When we explained how the size of a digital camera's sensor affects its performance, there was one detail we left out: smaller models are more susceptible to suffering from diffraction. With the latest 14 Megapixel cameras around the corner, it's time to take a look at this little-known optical phenomenon
The Basics
Diffraction is a physical phenomenon that can affect photography, causing discs of light, rather than precise spots, to fall on your digital camera's sensor. It increases as you reduce aperture (by increasing the 'f-number' shown on a dial on your digital camera or on the screen), and is more noticeable at f/8 than at f/4.
The extent to which it becomes visible in the final photo depends on both how open the lens is and the size of the pixels on your camera's sensor. The larger the individual pixels, the less able they are to detect these spots of light.
The most recent cameras, whose sensors are crammed with pixels, are particularly susceptible to this effect.
How It Works
Diffraction is a simple physical principle: when light waves travel through a small hole--like the lens of your camera--they spread out. If you've ever watched waves rolling into a bay, you'll have seen the same effect: although the wind pushes the waves towards the shore in a straight line, once they roll past the pier, they spread out and the whole of the harbour is choppy, and not just the space in front of the entrance.
We hope any scientists who are reading will forgive us for this gross oversimplification, but when light enters the lens of your camera (in blue here), it flares out in a cone shape. As a result, instead of producing small, precise dots on the sensor (in green), it ends up producing larger round discs that are roughly circular. The exact shape depends on the form of the diaphragm in front of the lens, which is why using more blades in the iris is a distinct advantage.
In our case, the diameter of the disc produced by the diffracted light depends on the aperture of the lens, or its 'f-number', f/2.8, f/5.6, etc. The smaller the aperture (that is, the higher the f-number), the wider the area into which the light spreads out. The colour of the light also has an effect, with red light spreading out further than blue.
So, when you use a narrower opening, the photo ends up more blurry, with rays of light reflected from every object in the scene spreading out into a larger disc. This means that diffraction effectively limits the maximum resolution of a photo, imposing a limit on the smallest possible detail that can appear in a photo, usually measured in dots per inch or lines per millimetre.
Sensors
But the lens isn't the only part of the camera which plays a role here. Diffraction isn't a problem if the gap between two pixels on the sensor is larger than the smallest diffracted disc that the lens can produce. The maximum size is calculated according to various factors, and is around 1.8 µm for a lens at f/2.8, 5 µm at f/8, 10 µm at f/16 and so on. We've used green light to do these calculations.
Let's take a look at what happens when light is diffracted into a disc that measures 5 µm across, corresponding to a lens at f/8. It hits just a single large 8.4 µm pixel on the Nikon D700 on the left, producing a sharp result, but just begins to overlap on the EOS 7D (4.5 µm pixels) in the middle and completely covers several 1.4 µm pixels on the recent point-and-shoot on the right. In this last case, it doesn't matter how many pixels you have: they're just too small to be able to pick out enough detail and the photo will end up looking blurry.
In Practice
Here are some test shots from our lab, taken with two different cameras at five different apertures:
These are some full-size crops from photos we took using the 6 Megapixel Canon EOS 300D (left) and the EOS 550D (right), which has an 18 Megapixel sensor. In both case, the lens is a new Canon EF-S 18-135 mm, and we saved the photos as RAW files to make sure the results were comparable.
The pixels on the EOS 300D are 7.4 µm across, and as you can see, there is as much detail at f/16 as at f/11. At f/22, the results become less sharp. The pixels on the EOS 550D measure 4.3 µm, and at f/16, the loss of detail is definitely visible, especially on the connectors of the individual components. The loss of detail at f/22 is incredible.
Still, we have to be careful about drawing any hasty conclusions. Look at the silver connectors of the long black chip on the left. With the 6 Megapixels on the EOS 300D, they're never actually clear, but when you have 18 Megapixels, you can still see them at f/16, even if diffraction has blurred them a little.
In more realistic conditions, like looking at photos online or making an A4 print from which we've taken an extract on the right, the difference is limited. On the left, the EOS 300D does produce a slightly sharper photo, but the extra pixels that the EOS 550D has on the right help it avoid falling into other traps like the moiré that makes the connectors look blue and yellow on the left, rather than grey.
Compact Cameras
Strictly speaking, if a lens is open to f/2.8, light be will be diffracted to a diameter of 2 µm. A 1/2.3'' sensor measures exactly 4.62 x 6.16 mm, so its surface can only contain eight million pixels capable of capturing this 'smallest visible detail.' Going above 8 Megapixels won't change anything! And remember, plenty of compacts can't even manage f/2.8, and lots don't go anything further than f/5 when you start zooming in ...
However, as we saw with the EOS 550D, having individual pixels that are smaller than the spot of light diffused from the lens does actually allow you to cram in a few extra details. Nevertheless, at f/22, where the spot measured 14 µm across, the loss of detail is visible even on an A4 print if the pixels measure 4.3 µm. Using the same logic, if you have a compact camera with 1.5 µm pixels, as is the case for 12 Megapixel 1/2.3'' sensors, the limit for an A4 print would be f/5.6, and you'd be able to spot the missing details on a full-size image on anything from f/4 and wider.
These limits are a long way off for digital SLRs. The addition of extra pixels has some disadvantages (especially when it comes to the amount of memory needed to store your photos), but with a lens at f/8, increasing the resolution does allow you to capture details that would otherwise remain invisible.
On a point-and-shoot camera, though, it's often the case that diffraction limits the amount of detail that can be captured even with the lens as wide open as possible (the lowest available f-number). Worse still, the lenses found in many of these cameras can't deliver a perfect image to the sensor anyway, to the blur introduced by diffraction joins their other optical failures.
