Monday, September 1, 2008
Megapixels: How Much Is Enough?
Pixel count certainly plays a key role in photography, but we’re approaching a point of diminishing returns and potentially reduced image quality as more pixels are packed onto a sensor
How small is small? The individual photodetectors in the Canon EOS 40D D-SLR are about 5.7 micrometers across. That’s 5.7 millionths of a meter (or 5.7 thousandths of a millimeter). By contrast, the width of a typical human hair is about 50 micrometers. For a sense of the downward trend in photodetector sizes, we can compare the EOS 40D to the EOS 10D that preceded it by just 41⁄2 years. The photodetectors on the EOS 10D were about 7.4 micrometers across, meaning in those 41⁄2 years they have decreased in size by about 23%. Considering the tiny scale we’re talking about, that’s significant.
The number of electrons that can be held by an individual photodetector (which is measured as an electrical voltage) determines the definition of “full” in the context of that photodetector (this is referred to as a measure of “full well”). The larger the photodetector, the more electrons it can fit, and thus the stronger electrical charge it can record. The minimum charge (ostensibly, no charge at all) compared with the maximum charge provides a measure of the dynamic range of each individual photodetector—the darkest and brightest values each photodetector can measure.
This is where we start running into problems with the math. If we need the size of each photodetector to decrease as the number of megapixels increase, but we need bigger photodetectors to increase dynamic range, doesn’t that mean with each jump in megapixels the dynamic range of the imaging sensor is decreasing? Yes, it probably is. Granted, there are other technological advances that mitigate this problem, but it’s a clear example of how conventional wisdom can be dead wrong. (“I need more megapixels to produce a better image.”)
The size of photodetectors also has an impact on the amount of noise present in a digital photo. Noise results from a variety of factors, but it represents errors in the measurement of pixel values. In effect, noise results when the camera has to guess at the value of a pixel it isn’t able to see very well. The best way to avoid noise is to have more signal (thus, the term “signal-to-noise ratio” when describing the amount of noise with electronics, in general). If you have more signal (or a higher-quality signal), you’ll have less noise. Well, just as the size of the photodetectors determines the dynamic range of the sensor, that size has an impact on the noise in the final image.
Page 2 of 4







