What are the characteristics of Sigma x3 series sensors? Is it an advanced concept?

As one of the few people in China who have really used Sigma FOVEON sensor camera, let me say a few words.

I won't repeat the concept of X3 sensor. In short, it is a multi-layer sensor. You can see other people's answers. The easiest way is to look at the standard explanation of Sigma's official website.

What I want to explain is the fallacy of China people with some unscientific eyes.

The biggest fallacy is that X3 sensor, like film, has a color mask between different layers, so many people think that X3 sensor has the worst performance for the red photosensitive layer placed at the bottom.

Actually, this is not right.

Because of the basic principle of X3 sensor, please remember:

When light passes through a silicon wafer, it has different absorption characteristics at different thicknesses. At the lowest level, only red wavelength energy is left.

This is the basic principle of X3 sensor.

Therefore, X3 sensor does not have the problem of favoritism in color reproduction.

The picture above shows the real principle of X3 series sensors.

Like many technical routes, X3 sensor has its advantages and disadvantages.

I used sd Quattro myself for a long time.

This camera has huge calorific value, slow response speed (almost no continuous shooting), huge volume and almost no high sensitivity, all of which are caused by the inherent deficiency of X3 sensor in camera design.

But it also has the characteristics of very pure picture quality and heavy color. The picture has a strong three-dimensional sense.

This kind of camera is loved by people who love it and almost accept its shortcomings, such as me; People who hate it can hate it until they never use it.

As a photography enthusiast, I am a loyal user of Sigma SLR. 14 years, with four sigma machines. Let me tell you how I feel. The first machine was SD 10, which was bought in 2005. I was poisoned when I saw the SD9 photo of the pig in the endless shadow. The color is particularly solid. The image has a strong three-dimensional sense. But I'm really worried about the color cast and manipulation of the machine. After waiting for a year, SD 10 went on the market. I heard that the color and control have improved, so I bought it without hesitation. When I came back, I found that the color was really good and the control was really poor. The film rate is less than 10%. The pictures are all patients with jaundice. It takes two minutes to take a photo. But it's really good to come out occasionally. About 1 1 year, SD 15 was changed, and the color was greatly improved. Except for the photographer, other subjects can be photographed without too much post-color correction. The biggest problem is out of focus Have you ever seen your own fuselage with your own lens out of focus? Sima is the only one here. Later, a backup machine DP 1S was added, and it really didn't run out of focus. It is slower to control than Sd 10. Such a small machine uses the speed of medium format. Later 18, SDQH was changed. This time it is finally a working camera, with no color cast and out of focus. The reaction speed is still slow, but it is acceptable. But the price is that the pixel structure is compromised and the color is really not solid enough. But I think the price is worth it. After all, it is the first priority to take usable photos. At this time, the standby machine is RX 1R2. The two machines are highly complementary and the output ratio is much higher. My experience is that you can buy the right horse. But depending on what subject you are shooting, the picture quality is unique. Also, don't buy odd products. In my opinion, SD9, SD 14 and SD 1M are not mature enough, so they can only be regarded as experimental products. It's safer to buy their improved model. Don't be a mouse.

Thank you for inviting me!

As a vice factory, Sigma is really a "great" company. Often whimsical, unconventional, especially with its "stacking materials", self-proclaimed "black technology."

Bayer array refers to a common method for collecting digital images when CCD (Charge Coupled Device) or CMOS (Complementary Metal Oxide Semiconductor) is used as photosensitive element.

Bayer array sensor was invented by Bryce Edward Bayer, an engineer of Kodak Company, and registered a patent in 1976. This great inventor died at the end of 20 12 in Maine, USA.

Because the photodiode, the basic unit of the image sensor, can measure the light intensity, but can't perceive the color and can only take black and white photos, Bayer creatively designed a color filter matrix with 50% green, 25% red and 25% blue to cover the photodiode to achieve the purpose of taking color photos.

