The Optical Wavefront Camera

This style of camera uses microlenses placed in front of a sensor which causes the light rays to defract and place the resulting image over a number of pixels. In post-processing the decompiling algorithm has an adjustable "mask" which allows the user to select the focus plane after the fact. This allows for 3D microphotography, extreme DoF, selective DoF and adjustable focus well after the actual photograph has been taken. This technology isn't actually new, but the practical application outside of the scientific community has been limited by computer processing horsepower and the pixel-density in the sensors. If a camera sensor could have 200+ MP, wavefront photography would be viable for general-purpose photography.

The Digital Optical Wavefront Sensor

Non-Bayer Array Sensors

Foveon has been the only company to bring a commercially viable non-array sensor to market. This sensor uses all three green, red and blue detectors at each pixel location. By stacking the detectors vertically on the sensor, each pixel location has full visibility of the entire visible color spectrum. As mentioned previously for something else, the technology is sound but the execution has been poor. Foveon's sensor is slow in the ability to move the image off the chip. This isn't a problem with the stacked sensors, but is a problem with all the supporting circuitry and design. This is most certainly a money issue as chip design is extremely expensive and there are many cross-licensed technologies to both contend with and pay for.

But the single-pixel approach is something that makes a lot of sense and may be incorporated in one of the four following methods illustrated here in a cross-section of a sensor. All of these methods are becoming viable through the continuing developments in micro devices and molecular science. 

The first sensor on the left uses the standard tri-color sensor, but the incoming light into enters a doped fiber-optic chamber that is lit with an UV emitter from the bottom. The UV light collides with the incoming photons and they are knocked loose--essentially fluorescing in the appropriate colors to be detected by the green, red and blue sensors. The actual wavelength the detectors are sensitive to would not actually be in the visible light band! This technology is used today in scientific application as well as a variation is used in fiber-optic networks where the signal is amplified by the injection of another high-powered laser into a segment of chemically doped glass fiber.

The second sensor uses a refraction grating within the the pixel position itself. Actual detectors wouldn't necessarily be specific color sensitive as the refraction grating acts as a prism which breaks down the incoming light into the color components.

The third sensor would use micro-mirrors which reflect certain colors and pass on the rest. Again, the detectors themselves do not need filtering as the incoming light to the detector is already of the color range that detector needs to see.

The final sensor design in this illustration uses tuned laser detectors. These detectors are not color filtered, but by using a tuned chamber the presence of light of the appropriate wavelength will generate the electrical signal to be converted into digital data.

Not illustrated is probably the most likely scenario. With the continuing shrinking of pixels, we'll see all three color detectors placed under a single microlens. The converting of the RGB colors into a single full-color pixel could happen on-sensor, but would most likely be stored individually as it is done today. The three pixels would be combined and mapped as a single pixel in conversion.

One other approach which is used extensively in video cameras is a beam-splitter and three sensors. A search of patents reveals a revitalization of this concept and we are very likely to see this make an appearance in a still camera before long.

Conclusion

The pixel wars are not over. This is guaranteed. As a marketing tool, counting pixels is no longer a major selling point. This will drive the companies into bringing out sensors with better dynamic range--probably in the 18-24 stop range within the next three years, backlit sensors and even sensors with built-in LCD shutters!

We are entering the next phase of sensor development. Every technology mentioned in this article is either currently in development or has already been tried but is just waiting for the manufacturing ability to evolve to the point where it is viable to produce.

I believe that the multi-colored arrays is the most likely scenario to come out. It isn't the only way to address the wide-gamut issue, but does present distinct advantages over a three-color array. Three-color arrays will continue to evolve too with increased abilities to derive colors.

I have stayed away from sensor size in this article. We are getting to the point where sensor size has little to do with noise, color-fidelity and dynamic range, but is a choice based on DoF and Bokeh preferences. All one has to do is look at the image quality from Canon's S90 and G11 and you can see that the technologies applied to tiny sensors are also applicable in larger sensors.

This article is Copyright 2009 Kenneth E Norton

November 14, 2009.