On 05/09/12 08:01, Godfrey DiGiorgi wrote: > On Tue, Sep 4, 2012 at 7:38 PM, Steven Desjardins<drd1...@gmail.com> wrote: >> You could do a better job with many aspects of this camera. Maybe low >> light is a sensor problem? > A standard Bayer mosaic sensor is a planar surface with a filter array > on top of it. This means that the light at each of the photosites for > all three colors is essentially equal, minus the amount that the > particular color filter absorbs for each photosite. > > A Fovean sensor stacks three photosites in each pixel location on the > sensor, with a color filter between them. I don't know precisely how > they order the stack, but it would make sense to order it blue, green, > red since short wavelengths are more easily absorbed than longer > wavelengths. But what this means is that the light energy hitting the > bottom layer is inevitably reduced compared to what hits the top > layer. I thought one the main premises was that each sensor would absorb all the light for a certain range of wavelenghts, and not at all respond light outside that range, so that it would continue without any loss of energy to the next layer(s) (absorbing different wavelengths.) And also, that there weren't filters as such, just differences in the way the photo diodes respond to light, or whatever.
But maybe the problem is that it's not quite like that in the real world... - Toralf > They compensate to some degree by making the three layers > differentially sensitive, either by adjusting the signal amplification > or by staggering the photosite sizings, I don't know. > > But what this means is that for a given amount of light energy per > unit area falling on the sensor plane, a Fovean sensor has to work > harder to achieve the same signal to noise characteristic than a Bayer > matrix sensor. > > The difference is that a Fovean sensor is acquiring R, G, and B values > at each pixel position, producing a non-interpolated chroma resolution > at the same pixel resolution, where a Bayer mosaic takes a minimum of > an 8x8 pixel array in a GRBG/RBGG/BGGR row*column (or similar > ordering) to resolve each pixel's position and RGB value by > interpolation. The article makes a big deal of the Fovean design not > needing an AA filter to combat chroma irregularities due to the > interpolation of pixel position. Well, as pixel density increases, the > need for an AA filter to minimize chroma irregularities reduces such > that with today's sensors the spatial resolution is very nearly > identical to the Fovean sensor's capabilities at the same single-layer > resolution. > > AA filters do more than simply minimize chroma irregularities ... they > also minimize moire artifacts, the effect of discrete line crossings > on a cartesian grid when the cell size of the grid is close to the > width of the line. Again, as grid cell size decreases, the tendency to > moire also decreases. > > All engineering solutions to the problem of obtaining enough color > information to accurately characterize the incoming light are > tradeoffs. Fovean technology means more complex to manufacture, > multi-layer sensors, which have known plusses and minuses compared to > Bayer mosaic sensors. My feeling is that the quality of Bayer mosaic > sensor technology is such that for practical purposes, the Fovean > design is mostly irrelevant except for some very critical color niche > market uses. > > If the cost of making Fovean sensors were to drop, and a solution to > the multi-layer light absorbtion issue resolved, it might stand a > chance of becoming more popular than a Bayer mosaic design ... as then > there would be more demand for the sensors and the price would drop > further by the scaling effect of large volume manufacturing. -- PDML Pentax-Discuss Mail List PDML@pdml.net http://pdml.net/mailman/listinfo/pdml_pdml.net to UNSUBSCRIBE from the PDML, please visit the link directly above and follow the directions.