This design solves the problem of color recognition with only one image sensor. The method is to set a color filter array in front of the image sensor, and the array is covered with filter points, which correspond to the lower pixels one by one. Each filter point can only pass one color of red, green and blue, which means that the pixels below it can only have three colors: red, green and blue, or nothing (black). The arrangement of filter points of different colors is regular: around each green point, there are 2 red dots, 2 blue dots and 4 green dots. This means that there are twice as many green dots as the other two colors. This is because the research shows that the human eye is most sensitive to green, so the filter layer has the most green spots.

If a pixel can only have three colors, how can you take a color photo? As mentioned above, there are filter points of other colors "regularly" distributed around each filter point, so the primary color of light can be judged by combining their values. Take the yellow light as an example. It is a mixture of red light and green light. After passing through the filter layer, the pixels below the red dot and the green dot will have values, and the pixels below the blue dot will have no values. Therefore, by looking at the color distribution around a pixel-red, green and blue-it can be inferred (interpolation calculation) that the original color of this pixel should be yellow.

This method of calculating color is called "demosaicing". The lower part of the above picture is a "mosaic" image generated by the image sensor, and all pixels have only four colors: red, green, blue and black. The upper part is the effect after "demosaicing", that is, the result of interpolation algorithm processing. At present, most of the photosensitive elements of digital cameras use Bayer array to generate color digital photos. In memory of the inventor Braez Bayer, it is called "Bayer pattern" or "Bayer filter", and Bayer is also called "the father of digital images".

The biggest problem of Bayer array CMOS sensor is moire fringe and pseudo color. The solution is to install a low-pass filter on the sensor surface to reduce the degree of moire fringes and pseudo color, at the cost of reducing the image quality of the details. However, in recent years, with the accumulation of a large number of pixels in Bayer array CMOS sensor and the design of removing low-pass filter from the camera, this kind of problem has been solved to some extent.

(1) Source of FOVEON X3 sensor

The Foveon X3 sensor was originally developed by Foveon Company founded by Professor Carver Andress Mead of California Institute of Technology. At that time, the team led by Dick Merrill, the chief engineer, and Dick Lyon, the chief scientific officer, invented the Foveon X3 sensor and successfully commercialized it.

The first camera equipped with Foveon X3 sensor was Sigma SD9 digital SLR camera released in 2002. Sigma SD9 is equipped with APS-C Foveon X3 sensor, with a size of 20.7× 13.8mm, a single-layer pixel of about 3.5 million, and a * * * three-layer pixel arrangement.

(2) 2) The principle of FOVE ON X3 sensor

This structure of Foveon X3 sensor is similar to that of film. Each pixel is actually composed of three layers of red, green and blue pixels, and each pixel can fully perceive three colors. However, the traditional Bayer array sensor has only one layer of pixels, on which there are three kinds of pixels: red, green and blue. Each independent pixel can only perceive one color and lose the other two colors. The final color is completed by later interpolation calculation (color guessing).

Simply put, because the wavelengths of different colors of light are different, the red wavelength is greater than the green wavelength, and the green wavelength is greater than the blue wavelength. The order of arrangement is short wave first, long wave last. It is not a direct color sensitive layer like film, but depends on sensing different wavelengths, absorbing corresponding color wavelengths and calculating corresponding colors. In the three-layer structure of X3 sensor, the first layer feels full color including three primary colors of red, green and blue. After passing through the first layer, the blue light is absorbed by the sensor, and the remaining red light and green light reach the second layer to be felt and absorbed by the green light wave. The light wave reaching the bottom layer is only red, and the third layer of the sensor feels and produces the only color that can be recognized without calculation-red electric signal.

Therefore, theoretically, the Foveon X3 sensor can directly calculate the RGB value of each pixel position, and the high sampling rate also means that the incidence of moire fringes is greatly reduced, and high-frequency information can be filtered without a low-pass filter, which can show rich colors and amazing details.

