Is there still a chance for GaAs in the RF PA field?-EEWORLD

Collect

Once you enter RF, your fate will be determined by the technology.

The wireless world is an infinite world. RF PA is born from radio and life is wonderful because of wireless communication.

There are many industry articles and discussions. GaN (gallium nitride) is here, SOI and SiGe RF front-end chips are increasingly used, does GaAs (gallium arsenide) still have a future? At the invitation, I had a discussion and exchange with Yole of France, data analysis and future predictions. I wrote this article to start a discussion. We welcome your comments and suggestions.

Digital chips are manufactured based on CMOS technology. The more advanced the process, the better. It depends on the positioning and application of the chip. Currently, the mainstream process for RF PA is GaAs. In addition to the mainstream, CMOS, SOI and SiGe processes are also used in RF PA.

Whatever process a domestic RF PA company chooses, it will determine the road it faces. If they choose GaAs process, it will be a difficult road, with everyone catching up and it will be difficult to widen the gap. If they choose other processes, it will be a risky road, with either success or failure.

RF PA company failed due to process selection

As Drucker said, a company has only two basic functions: innovation and marketing. Foreign chip companies are keen on innovation, while Chinese chip companies are keen on marketing. We often hear that there are no products that cannot be sold, only people who cannot sell.

After Avago, RFMD (Qorvo), and Skyworks emerged as GaAs PA manufacturers, many foreign startups have chosen to replace GaAs PA with CMOS PA. Only innovation can bring value.

In 2009, Skyworks acquired Axiom Microdevices. At that time, Axiom had already achieved tens of millions of CMOS PA shipments for 2G mobile phones.

In April 2013, Avago Technologies completed the acquisition of Javelin Semiconductor. RFMD acquired CMOS PA startup Amalfi.

In June 2014, Qualcomm announced the acquisition of BlackSand. Founded in 2005, BlackSand released the world's first 3G CMOS PA, and achieved an efficiency of 39%, making CMOS a reality in replacing GaAs. Its BST34 series power amplifiers are designed as replacements for existing 3G GaAs RF power amplifiers, and are fully functional and pin compatible. Switching from GaAs to CMOS can benefit from an improved supply chain and lower costs.

In the same year, Qualcomm's RF360 RF front-end solution was officially adopted by ZTE's flagship mobile phone Grand S II LTE. At that time, the market felt that in the future, CMOS PA would be popular in the mobile terminal market application of LTE and LTE-A carrier aggregation (CA) technology, subverting the RF front-end industry landscape.

The result? In 2017, Qualcomm launched a series of comprehensive radio frequency front-end (RFFE) solutions, including the first gallium arsenide (GaAs) multi-mode power amplifier (MMPA) module, declaring Qualcomm's CMOS PA road a complete failure.

In May 2016, Skyworks announced the acquisition of RFaxis. Founded in 2008, RFaxis is determined to become a pioneer in the next generation of wireless radio frequency chips. Previous comments by its executives can be found online: "Since the launch of the iPhone in 2007, everyone has been discussing how to simplify RF. Traditional wireless chips have always used gallium arsenide (GaAs), which leads to high costs and low production capacity. Silicon is God's gift to mankind. Almost the entire chip industry has received this gift, and the wireless communication field should get this gift as soon as possible. Just like the Internet of Things is the inevitable result of 'natural evolution', although gallium arsenide was developed by the US government at a cost of billions of dollars (for national defense), as people master more technology on how to apply silicon, the 'death of GaAs' is inevitable."

GaAs is not dead, but those foreign startups that made CMOS PA have disappeared. However, the spirit of innovation will not disappear, and the chip industry needs innovation. Why did foreign CMOS PA startups fail? Because the wheel of wireless communication technology is rolling forward, the requirements for RF PA technical indicators are getting higher and higher. It is difficult for design to make up for the shortcomings of process, but it is relatively easier for process to make up for the shortcomings of design.

