Doug Franklin wrote: > Joseph Tainter wrote: >> http://www.digital.pentax.co.jp/ja/35mm/k10d/image/k10d_09.gif >> >> The left graph is the RAW analogue signal from CCD. Middle is the prevalent >> 12 bit conversion and the right of course is the 22 bit conversion. >> >> Ken >> >> ----- >> >> Thanks, Ken. But, er, am I the only one who doesn't understand what >> these charts are showing? There are no captions (not even In Japanese, I >> think), no labels for the axes, no labels for what the red lines and >> dots mean. > > The red line in the left chart is a hypothetical actual, continuous, > analog signal. The middle chart represents analog -> digital conversion > with a 12-bit output. The right chart represents analog -> digital > conversion with a 22-bit output. > > In the middle and right charts: > > a) the same analog signal is represented as the black line > > b) the horizontal axis indicates the area of each digital pixel > > c) the vertical axis represents the color values that the analog -> > digital conversion can return, black at the bottom and white at the top > > d) the grid lines for the vertical axis on the right are closer together > than the ones in the middle chart to show the finer gradations of output > color possible with the
Oops. That should be "possible with the 22-bit analog -> digital conversion". > e) the red dots show the output values for at each pixel under either > 12-bit or 22-bit digital output The important thing to notice here is that those nice arcs on the right chart are horizontal lines on the middle chart. This represents the difference in "quantization error". When converting an analog signal to a digital one, quantization error is always present. Quantization error describes the difference between the actual analog values and the values that the values that the digital system can represent. On the middle and right charts, the vertical distance between the black line and the red dot is the quantization error for that pixel. The wider the horizontal lines on a graph like this, the more quantization error. Since the 22-bit system can represent 1,024 times as many different values as a 12-bit, each 22-bit value is closer to the actual value than the 12-bit values are, and the 22-bit digital representation of the signal has less quantization error, thus higher fidelity with the original analog signal. -- Thanks, DougF (KG4LMZ) -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net
