The Foveon X3 sensor is a CMOS image sensor for digital cameras, designed by Foveon, Inc. (now part of Sigma Corporation) and manufactured by Dongbu Electronics. It uses an array of photosites, each of which consists of three vertically stacked photodiodes, organized in a two-dimensional grid. Each of the three stacked photodiodes responds to different wavelengths of light; that is, each has a different spectral sensitivity curve. This difference is due to the fact that different wavelengths of light penetrate silicon to different depths. The signals from the three photodiodes are then processed, resulting in data that provides the amounts of three additive primary colors, red, green, and blue.
The development of the Foveon X3 technology is the subject of the 2005 book The Silicon Eye by George Gilder.
The diagram to the right shows how this works in graphic form. Depicted on the left is the absorption of colors of the spectrum according to their wavelength as they pass through the silicon wafer. On the right, a Foveon X3 layered sensor stack in the silicon wafer for each output pixel is shown depicting the colors it detects at each absorption level. The color purity and intensity of blue, green and red depicted for the sensors are for ease of illustration. In fact, the attributes of each output pixel that are reported by a camera using this sensor result from the camera's image processing algorithms that employ a matrix process to construct the single RGB color from the data sensed by the photodiode stack. The results, in terms of color accuracy (metamerism index), were state of the art at the time of their invention.
Because the depth in the silicon wafer of each of the three layer Foveon X3 sensors is less than five micrometres, it has negligible effect on focusing or chromatic aberration. However, because the collection depth of the deepest sensor layer (red) is comparable to collection depths in other silicon CMOS and CCD sensors, some diffusion of electrons and loss of sharpness in the longer wavelengths occurs.
The first digital camera to use a Foveon X3 sensor was the Sigma SD9, a digital SLR launched in 2002. This used a 2268x1512×3 (3.54×3 MP) iteration of the sensor, and was based around a Sigma-designed body using the Sigma SA mount. The camera was followed in 2003 by the improved but technically similar Sigma SD10, which was in turn superseded in 2006 by the Sigma SD14, which used a higher-resolution, 2640×1760×3 sensor. Sigma announced a successor, the Sigma SD15, in 2008, although the camera did not go on sale until June 2010. It used the same 2640×1760×3 (4.7×3 MP) sensor as the SD14. As of 2011, it is Sigma's current prosumer digital SLR. In September 2010, the company announced the Sigma SD1, which uses a new, 4800×3200×3 sensor, and was aimed at the professional market.
In 2004, Polaroid Corporation announced the Polaroid x530, a compact camera based around a 1408×1056×3, 1/1.8" sensor. The camera received a limited release in 2005 but was recalled later in the year for unspecified image quality problems. Sigma announced a prototype of their own Foveon-based compact camera in 2006, the Sigma DP1, using the same 14 MP sensor as the SD14 DSLR. A revised version of the prototype was exhibited in 2007, and the camera was eventually launched in Spring 2008. Unlike the Polaroid x530, the DP1 was based around an APS-C-sized sensor, with a 28mm equivalent prime lens. The camera was subsequently revised as the DP1s and the DP1x. In 2009, the company launched the DP2, a compact camera based around the same sensor and body as the DP1, but with a 41mm-equivalent f/2.8 lens.
The operation of the Foveon X3 sensor is quite different from that of the Bayer filter image sensor more commonly used in digital cameras. In the Bayer sensor, each photosite in the array consists of a single light sensor (either CMOS or CCD) that, as a result of filtration, is exposed to only one of the three primary colors, red, green, or blue. Constructing a full color image from a Bayer sensor requires demosaicing, an interpolative process in which the output pixel associated with each photosite is assigned an RGB value based in part on the level of red, green, and blue reported by those photosites adjacent to it. The Foveon X3 sensor creates its RGB color output for each photosite by combining the outputs of each of the stacked photodiodes at each of its photosites. This operational difference results in several significant consequences.
Because demosaicing is not required for the Foveon X3 sensor to produce a full-color image, the color artifacts ("colored jaggies") associated with that process are not seen. The separate anti-aliasing filter commonly used to mitigate those artifacts in a Bayer sensor is not required. This is because little aliasing occurs when the photodiodes for each color, with the assistance of the microlenses, integrate the optical image over a region almost as big as the spacing of sensors for that color. On the other hand, the method of color separation by silicon penetration depth gives more cross-contamination between color layers, and therefore more issues with color accuracy especially with the red channel.
