Image source @ Visual China
Text | Ray Tech leitech
According to the past practice, with the release of Qualcomm and MediaTek's new mobile flagship SoC, a wave of new machine releases will be ushered in from January to March every year, and the annual new flagships of mainstream Android mobile phone manufacturers will be unveiled in this period. The arrival of the new model also means the improvement of mobile phone image strength, compared with the previous high-profile appearance of ultra-high pixel sensors, this year's mobile phone CMOS market has undergone some new changes.
In mid-to-late December this year, Motorola and Xiaomi released new models equipped with snapdragon 8, and on CMOS, the Motorola edge X30 and Xiaomi Mi 12/12X both use high-pixel midsole sensors. It is worth noting that the Xiaomi 12/12X deprecates the 100 million pixel sensor of the previous generation and uses the IMX766 that many mobile phone manufacturers are using.
Not only Motorola and Xiaomi, but many models this year have chosen to use high-pixel midsole CMOS. The choice of manufacturers illustrates the change in the direction of the market, away from the one-sided numbers no longer blindly pursue ultra-high pixels, the use of more mature hardware solutions.
Ultra-high pixels are not directly equivalent to high quality
It should be noted in advance that the CMOS image sensor is part of the mobile phone camera module, which is responsible for converting the light signal through the lens into a digital signal that the chip can understand. In addition, the mobile phone camera module also includes a lens and a lens module, the former is responsible for transmitting light, and the latter is to package the lens, CMOS, OIS optical image stabilization and other modules together. The so-called "high-pixel midsole CMOS" refers to a sensor with a specification of 5000 pixels and about 1/1.5 inch, which is different from the ultra-high pixel outsole sensor of about 100 million pixels and about 1/1.3 inch.
To judge the quality of a CMOS, you can look at the total pixels, the size of the pixels per unit, the sensor's light-sensitive area and focus technology. Smartphone body can be used in limited area, for mainstream mobile phones directly stuffed with a super-size CMOS is not realistic, in recent years, manufacturers in order to improve the quality of mobile phones began to try to take the ultra-high pixel route, by reducing the number of pixels per pixel to increase the total number of pixels of the sensor. For example, Samsung's HM1 has 108 million pixels, but the unit pixel size is only 0.8μm, which is much smaller than the 1.8μm of the 12 million pixel Sony IMX555.
The advantage of ultra-high pixel sensors is that, on the one hand, it can increase the unit pixel size with multi-pixel synthesis, and Samsung HM1 pixels can output 12 million pixels /2.4μm photos after nine-in-one; on the other hand, a 100 million pixel sensor is conducive to product promotion, and it is easy to rely on the product characteristics of ultra-high pixels to attract consumers to buy, especially in the sinking market, some users also think that high pixels must be equivalent to high image quality.
The ideal is very full of reality is very bone, after several years of development of 64 million pixels and 100 million pixel sensors have not become the mainstream of the market. Not only the Xiaomi Mi 12 series returns to the 50-megapixel sensor, even Samsung has not popularized a hundred million pixels in its own Galaxy S series, the S20 and S21 generations of products only ultra models are equipped with 100 million pixel CMOS, and several other models use "conservative" 12 million pixel sensors. Ultra-high pixels are difficult to popularize, mainly including the following major reasons.
First of all, the ultra-high pixel sensor did not bring a qualitative leap in the imaging quality of the mobile phone, but increased the computational pressure of the ISP, causing problems such as photo stuttering and abnormal heat, affecting the user experience. Whether it is directly output 100 million pixel photos or pixel all-in-one output of 12 million pixel photos, the ultra-high pixels greatly increase the amount of ISP computing in a short period of time, and it is difficult to achieve an insensible photo experience (that is, shooting and getting). For example, the Xiaomi CC9 Pro equipped with a 100 million pixel sensor has been complained about by some users when taking pictures, and the viewfinder is stuck, and there will be a waiting process when quickly tapping the shutter.
Secondly, the higher the pixels and the smaller the size per pixel CMOS, the more sophisticated process technology is required, that is, the cost of manufacturing ultra-high pixel CMOS is higher. CMOS is different from mobile phone SoC, does not need to chase the top process, but the reduction of CMOS unit pixel size makes the traditional 40nm + production line unable to achieve accurate cutting, can only use 40-28nm process.
After The Android mobile phone entered the 48 million/50 million pixel era, the important reason why Samsung caught up with Sony in the CMOS market is that it has an advanced process production line, and Sony's own old process production line cannot meet the requirements of manufacturing ultra-high pixel CMOS, and now it can only find TSMC to cooperate.
