The ups and downs of CMOS: From a two-horse race to domestic substitution

In a sense, the war of contemporary mobile phones has turned into a war of photography.

From single camera to quad camera, from 8 million pixels to 100 million pixels, in today's world where there are no revolutionary changes, the ability to automatically beautify photos taken at night for selfies has become the main battlefield for differentiation among brands.

Here, Huawei, Xiaomi and other mobile phone companies are competing fiercely, but fundamentally, the quality of a mobile phone's camera performance, the running scores and publicity of mobile phone manufacturers are just superficial work. What really determines the quality of a mobile phone's camera function is a chip called CMOS image sensor .

The so-called image sensor can convert the light signal captured by the lens into a digital signal that can be understood by the machine. In other words, it can draw the real world into the world of the chip. Generally speaking, the larger the sensor area, the larger the canvas area, and the more details that can be drawn. The performance of the sensor directly affects the quality of the camera, which is equivalent to the "eyes" of the camera.

This chip is not big in size, but it accounts for 52% of the cost of the entire camera module . The camera module is often the third largest cost source in a mobile phone, second only to the SoC chip and the screen.

At present, the three major players that can produce this "canvas" include Sony, which allows Apple, Huawei, and Xiaomi to roll up, as well as Samsung and OmniVision.

The history of global mobile phone image sensors can be summarized as the struggle between Sony and OmniVision, with Samsung reaping the benefits.

During the process, the changes in technology and production capacity have successively pushed different companies to the throne. During this period, OmniVision also changed from an American company to a Chinese company, leading China's CMOS to start another war from scratch to impact the throne:

From the 1970s to the 1990s, the global image sensor market was monopolized by Japan for more than 20 years.

During that period, the world's image sensor products were mainly high-pixel, high-difficulty-to-manufacture CCD (charge-coupled device image sensor) devices, with Sony and Panasonic being the leaders among them.

Relying on its leading position in the CCD and storage fields, Japan once accounted for more than half of the world's semiconductor production, and at one time, Japan accounted for 6 of the top 10 semiconductor companies in the world.

As the inventor of CCD, Americans were unwilling to lose the market. During the 20 years when Japan dominated the market, American companies had been looking for an alternative technology that could not only achieve the high-quality imaging of CCD, but also be low in production difficulty and cost to enable American companies to defeat traditional CCD.

In 1995, several Chinese students studying abroad in China founded a company called OmniVision in Silicon Valley, bringing hope to the United States.

This young company has been trying to use traditional CMOS technology (complementary metal oxide semiconductor image sensor) to replace traditional CCD in image sensors.

Difference between CCD and CMOS

If we compare the two processes directly, under the process conditions at the time, the image quality of CMOS was slightly inferior to that of CCD, but it could greatly reduce power consumption and cost, making it more suitable for the then-emerging consumer electronics market. That year, OmniVision exhibited its products at the industry-renowned Comdex computer show, and after being purchased by many Taiwanese customers, it achieved mass production in just one month [1].

Since then, CMOS technology has gradually become the mainstream of the image sensor market. OmniVision, which has been in the limelight, began to move on to the fast track and successfully listed on NASDAQ in 2000. In 2007, it entered the Apple mobile phone industry and became the earliest company in the Apple industry chain.

But the war of CMOS has just begun.

In 2000, when OmniVision was listed on NASDAQ, the old wizard Sony finally gave up its signature CCD and turned to the CMOS market.

In sharp contrast to its dominance in the CCD market back then, ten years after the transformation, until 2010, Sony's market share in CMOS was only 7%, while the market share of newcomer OmniVision had reached 50%.

No one could have imagined that the changes would happen so fast: In October 2011, Apple launched the iPhone 4S. Before the launch, some media reported that OmniVision would no longer be Apple's main rear camera supplier and would be replaced by Sony.

At that time, many people thought it was just a fake news that came from nowhere, but they never thought that on the day of the press conference, Apple took actual actions to prove the authenticity of the message. For the latest iPhone 4S, Apple will significantly reduce the supply share of OmniVision and instead support Sony as Apple's main CMOS supplier.

It’s not that Apple abandoned the company after having a flirtation with it, but that despite the surge in demand for Apple products, OmniVision’s production capacity has long been unable to keep up with Apple’s pace despite its leading technology. This involves two classic models of semiconductor production, IDM and Fabless.

1. Sony adopts the IDM model: that is, it handles the entire industry chain from manufacturing, design, packaging and testing. The advantage of this model is strong production capacity and the ability to meet customer needs in all aspects, but the disadvantage is that the company has heavy assets and huge investment costs.

2. OmniVision adopts the fabless model: the company is only responsible for design and development, but outsources production and packaging and testing to third parties. The advantage of this approach is that it requires little investment, can reduce the huge capital and manpower costs in the production process, and can concentrate funds on technology upgrades, but the disadvantage is that there is a risk of production capacity.

