/stop. The tiny sensor and high-crop lens produce plenty of depth of field. They trade off some of the loss in acuity due to diffraction against better focus across a wide portion of the scene. The beautiful color saturation and tone curves of the H-Series cameras add to the sense of three-dimensionality that makes their images "pop" so well. How sharp is sharp enough? Sharp enough that a picture looks good hanging on the wall and viewed at a comfortable distance. Can you create an image like that with the H-Series cameras, even at the highest f/stops? Yes. Does diffraction matter? Yes. Does it seriously limit these cameras? No. My Recommendations I generally like to take the sharpest images I can with the most detail possible. If I start with sharp, I can tone it down. You can't add detail. So I generally avoid f/8. I limit myself to f/7.1, and shoot as much as possible from f/2.8 to f/4. However, I'm not afraid to use the higher f/stops if I need them. If f/8 gives me the depth of field I need to get sharp focus over a huge range, I'll use it without a second thought. If I need f/8 to slow down water on a bright day, I'll use it. I have no compunctions about post-processing my images to bring out the best of what's in the image, and you shouldn't either. All cameras have limitations. DSLRs suffer from the same problem, but usually at f/16 to f/32. On the other hand, most DSLRs don't produce particularly sharp photos wide open. The H-Series cameras do. On those occasions when sharpness is critical, use the lower apertures. For other images, use the f/stop that's best for the composition and the purpose of the shot. Remember, no one is going to look at your shots at 100% crop.