Pixels vs. Pixels

There are a number of factors that ultimately determine which digital camera is best for you, or for a given use.  At the heart of every digital camera is the imaging sensor.  Trust me not all sensors are alike and not all pixels are created equally.

The two basic types of camera image sensors are either a CCD (charged-coupled device) or a CMOS (complementary metal oxide semiconductor) type.  Both have a grid of light gathering photo sites, incorrectly, but often referred to as pixels, to sort the light into Red, Green and Blue groups of digital information (see figure 4).

Until very recently, most digital cameras have relied on the use of the CCD sensors.  CCD manufacturing uses an obsolete process called N-MOS.  The process, requiring hundreds of steps over weeks, makes CCD sensors complex and expensive to manufacture, especially in large sizes.  Even though increased demand for CCD sensors has yielded some economies of scale, these have not been sufficient to lower production costs to easily affordable levels.

A CCD and a CMOS sensor may look the same, but their operational methods are very different (see diagram 1 and 2).  The CCD gathers the light during the exposure into a grid of Red, Green, Blue photodiodes.  The built up charge resulting from the exposure is sent to the register buffer within the chip and onto the voltage converter, then into the amplifier circuit, followed by the analog to digital converter and finally to the camera’s main processor for further color and sharpening work as needed.  The CCDs numerous steps demands high power consumption, greater battery consumption per picture.  The multiple steps listed also allows for the introduction of digital noise, which requires increased and various solutions to correct and reduce it.

The CMOS sensor has the same grid of photo sites, but each site has connected circuitry for converting light into digital electron signals within itself, before the signal is sent to the main camera processor.  This CMOS is able to boost the signal value at the site, thus reducing noise (static) in the image. The CMOS uses less power, has a clean digital signal, and is slightly less costly; therefore utilized mainly for larger sensor camera models.

The number of imaging sites (megapixels) refers to the resolution of information gathered with each exposure.  The greater the information gathered the sharper the photograph will be and the larger it may be printed.  CCD sensor chips used in most small digicams are 3.5x4.5 mm (approximately 9/64 x 11/64” – a quarter of your little finger nail).  The newer 8 mp point and shoot cameras now use the new 6.6x8.8 mm (approximately 17/64 x 11/32”) sensor.  The partial frame size sensor used in most DSLR cameras is 15.1x22.7 mm (19/32 x 57/64 “).  The very few full-frame sensors are the CMOS type and are used in cameras with a frame of 24x36 mm (15/16 x 1 and 27/64” with 10 micron photo sites).  [See figure 1 and 2 below]  When discussing image quality the digital image sensor size and the photo site size is just like film, in that the larger the piece of film, the greater the image quality.

The larger the image sensor chip, the larger the individual image sites may become, thus offering cleaner and better information.  The 15.1 x 22.7 mm image sensors, like those used on most 6 megapixel DSLR’s with a 1.5 (Nikon D70) or 1.6 X (Canon 10D) magnification factor have photo sites (pixels, if you must) of 6~7.5 microns each in size.  The new Canon Powershot Pro 1 has an 8 megapixel sensor approximately 6.6x8.8 mm in size which has photo sites (pixels) which are about 2.7 microns each.  Remember that a micron is mighty small, at one millionth of a meter.  To put things in perspective a human hair is about 8 microns in diameter.

Now we can discover how a Canon Pro 1, and all other manufacturers, can put 8 megapixels on a chip that is smaller than the Canon 10D’s (or others) 6 megapixel sensor.  The photo sites themselves are smaller, a whopping 2 and 1/2 times smaller.  SO – things really are not alike.

One big digital problem is a noise problem, something referred to as digital grain.  Noise is referred to in electronic terms as the Signal to Noise ratio.  The S/N is the ratio of the signal information to the empty signal of just electronic noise (static), which is always there.