Sony, which has its own factories, has made no secret of its ambition for expansion: in 2011, Sony invested 40 billion and 100 billion yen in the development of CMOS in its Kumamoto and Nagasaki R&D centers to cope with the surge in market demand.

On the other hand, although OmniVision operates with light assets, faced with the sudden surge in demand, OmniVision, which does not have its own wafer fab, can only silently swallow the loss and fall from the first to the second supplier.

Later in 2013, it developed a stacked CMOS that was more advanced than OmniVision's. While OmniVision had not yet caught up with the technology, it had workers work overtime to produce it, further consolidating its position as the leading supplier of CMOS for Apple.

As for OmniVision, although Apple began to use Sony CMOS for the rear cameras of iPhone 4S and iPhone 5 in 2011, it still used OmniVision solutions for the front cameras to save face. After Sony officially launched a double counterattack of technology + production capacity, starting from iPhone 6, all models of iPhone, regardless of front and rear cameras, all used Sony CMOS, and OmniVision was completely kicked out of the Apple supply chain.

After 2011, the performance of OmniVision and Sony CMOS has fluctuated

The industrial model led to different outcomes for Sony and OmniVision. After 2012, OmniVision's market share dropped from the top to the third with only 11%. Sony completely took over OmniVision's previous high-end market customers, with a market share of more than 40%. While Sony and OmniVision were fighting, Samsung took advantage of the terminal support and the latest ISOCELL technology to tear off 20% of the market share from the two giants.

At this point, the two-strong CMOS situation has been determined, and it seems that OmniVision has no chance of success, but the story is not over yet.

In 2013, OmniVision could not beat Sony in the high-end market, so it began to move to China to enter the mid- and low-end markets. In 2019, it was acquired by Chinese semiconductor company Will Semiconductor. According to the development of the losing battle-being acquired plot, OmniVision is likely to decline, but why is OmniVision not only alive now, but also has the possibility of catching up with Sony?

This is related to the nature of CMOS image sensors. CMOS is a special product. As we all know, traditional digital chips such as CPUs and GPUs rely on advanced processes, while analog chips such as RF and power management chips rely more on the experience accumulated by veterans. But CMOS is a "double-headed eagle": it is an integration of analog and digital circuits.

On the analog side, it needs to sense subtle changes in light and convert light signals into electrical signals. However, a single pixel is micron-sized, so analog chips do not require particularly advanced processes.

But at the same time, CMOS also needs to convert electrical signals into digital signals that can be understood by the mobile phone SoC. This part is realized by the digital chip part integrated on the CMOS in the second half.

Since the key to imaging lies in the first half, and the optimal number of mobile phone pixels has long been stuck at 12 million pixels, Sony can not only easily handle it by operating its own factory, but also accumulate a lot of industry experience, gaining advantages in both cost and technology.

Until recent years, sensor manufacturers, led by Samsung, have launched the "small pixel" technology route, pursuing smaller pixel size and higher pixel density, and copied Sony's back road with a cost-effective route.

Following this trend, Samsung and OmniVision have successively reduced the pixel size to below 0.8 microns, causing the number of pixels in sensors of the same size to increase suddenly. Sensors that previously could only have 20 or 40 million pixels have now been increased to 48 million, or even 64 million or 100 million pixels. This has brought consumers a great sense of satisfaction: a mobile phone purchased for 3,000 yuan, if you only look at the parameters, has several times higher pixels than Apple.

(Even Sony has used small pixels to increase the number of pixels without changing the CMOS area)

This puts Sony in a dilemma. If it follows suit, it will have to use old equipment to practice its craftsmanship, and to make small pixels it will have to switch to new equipment with advanced processes, and a lot of its experience will be wiped out. If it doesn't follow suit, the market for mobile phones that focuses on price-performance ratio will start to be eroded.

As for OmniVision, it used TSMC's advanced technology to quickly enter the market below 3,000 yuan, which has a huge demand for small pixels, and began to break Sony's iron wall.

On the one hand, the technology route is evolving, and on the other hand, the pattern of CMOS and even the entire semiconductor industry is also changing. For example, the birth of TSMC has made vertical professional division of labor possible in the semiconductor industry. 10 years ago, Sony's IDM model had advantages, but as TSMC's production capacity and process level became increasingly advanced, Sony's factories have begun to fall behind TSMC.

Therefore, TSMC became the watershed between OmniVision and Sony, which can be analyzed in two time periods:

1. 2010 period: CMOS mainly competed in production capacity and design technology. At this stage, OmniVision adopted the fabless model and handed over production to TSMC. Sony adopted the IDM model. At that time, Sony's CMOS foundry level was stronger than TSMC. In terms of design, Sony's CMOS products were also more advanced than OmniVision, so Sony completely crushed OmniVision during this period.

2. 2020 period: During this period, TSMC not only ranked first in the industry in terms of wafer production capacity and process, but also had the 40-28nm CMOS process required for small pixels.