Selecting a successful RF PA company based on its process

Domestic RF PA companies are concentrated in two tracks: mobile phone RF PA and WIFI RF PA. There are no less than 10 RF PA design companies in these two tracks. Because no one succeeds, everyone has a chance. Therefore, RF chip investment looks more like a gamble.

Acquisition or merger of domestic chip startups rarely happens. What awaits you is either success or failure, and there is no middle option of acquisition. Why? Because the vast majority of domestic chip startups have no innovation, no differentiation, and are following the path of repetition.

Among domestic RF PA companies, two have chosen the process differentiation route. One is Hantianxia and the other is Huizhiwei. Hantianxia is following the path that foreign companies have taken, while Huizhiwei is taking the uncharted territory that foreign companies have not taken.

The success of Handianxia lies in 2G PA, and the success of 2G PA lies in the choice of CMOS process. Although mobile phones have evolved from 2G to 3G, and then to 4G, and now into the 5G era. In recent years, the market demand for 2G PA has remained at around 80KKpcs/M. Handianxia can match GaAs PA in performance through CMOS 2G PA, and can crush 2G GaAs PA in cost, and has obtained more than 60% market share in one fell swoop. In 2016, Handianxia also won the "10th (2015) China Semiconductor Innovation Product and Technology Project Award" for 2G CMOS PA.

When talking about domestic RF PA technology, we cannot avoid Huizhiwei. A founder of an RF PA company once told me that domestic RF PA competition is excessive, and Huizhiwei will have an advantage in the long run. He is a senior expert in the field of GaAs RF and a friendly company. I was quite surprised when he said this. There is a saying that laymen watch the excitement, while experts watch the doorway. People in the market watch the market, people in technology watch the technology, and I am a product person. I like to look at technology and craftsmanship from the perspective of products.

Based on the fact that SOI's RF performance characteristics and cost are between CMOS and GaAs, Huizhi Micro has innovated the 4G PA architecture, using SOI for the first and second stages and GaAs for the third stage. Compared with the 4G PA that uses GaAs technology for all three stages, it has a cost advantage. However, doing new things is risky and requires continuous high-intensity R&D investment. Taking this route requires more effort than other domestic RF PA manufacturers that follow Skyworks. There is no one to learn from or imitate, and it all depends on one's own hard work and persistence. Once a product is successful, a technical barrier will be established.

Only after starting a business do you know the pain of starting a business, and only after innovating do you know the difficulty of innovation. It is difficult to have a chance to start a business by following the same path as others. To succeed, the only way is to innovate. Sanwuwei chose to take the pHEMT process route, and many people were puzzled and not optimistic. ESD was the first obstacle. The ESD of the switch made by SKY using pHEMT was only 120V, and Sanwuwei achieved 1000V, so that Hongkang suspected that the test was wrong after seeing the results after testing ESD. Because they didn't believe it, overseas customers took the Sanwuwei switch samples and re-tested ESD. It is not easy to make LDO and control circuits on pHEMT. R&D tests and technical breakthroughs are repeated over and over again in order to take a distinctive path. Each process has its own advantages. Taking its strengths and avoiding its weaknesses is to find market opportunities and define products that are highly compatible with processes and technologies.

Does GaAs technology still have a future?

Gallium arsenide (GaAs) is divided into three categories: HBT, pHEMT, and MESFET. Spectrum range: 1GHz to 100GHz, meeting low-frequency to high-frequency applications. In RF front-end chip products, RF PA uses HBT technology, and RF switch and LNA use pHEMT technology.

RF switch and LNA have turned to SOI process, which is the general trend. It is reported that SONY has closed the pHEMT process line. SiGe has taken a part of GaAs HBT market share, and there is a trend of expansion. CMOS has taken away the low-frequency GaAs HBT market. With the advent of 5G, Qorvo predicts that gallium arsenide below 8GHz will still be the mainstream, and gallium nitride above 8GHz will have a clear trend of replacement. Gallium nitride (GaN), as a wide bandgap semiconductor, has the characteristics of high power density, low energy consumption, suitability for high frequency, and support for wider bandwidth. International RF PA giants have invested huge amounts of money in GaN research.