Arguably, because there is no need to interpolate and demosaize RGB data, the data storage speed of Foveon X3 sensor should be faster, but in fact, the extremely high sampling rate is accompanied by a substantial increase in data volume. At the same time, the light will inevitably lead to data loss after penetrating each layer of pixels, so the amount of light felt by each layer of pixels is different (the bottom red pixel is the most unfavorable), which leads to the deviation of the three color information in the final synthesis, and it is easier to cast color when reflected in the photo; In addition, the loss of optical signal caused by this three-layer pixel arrangement structure is bound to be the reduction of signal-to-noise ratio and high sensitivity.

(3) Foveon X3 Quattro full color image sensor.

In order to solve the problem that the image file capacity is too large when the RAW format stores the image data of RGB information of each pixel position, the structure of the new generation Foveon X3 Quattro sensor is changed from the previous1:1to 1: 65438+4 to allocate red, green and blue pixels. The number of photodiodes in the top layer responsible for recording blue light and dark information is four times that of recording green light and red light, respectively. The photodiodes in the middle layer and the lower layer are only responsible for recording the corresponding colored light, and do not participate in recording the light and dark information. Subsequently, in the process of data processing, the light and dark information of the upper layer is also applied to the middle layer and the bottom layer. According to official data, the resolution of Foveon X3 Quattro sensor with new structure is improved by 30%, the data processing speed is faster, and the camera life is effectively improved.

Take Sigma's SD Quattro as an example, the blue top layer is19.6 million pixels, and the green and red layers are 4.9 million pixels. After weighting the brightness information of the top layer to the lower two layers, the realization that "APS-C frame can still display the equivalent image quality of 39 million pixels (from Sigma official website)" is actually the biggest controversy of Sigma Foveon X3 sensor-the three-layer superposition algorithm.

Although the structural change does reduce the overall data obtained by the sensor and the data processing speed is faster, there is a natural fatal defect in this way of laying three layers of photodiodes vertically, that is, the "absolute sensitivity threshold" is too high. The absolute sensitivity threshold is the number of photons that make the signal equal to the noise generated by the sensor. This is an important index, because it represents the minimum amount of light required to observe any meaningful signal theoretically. In the weak light environment, with the decrease of the absolute light entering the sensor, it is bound to be accompanied by the gradual attenuation of the light. In other words, compared with the traditional Bayer array CMOS sensor, X3 sensor needs more incident photons to obtain the same signal-to-noise ratio, which ultimately leads to extremely weak high sensitivity of X3 sensor. Simply put, the actually available ISO range is only the lowest-ISO100.

In a word, X3 structure of Sigma does have its own uniqueness in design concept. In the static environment with sufficient ambient light and low sensitivity (ISO 100), it is indeed possible to obtain a more delicate and clear image, with "rich colors and amazing details". In order to emphasize this ability more, Sigma subsequently introduced SFD (ultra-fine detail) shooting mode. That is, the so-called "Foveon Quattro sensor extracts the maximum number of details from the image while controlling the image noise at a very low level"; In fact, it is centered on standard metering. The camera automatically takes seven photos with the exposure range of -3 EV to +3 EV (that is, the exposure of plus or minus 3 EV), and then saves them as a single X3I file, and synthesizes HDR images in SIGMA PhotoPro(SPP), a special RAW image conversion software for Sigma. However, this method of multi-frame exposure and post-synthesis is only suitable for still life or completely windless environment. If the subject moves a little, the post-synthesis will be quite simple and rude, and procrastination and dynamic blur are completely inevitable.

Although Sigma constantly emphasizes that the complete data provided by Foveon X3 Quattro is captured from each pixel unit, it does not need computer interpolation to supplement, and it has the characteristics of faithfully presenting the scenery it sees. (From Sigma official website) "However, its natural structural defects lead to weak high sensitivity, and there is no possibility of technical solution in the short term. Therefore, the smart camera developed based on X3 technology is destined not to be the darling of the mainstream market, but only to be a niche player who pursues extreme alternatives. As the saying goes, "those who love it caress it, those who hate it laugh at it."

There are no stupid questions, only boring answers.

Thank you for watching!