Tuesday, October 13, 2009
In the race for higher resolution, there are trade-offs with image quality. It all comes down to the limits of the image sensors and the individual photosites on those sensors.
Sheer photosite size doesn’t necessarily mean better light-gathering ability, however. What’s important is the size of the light-gathering area. A photosite consists of the light-gathering “well,” plus the necessary electronic circuitry to operate it and move the data along the imaging pipeline. Manufacturers work hard to minimize the size of this circuitry so the light-gathering area can be larger for a given photosite size. From a practical standpoint, this means newer D-SLRs, with their greater “fill factors,” produce better image quality than earlier ones, even with the same, or slightly smaller, overall gross pixel size.
Cameras with sensors that have larger photosites tend to generate better color and higher dynamic range (because of a bigger, clearer source in relation to the background noise from the charged chip) and better performance at higher ISO. For example, the Nikon D700 and Nikon D3 rank #1 and #2 in DxO Labs’ high-ISO performance rankings, but #3 and #4 overall (www.dxomark.com). These full-frame (Nikon FX-format) cameras produce the best high-ISO performance, in part, because they have the biggest pixels among current models.
Heat is a major concern among camera manufacturers because heat on the sensor increases image noise. To mitigate the heat, manufacturers use several tactics. Ultimately, heat comes from power consumption. All other things being equal, a CMOS sensor will draw less power and generate less heat and noise than a CCD sensor. However, in practice, all things aren’t equal, and manufacturers work to reduce heat and image noise through several approaches. Choice of sensor technology is one approach, as are pixel spacing on the sensor and onboard image-processing technology.
Noise also affects bit depth, the number of tones from white to black (or number of shades of a color) a sensor can deliver. The camera’s A/D converter converts the analog data from the image sensor to digital data, which theoretically consists of 12-bit (4,096 possible tones) or 14-bit (16,384 tones) output per color channel (red, green and blue). But the noise level isn’t identical for each channel, so the actual bit depth of the image is less than the theoretical 12 or 14 bits. (And keep in mind that JPEG images are 8-bit, with just 256 tones.) Among today’s D-SLRs, those with the most megapixels tend to produce the highest actual bit depths; in DxO Labs’ RAW sensor tests, the 60.5-megapixel, medium-format Phase One P65+ produced the highest actual bit depth (in part because of its 16-bit A/D converter) followed by Nikon’s 24.5-megapixel D3X; the highest under-24-megapixel camera rated 10th on the list. (Again, it’s a balance between pixel count and sensor size; the highest-rated APS-C D-SLR ranked 14th.)
Phase One’s P+ Sensors