So, what's the point of having 14 Megapixels on a compact camera? Your photos won't have any more details, but all of the problems of having an over-crowded sensor--poor sensitivity and a narrow dynamic range--will be there, and your files will be larger as well …
Diffraction is a physical phenomenon that can affect photography, causing discs of light, rather than precise spots, to fall on your digital camera's sensor. It increases as you reduce aperture (by increasing the 'f-number' shown on a dial on your digital camera or on the screen), and is more noticeable at f/8 than at f/4.
The extent to which it becomes visible in the final photo depends on both how open the lens is and the size of the pixels on your camera's sensor. The larger the individual pixels, the less able they are to detect these spots of light.
The most recent cameras, whose sensors are crammed with pixels, are particularly susceptible to this effect.
How It Works
Diffraction is a simple physical principle: when light waves travel through a small hole--like the lens of your camera--they spread out. If you've ever watched waves rolling into a bay, you'll have seen the same effect: although the wind pushes the waves towards the shore in a straight line, once they roll past the pier, they spread out and the whole of the harbour is choppy, and not just the space in front of the entrance.
We hope any scientists who are reading will forgive us for this gross oversimplification, but when light enters the lens of your camera (in blue here), it flares out in a cone shape. As a result, instead of producing small, precise dots on the sensor (in green), it ends up producing larger round discs that are roughly circular. The exact shape depends on the form of the diaphragm in front of the lens, which is why using more blades in the iris is a distinct advantage.
In our case, the diameter of the disc produced by the diffracted light depends on the aperture of the lens, or its 'f-number', f/2.8, f/5.6, etc. The smaller the aperture (that is, the higher the f-number), the wider the area into which the light spreads out. The colour of the light also has an effect, with red light spreading out further than blue.
So, when you use a narrower opening, the photo ends up more blurry, with rays of light reflected from every object in the scene spreading out into a larger disc. This means that diffraction effectively limits the maximum resolution of a photo, imposing a limit on the smallest possible detail that can appear in a photo, usually measured in dots per inch or lines per millimetre.
Sensors
But the lens isn't the only part of the camera which plays a role here. Diffraction isn't a problem if the gap between two pixels on the sensor is larger than the smallest diffracted disc that the lens can produce. The maximum size is calculated according to various factors, and is around 1.8 µm for a lens at f/2.8, 5 µm at f/8, 10 µm at f/16 and so on. We've used green light to do these calculations.
 |
 |
 |
| 12 Megapixels Full Frame |
18 Megapixels APS |
14 Megapixels 1/2.3'' |
Let's take a look at what happens when light is diffracted into a disc that measures 5 µm across, corresponding to a lens at f/8. It hits just a single large 8.4 µm pixel on the Nikon D700 on the left, producing a sharp result, but just begins to overlap on the EOS 7D (4.5 µm pixels) in the middle and completely covers several 1.4 µm pixels on the recent point-and-shoot on the right. In this last case, it doesn't matter how many pixels you have: they're just too small to be able to pick out enough detail and the photo will end up looking blurry.
In Practice
Here are some test shots from our lab, taken with two different cameras at five different apertures:
These are some full-size crops from photos we took using the 6 Megapixel Canon EOS 300D (left) and the EOS 550D (right), which has an 18 Megapixel sensor. In both case, the lens is a new Canon EF-S 18-135 mm, and we saved the photos as RAW files to make sure the results were comparable.
The pixels on the EOS 300D are 7.4 µm across, and as you can see, there is as much detail at f/16 as at f/11. At f/22, the results become less sharp. The pixels on the EOS 550D measure 4.3 µm, and at f/16, the loss of detail is definitely visible, especially on the connectors of the individual components. The loss of detail at f/22 is incredible.
Still, we have to be careful about drawing any hasty conclusions. Look at the silver connectors of the long black chip on the left. With the 6 Megapixels on the EOS 300D, they're never actually clear, but when you have 18 Megapixels, you can still see them at f/16, even if diffraction has blurred them a little.
In more realistic conditions, like looking at photos online or making an A4 print from which we've taken an extract on the right, the difference is limited. On the left, the EOS 300D does produce a slightly sharper photo, but the extra pixels that the EOS 550D has on the right help it avoid falling into other traps like the moiré that makes the connectors look blue and yellow on the left, rather than grey.Compact Cameras
 On the Panasonic LX3, whose pixels are 2 µm apart, you can spot diffraction as early as f/5.6 ...
|
However, as we saw with the EOS 550D, having individual pixels that are smaller than the spot of light diffused from the lens does actually allow you to cram in a few extra details. Nevertheless, at f/22, where the spot measured 14 µm across, the loss of detail is visible even on an A4 print if the pixels measure 4.3 µm. Using the same logic, if you have a compact camera with 1.5 µm pixels, as is the case for 12 Megapixel 1/2.3'' sensors, the limit for an A4 print would be f/5.6, and you'd be able to spot the missing details on a full-size image on anything from f/4 and wider.
These limits are a long way off for digital SLRs. The addition of extra pixels has some disadvantages (especially when it comes to the amount of memory needed to store your photos), but with a lens at f/8, increasing the resolution does allow you to capture details that would otherwise remain invisible.
On a point-and-shoot camera, though, it's often the case that diffraction limits the amount of detail that can be captured even with the lens as wide open as possible (the lowest available f-number). Worse still, the lenses found in many of these cameras can't deliver a perfect image to the sensor anyway, to the blur introduced by diffraction joins their other optical failures.
So, what's the point of having 14 Megapixels on a compact camera? Your photos won't have any more details, but all of the problems of having an over-crowded sensor--poor sensitivity and a narrow dynamic range--will be there, and your files will be larger as well …
DigitalVersus on...