GaAs HBT technology will maintain a certain proportion in the future due to its stability and good cost performance. As the mainstream technology of RF PA, its research and development work is still ongoing, which will surely expand the application space of GaAs HBT. GaAs pHEMT is also a very good technology. If it can be continuously developed and utilized, it will also have a good market prospect.

Based on the global consumption quantity and forecast of 6-inch GaAs wafers, French research organization Yole has made a preliminary statistics and classification. Due to the impact of the epidemic, global mobile phone shipments will decline in 2020, and the following data will be further updated.

Yole predicts that from 2019 to 2025, GaAs wafer 6" consumption will maintain a compound annual growth rate of 4%, and wafer demand will increase from 744,000 pieces/year to 941,000 pieces/year.

Among them, mobile phone PA has the largest demand for GaAs wafers, accounting for 57%, and the wafer consumption is expected to be 459,000 pieces in 2020 (the data will be revised downward). In addition to mobile phone PA, mobile phone WIFI PA and router WIFI PA consume 26.5% of GaAs wafers, and the wafer consumption is expected to be 212,000 pieces in 2020 (the data will be revised downward).

Although with the advent of the Internet of Things, for example, the demand for GaAs PA has increased for LTE Cat.1, overall, the main chip applications such as WIFI, Bluetooth, NB-IoT, Lora, etc. that integrate CMOS PA are the trend, and GaAs wafer consumption is negatively growing, with a predicted compound annual growth rate of -7%.

According to forecasts, the demand for GaAs PA and GaAs RF front-end chips for base stations will also have an annual growth rate of 7%, from 34,000 pieces in 2019 to 52,000 pieces in 2025.

Surprisingly, Yole predicts that automotive radar will almost abandon the GaAs process, decreasing from 6,900 units in 2018 to 800 units in 2025.

According to statistics from a representative wafer fab, the demand for gallium arsenide wafers from WIFI PA is not small. The data changes our previous understanding that the WIFI PA market is actually not a niche market.

In 2018, the wafer factory's WIFI PA accounted for 25-30%, only 10% less than the mobile phone PA. Mobile phones use gallium arsenide WIFI PA, and brands such as Apple, Samsung and Huawei all use gallium arsenide WIFI PA. Routers also use gallium arsenide WIFI PA.

In 2019, router main chips began to integrate 5.8G PA, and the demand for WIFI PA decreased. In terms of mobile phones, except for Apple, which continued to use Murata's GaAs WIFI PA, other mobile phone manufacturers have turned to SiGe process WIFI FEM, so the GaAs wafers required for WIFI PA decreased in the first two quarters of 2019. With the arrival of WIFI6, the demand for GaAs WIFI PA has increased significantly, and the corresponding GaAs wafer consumption accounted for 20-25% in Q3 2019, 15 percentage points less than mobile phone PA, which accounted for 35-40%.

Conclusion

Choices determine destiny, and environment shapes life. In the world of RF, technology is the choice we have to face. Foreign RF PA giants can have different teams working on different technology directions and running side by side; domestic RF PA startups have limited teams and funds and can only move forward in one technology. The choice of technology determines the fate of RF PA companies.

I am optimistic about every process, and every process has its own suitable products and markets. Don't expect every process to quickly become the mainstream of the future market, the future is in our unknown world. Big dreams lead small dreams, and small dreams help big dreams. Wafer factories are big dreams, and RF chip design companies are small dreams. As long as the wafer factory is determined in its direction, there will be chip design companies to follow; as long as there is soil for technology, the seeds of design will grow. A beautiful world is achieved between each other.

Keywords：GaAs Reference address：Is there still a chance for GaAs in the RF PA field?

Previous article：Academician Wang Jiangzhou: China leads in 5G equipment, while the United States is the leader in chips
Next article：Take a look at the true face of Huawei's visible light communication technology

Popular Resources
Popular amplifiers