A law of physics suggests that the smaller individual photo sites become the harder it is for the photons of light to reach it.  A 2.7 micron photo site is incredibly small, and this means that the limited number of photons of light that can reach the photo site can be so small that the noise overwhelms it, and we end up with unacceptably noisy images.  Where does this noise come from?  Part of it from the camera system's electronics, part from the random motion of atoms within the silicon and part from ambient cosmic rays.  So, the weaker light, at the photo site, produces a weak signal with a low (bad) S/N ratio, which creates a noisier image.  This is a challenge that manufacturers must face when designing a chip with 2.7 Micron, or smaller, pixel/photo sites.  Many manufacturers use a micro-lens technology to help with the light amplification to each pixel for increased signal strength (See diagram 3).

If you are still wondering why sensor size matters so very much, the total answer really is a complex issue.  Since the larger photo sites produce better image quality and lower noise levels, as in larger sensor cameras like the Canon 10D, they can operate at greater ISO equivalents.  The large sensors may operate at the amplified signal strength equivalent of ISO 3200, although the image may get a little noisy, while many smaller consumer digicams with small sensors cannot operate above ISO 400 because the S/N ratio, in tiny photo sites, becomes excessive.

A factor of image quality also occurring here is that small sensors tend to be of a different type than large sensors.  Small sensors used on all consumer digital cameras, use a scheme, which can read the data from the sensor in “real time.”  This provides the previews, and is called “interline transfer”.  Therefore the CCD electronics must gather image information and control the exposure time, rather than using a mechanical shutter.  Large sensors used on more expensive Digital SLRs are often of a different design style known as “full frame transfer”, which is not referring to their sensor size, but to their design.  The “full frame transfer” style of sensor does require the use of a mechanical shutter.  They do not read out and display the data in real time, but provides the viewable image only after the exposure.  That logically means they cannot give real time LCD previews or record video.  The advantage of this plan is that the sensitivity of the whole photo site can just be used to capture the light.  With the smaller “interline transfer” style CCDs utilizing part of each pixel to help store image charge and part to broadcast an image preview; therefore they compound their own problems.  The smaller sensor with small photo sites, generating a low S/N ratio, has now a significantly greater image problem than initially realized.  Now we see that the smaller “interline transfer” sensors, used in the small consumer digital cameras, yield lower resolution images than those used in higher end DSLRs.  On the good side, the lack of a mechanical shutter makes the cameras cheaper, simplifies their construction, and allows them to do more "tricks" like recording video clips and giving a live image display on their LCD screen.  There is always a trade off.  Nothing is as simple as it may appear.

A small size sensor also means that it only requires a short focal length lens for the image to cover the sensor.  For example, a typical consumer digicam may only need a 7mm lens to give the same field of view, as you would get using a 35mm focal length lens on a 35mm camera.  This has consequences on the photographs made and their depth of field.  It means that most consumer digicams have a large depth of field; say about from here to Milwaukee.  Great if you want everything in focus, not so great if you want a blurred background.  All shallow depth of field effects must be performed in an editing program, such as Adobe’s Photoshop.  Banking on the fact that a large number of small digicams are used for personal memory gathering, the large depth of field will be perfect for the application.  Wow, it seems as if there is a lot to be considered here.

Figure 1

Figure 2
Near actual sensor sizes
(should be 1 1/2 inch wide)
Internal Text Is The Same as Figure 1

Figure 3
View of Actual Sensors

 

Figure 4
The Bayer Pattern Pixel Grid on an Imaging Sensor

 

Diagram 1 - Image processing within a CCD image sensor

 

Diagram 2 - Image processing within a CMOS image sensor

Diagram 3 - Micro-lens technology

For additional information on the confusing subject of image resolution, see the article on Image Resolution vs. Printer Resolution.

I would like to thank my friends and contacts at Intel Corporation and Unisys Corporation for their technical expertise on semiconductor fabrication and their great contributions to this article.

Stop by pictureline and see for yourself that not all pixels are created equal, and we will help you find what is exactly right for your needs.

Article submitted by askRodger@pictureline.com

Copyright 2004 pictureline inc.

For permission to reproduce this article CLICK here.