Sensors such as 100 million pixels will continue to exist, but high-pixel midsole sensors will become more popular and become standard in mainstream mid- and high-priced models. In addition, some "mature" sensors may become popular and be installed on more new models.
"Mature" CMOS will be more popular with vendors
The Sony IMX766 used by Xiaomi 12/12X is not the latest COMS, the earliest model is the OPPO Reno 5 Pro+ released at the end of 2020, but with strong comprehensive strength, IMX766 has been adopted by many manufacturers after the OPPO exclusive period. As an important model of Xiaomi, Xiaomi 12 did not use the latest CMOS to highlight the development characteristics of Xiaomi's stability, and bid farewell to a single CMOS hardware competition.
For mobile phone manufacturers, there are many benefits to using a CMOS that has been published for some time. In terms of algorithm debugging, there are more relevant models released on the market, and manufacturers have more reference cases and debugging time, thereby reducing the probability of product rollover. For example, Google has polished the IMX363 for several consecutive years, relying on the upgrading of the algorithm to enhance the strength of mobile phone photography.
At the cost level, the advantages of the mature CMOS at the cost level are not only reflected in the purchase price, but also in the development and adaptation of the algorithm, manufacturers do not need to re-develop a set of algorithm adaptation to the new model, more is to adjust the original optimization strategy according to different lenses and ISPs. Xiaomi Mi 11 Ultra's ultra-wide angle and telephoto lenses use the IMX586 (the main camera of the original Xiaomi Mi 9), a long period of polishing and good hardware specifications, so that the IMX586 this mature CMOS glows the second spring, the comprehensive imaging quality does not lose the specialized ultra-wide and telephoto CMOS.
The way to apply mature CMOS is not uncommon, but many Android mobile phone manufacturers prefer to chase new, and there are many mobile phones equipped with mature CMOS in the middle and high price segments, but this situation has changed in the past two years.
What is slightly different now is that the mature CMOS is more used for mobile phone main cameras, and some even appear a generation of CMOS for the third generation of models. The success of OPPO and vivo in the offline market shows that blindly chasing ultra-high pixels is not advisable, and most users do not care too much about whether some CMOS is new, but more about the specific effect of taking pictures. Similarly, telephoto macro and "microscope" lenses emphasize gameplay upgrades, both of which are mostly small-bottom low-pixel CMOS.
Personally, I believe that the smart phone industry has gone through a barbaric period of development only relying on hardware stacking, with the maturity of the mobile phone market and the development of the economy, many users are more concerned about the mobile phone manufacturers' own self-developed technology when purchasing mobile phones. The application of mature CMOS does not represent squeezing toothpaste, only using the new CMOS does not enhance the strength of self-developed algorithms is the real squeeze toothpaste, supply chain technology manufacturers can buy, but in order to seek a breakthrough in the high-end mobile phone market can only start from the self-developed technical level.
In addition to CMOS, self-developed software and hardware modules are already an industry trend
At this stage, the main reason for hindering the improvement of mobile phone camera/video capabilities lies in the physical layer rather than the software layer, such as the lens, the space limitation makes the number of lens elements of the mobile phone lens less than 10 pieces, and it is mainly plastic lenses with general light transmittance. Physical limitations make the changes that mobile phone manufacturers can not make much in the lens module, and can only turn to algorithms and image processing hardware.
Promoted by Google, "computational photography" has been considered to be one of the main ways to improve the bottleneck of mobile phone photography, such as handheld super night scene algorithm, multi-frame synthesis technology and wide-angle distortion adjustment algorithm, breaking through the limitations of lens and CMOS physical level to improve the imaging effect of mobile phones. Domestic manufacturers such as Huawei, Xiaomi and Vivo will emphasize the importance of self-developed algorithms at the press conference, reducing the influence of CMOS in the hearts of users. It can be said that the era when you can rely on third-party imaging algorithms to improve the quality of mobile phone photography is gone, and brands want to gain a foothold in the high-end mobile phone market, and they have to work software algorithms.
Self-developed ISPs and NPUs are another route to improve the strength of mobile phone images, more precisely, players who have a place in the high-end machine market are basically their own ISPs. This year, mi OV three are working on ISPs or NPUs, focusing on building a self-research system of algorithm + imaging hardware, and mastering more technical links that affect the imaging quality of mobile phones in their own hands.
Finally, it should be said that the technical leap from the feature machine to the smart phone is essentially a leap in chip technology, and the upgrade of many functions depends on the improvement of the chip computing power. For mobile phones with limited size, the role of "pixels" is getting smaller and smaller, one-inch CMOS seems to be the tipping point of mobile phone image development, the ceiling of the industry's physical level is overhead, since higher quality photos are difficult to rely on hardware "pile" out, it is better to rely on algorithms and chips to "calculate" out.