Due to the strong demand for chips, Sony has been unable to meet the production capacity needs of customers in the past two years, and the performance of its own 55nm process has gradually reached a bottleneck, and Sony's self-built process production barriers have begun to weaken. Therefore, at the end of 2019, Sony placed its first high-end chip order with TSMC to supplement its production capacity, and was also able to introduce TSMC's 28nm process.

In other words, in the CMOS production link, OmniVision and Sony started to stand at the same starting line, which provided the possibility for OmniVision to catch up with Sony. Moreover, OmniVision, which had been in decline for a long time, also encountered a new opportunity in 2019: the Well Group from the mainland acquired OmniVision with 15.3 billion yuan this year. After Well Group acquired OmniVision, it invested heavily in research and development, breaking through 48 million and 64 million pixel technology in 2019 and 2020 respectively, gradually catching up with Sony and Samsung in terms of technology, and OmniVision has the hope of returning to the first echelon.

But all of this is based on a premise: OmniVision, which has always made low-end and mid-range products in China, will anyone buy its high-end products now? This is the crux of the matter.

Soon, Howe had the major opportunity of being used by someone.

In May 2019, the US government included Huawei in the Entity List and banned Intel, Qualcomm and other companies from supplying Huawei. Huawei launched the "spare tire plan" and began to consciously cultivate domestic companies and accelerate the introduction of domestic suppliers to replace them.

According to the disassembly and analysis of Huawei's Mate 30 mobile phone launched in September 2019 by Japanese research agency Fomalhaut Techno Solutions, in the four months after the US ban, the utilization rate of Chinese-made components has increased significantly from 25% to 42% in terms of amount, while the utilization rate of US-made components has dropped from 11% to 1%.

OmniVision took on this round of domestic substitution opportunities. Nomura Orient International Securities once analyzed that in 2020, Huawei and OPPO purchased a large number of 32-megapixel and 48-megapixel products from Welltech (OmniVision). In the same year, the Xiaomi Mi 10 Ultra Commemorative Edition high-end flagship phone was launched, with the rear main camera equipped with OmniVision OV48C products. OmniVision began to return from the secondary camera to the main camera position of the flagship phone, gradually turning from a spare tire to a regular one.

GalaxyCore and Shengbang Electronics also seized the opportunity. GalaxyCore mainly focused on the domestic low-end CMOS market. In 2010, domestic mobile phones emerged, and GalaxyCore developed rapidly as a supplier to Xiaomi, OPPO, vivo and Transsion. GalaxyCore also took advantage of this wave of domestic substitution dividends. For example, the rear secondary camera of Xiaomi 10 Youth Edition introduced GalaxyCore GC02M1 product.

In terms of quantity, GalaxyCore's shipments are very large, even exceeding Sony, with more than 100 million units sold in a month. However, due to its low unit price, it is sometimes classified as "Others" in the statistical chart of the market calculated by sales amount. It can be seen that in the field of image sensors, the high unit price brought by technological leadership is an overwhelming advantage.

But GalaxyCore is also trying to make changes. In July 2020, GalaxyCore applied for listing on the Science and Technology Innovation Board to raise 6.96 billion yuan. The money was mainly used for two things: building a 12-inch integrated circuit project to increase production capacity and developing higher-end CMOS products. It is hard to say whether GalaxyCore can shift from low-end to high-end, but it is already prepared to take on this wave of domestic substitution opportunities.

There is also Shengbang Co., Ltd., which makes power management chips. Its LED drivers, operational amplifiers, unlocking motor drivers and many other products are used in mobile electronics, automobiles and medical fields. Not only are all its products independently developed, but their performance is also comparable to that of foreign counterparts. There is a lot of room for domestic substitution.

It is precisely because Huawei has triggered this wave of domestic substitution that many domestic companies have been able to enter the high-end market. It is basically difficult for companies to enter the high-end market by relying solely on their own technological research and development to catch up with foreign manufacturers.

The development path of domestic electronic enterprises is mostly to start from the low end and then transform to the high end. However, there are not many successful cases of such growth paths that enterprises have come up with on their own accord.

To achieve a low-to-high transition, relying on the opportunities of the times provided by domestic high-end enterprises is a relatively feasible choice. OmniVision, GalaxyCore, and Shengbang Shares are all examples. But we must be aware that when opportunities come, do companies have the core capabilities to seize them? The key to all development is to be strong.

As more and more Chinese brands emerge, the electronics industry will grow stronger. Only by investing in the long term to master the basics can domestic companies better embark on the path of transforming from the division of labor chain to the value chain.

The full text is over. Thank you for your patience in reading.

[1] Core Characters Single Chapter, Aijiwei

[2] Will Semiconductor: Leader in image sensor chips, Nomura Orient International Securities

[3] Weier Technology: Riding on the trend of domestic substitution, CIS leader continues to grow, China Galaxy Securities

[4] Nobel Prize winner in Physics: The difficult road from theory to practice, Science and Technology Daily

[5] Weier Technology: Three driving forces boost the rise of CIS leader, Dongxing Securities

Latest articlesabout