Another difference is that more of the photons entering the camera will be detected by the Foveon X3 photosensor than is possible with a mosaic sensor. This is because each of the color filters overlaying each photosite of a mosaic sensor passes only one of the primary colors, absorbing the other two. The absorption of these colors reduces the total amount of light gathered by the sensor and destroys much of the information about the color of the light impinging on each sensor element. Although the Foveon X3 has greater light gathering ability, the individual layers do not respond as sharply to the respective colours. Thus color-indicating information in the sensor's raw data requires "aggressive" matrixing (essentially, removal of common-mode signals) to produce color data in a standard color space, which can increase color noise in low-light situations.
According to Sigma Corporation, "there has been some controversy in how to specify the number of pixels in Foveon sensors." The argument has been over whether sellers should count the number of photosites, or the total number of photodiodes, as a megapixel count, and whether either of those should be compared with the number of photodiodes in a Bayer filter sensor or camera as a measure of resolution.
For example, the dimensions of the photosite array in the sensor in the Sigma SD10 camera are 2268 × 1512, and the camera produces a native file size of those dimensions (times three color layers). This amounts to approximately 3.4 million three-color pixels. However, it has been advertised as a 10.2 MP camera by taking account of the fact that each photosite contains stacked red, green, and blue color sensing photodiodes, or pixel sensors (2268 × 1512 × 3). By comparison, the dimensions of the photosite array in the 10.2 MP Bayer sensor in the Nikon D200 camera are 3872 × 2592, but there is only one photodiode, or one pixel sensor, at each site. The cameras have equal numbers of photodiodes, and produce similar RAW data file sizes, but the Bayer filter camera produces a larger native file size via demosaicing.
However, the actual resolution produced by the Bayer sensor is more complicated than the count of its photosites, or its native file size, might suggest. The reason has to do with both the demosaicing and the separate anti-aliasing filter commonly used to reduce the occurrence or severity of color moiré patterns that the mosaic characteristic of the Bayer sensor produces. The effect of this filter is to blur the image output of the sensor, thus producing a lower resolution than the photosite count would seem to imply. This filter is largely unnecessary with the Foveon X3 sensor and is not used. The earliest camera with a Foveon X3 sensor, the Sigma SD9, showed visible luminance moiré patterns, but not color moiré. Subsequent X3-equipped cameras have less aliasing because they include microlenses, which provide an effective anti-aliasing filter by averaging the optical signal over an area commensurate with the sample density, which is not possible in any color channel of a Bayer-type sensor. Aliasing from the Foveon X3 sensor is "far less bothersome because it's monochrome" according to Norman Koren. Therefore, in theory, it is possible for a Foveon X3 sensor with the same number of photodiodes as a Bayer sensor and no separate anti-aliasing filter to attain a higher spatial resolution than that Bayer sensor. Independent tests indicate that the "10.2 MP" array of the Foveon X3 sensor (in the Sigma SD10) has a resolution similar to a 5 MP or 6 MP Bayer sensor, and at low ISO speed even similar to a 7.2 MP Bayer sensor.
With the introduction of the Sigma SD14, the 14 MP (4.7 MP red + 4.7 MP green + 4.7 MP blue) Foveon X3 sensor resolution is being compared favorably by reviewers to that of 10 MP Bayer sensors. For example, Mike Chaney of ddisoftware says, "the SD14 produces better photos than a typical 10 MP dSLR because it is able to carry sharp detail all the way to the 'falloff' point at 1700 LPI whereas contrast, color detail, and sharpness begin to degrade long before the 1700 LPI limit on a Bayer based 10 MP dSLR." Another article judges the Foveon X3 sensor as roughly equivalent to a 9 MP Bayer sensor. A visual comparison between a 14 MP Foveon sensor and a 12.3 MP Bayer sensor shows Foveon has crisper details.
The Foveon X3 sensor, as used in the Sigma SD10 camera, has been characterized by two independent reviewers as noisier than the sensors in some other DSLRs using the Bayer sensor at higher ISO film speed equivalents, and specifically chroma noise has been noted. Another has noted higher noise during long exposure times. However, these reviewers offer no opinion as to whether this is an inherent property of the sensor or the camera's image processing algorithms.
With regards to the Sigma SD14 which uses a more recent Foveon X3 sensor, one reviewer judged its noise levels as ranging from "very low" at ISO 100 to "Moderate" at ISO 1600 when using the camera's Raw image format.
Sigma's SD14 site has galleries of full-resolution images showing the color produced by the current state of Foveon technology. The 14-MP Foveon chip produces 4.7 MP native-size RGB files; 14-MP Bayer filter cameras produce a 14 MP native file size by interpolation (demosaicing). Direct visual comparison of images from 12.7-MP Bayer sensors and 14.1 MP Foveon sensors show Bayer images ahead on fine monochrome detail, such as the lines between bricks on a distant building, but the Foveon images are ahead on color resolution.