In recent years, the pursuit of various differentiated performance has made it no longer a novelty for system manufacturers to develop their own chips, and the domestic smartphone manufacturer OPPO has also become an important player in this field.

The semiconductor industry observation published an article last year titled "The Victory of DSA Behind OPPO's First Self-developed Chip," which detailed the performance of this emerging chip company in this field and the behind-the-scenes story of their first chip - the imaging-specific NPU Mariana® MariSilicon X. According to Jiang Bo, Senior Director of OPPO's Chip Product, the chip has now reached a shipment of tens of millions of units and will be used in more mobile phones in the foreseeable future.

"OPPO's self-developed chip aims to provide value to users, which is the most important and fundamental starting point for OPPO's chip development planning," Jiang Bo continued. Driven by this idea, OPPO brought their second self-developed chip - the flagship Bluetooth audio chip Mariana® MariSilicon Y - to the "OPPO Future Technology Conference" held recently.

Advertisement

From image to audio: OPPO's "core" pursuit

"Because OPPO is closer to consumers, it has a clearer understanding of user pain points and product value propositions, so it is easier for OPPO to make decisions when making some directional choices," Jiang Bo told reporters from the semiconductor industry observation at the company's first chip launch last year.

This is precisely why OPPO chose computational imaging as the first landing scenario for the company's self-developed chip and launched MariSilicon X. Based on the company's advantage of being closer to customers than traditional chip companies, OPPO has chosen the audio market as the second focus for the company's chip.

As everyone can see, if you want to find new highlights for smartphones in the past few years, audio is undoubtedly one that cannot be ignored. Because, driven by Apple's AirPods and AirPods Pro, the entire industry chain, starting from the chip upstream, has undergone multiple rounds of innovation around the popular TWS. In addition to sound quality, many features including computational audio (currently represented by active noise cancellation and spatial audio) have also become the main demands of consumers for audio products.

The "2022 Audio Product Usage Status Research Report" also shows that more than 58% of respondents said they hope to obtain high-resolution or lossless sound quality. 61% of consumers hope to obtain better sound quality than MP3, including CD sound quality (44.1kHz/16bit), high-resolution sound quality (above 96kHz/24bit), and lossless sound quality. In addition, 66% of consumers believe that lossless sound quality will determine whether they purchase equipment.

In terms of computational audio, as the consumer market matures and technology develops, it has also entered the second stage. The biggest feature of this stage is personalization.According to the introduction, the traditional rendering of virtual space is based on a set of preset models, similar to buying ready-made suits. Although exquisite, there may still be some details that do not fit well. The new generation of spatial audio can customize the HRTF model (Head Related Transfer Function) according to the user's head and ear shape, to obtain the most personal listening experience, just like a high-end custom-made suit, every detail is tailored to the user's unique figure. With the integration of many advanced technologies such as AI technology, environmental perception, and natural language processing into the field of computational audio, a more personalized sound experience is imminent.

Specifically, for the current popular Bluetooth audio, there are several issues:

On one hand, the transmission rate is not the biggest advantage of this wireless connection technology. Therefore, even in the latest Bluetooth 5.3 standard, EDR (Enhanced Data Rate) supports a maximum rate of 3Mbps. However, in real-world applications, due to the impact of complex factors such as the actual signal environment, the theoretical maximum rate of 3Mbps can often only be half, that is, an actual rate of 1.5Mbps. This level of speed cannot carry the data of lossless audio transmission.

On the other hand, there are many mainstream encoding and decoding protocols on the market, and common formats such as AAC and LDAC are all lossy encoding. In the lossless encoding format, the compression rate of FLAC is 70%, the compression rate of ALAC is also 70%, and the better-performing L2HC has a compression rate of 60%. That is to say, the existing public lossless encoding and decoding schemes cannot meet the volume requirements of wireless transmission, so a stronger compression scheme is needed.

Moreover, in reality, the complex signal environment has a great impact on the transmission itself. When the signal is too complex, such as when the phone is in the pocket, the Bluetooth signal will be unstable.

Faced with the above problems, OPPO launched the company's second self-developed chip - Mariana® MariSilicon Y. According to Jiang Bo, the reason for defining such a Bluetooth audio chip is not only based on market demand but also has another layer of meaning, that is, OPPO believes that end-to-end connection technology is a particularly important basic technology in hardware capabilities.

"OPPO emphasizes the integration of all things. In the future, we will definitely explore and improve a lot in short-range communication and end-to-end experience. Although this chip is small in size, it is the first time we have achieved connection end-to-end through a private protocol. This connection's basic capability is something OPPO must establish." Jiang Bo told reporters.

He further pointed out that as a system manufacturer, OPPO has an advantage that traditional chip manufacturers do not have - a deeper understanding of user scenarios. Therefore, when the company invests in a technical direction, it is easier to know why to do this, and thus can clearly know the priority when making some performance trade-offs.

With the company's team's deep accumulation and investment in this area, a fully-equipped Mariana® MariSilicon Y was born.From Co-Processor to SoC: Technology Enters Deep Waters

As OPPO stated, the company's first self-developed co-processor chip, built on the DSA concept, requires pairing with a main platform chip to function. The company's MariSilicon Y, as OPPO's first SoC chip solution, can fully handle all the functions of a Bluetooth audio device. This not only signifies OPPO's first foray into the new field of connectivity chip design, equipped with a complete set of software and hardware capabilities for Bluetooth connectivity. It also means that OPPO has, for the first time, the capability to design a Bluetooth SoC platform with both computing and connectivity capabilities.

Jiang Bo summarized that the MariSilicon Y has three leading technologies, which are 192kHz/24bit ultra-high-definition lossless audio, 590 GOPS powerful computing power, and the advanced N6RF process.

It is reported that through the pioneering ultra-high-speed Bluetooth solution, the MariSilicon Y integrates an industry-leading 12Mbps Bluetooth rate that is 400% faster than traditional Bluetooth, coupled with OPPO's self-developed URLC lossless codec, achieving for the first time the wireless transmission of 192kHz/24bit lossless audio, freeing perfect sound quality from the constraints of wires.

"The advanced 12Mbps Bluetooth rate of the MariSilicon Y not only easily handles the data volume of 192kHz/24bit lossless audio but also has ample rate redundancy to carry data retransmission, system overhead, and signal control, significantly improving the overall Bluetooth connection quality," emphasized Jiang Bo.

As for the URLC (Ultra-Resolution Lossless Codec), it is OPPO's self-developed audio codec technology that, for the first time, raises the lossless compression rate to 50%. This means that the same lossless audio content can be compressed to a smaller volume for transmission. The URLC codec protocol, in conjunction with the design of the 12Mbps ultra-high-speed Bluetooth, truly integrates physical hardware and software codec technology to serve the transmission of lossless audio. At the same time, URLC is also one of the only two codecs on the market that support 192kHz/24bit. This is also the highest specification of audio quality in wireless Bluetooth earphones (other Bluetooth earphones currently only support up to 48/24 lossless).

After appreciating its leading audio bandwidth and decoding capabilities, the powerful capabilities of this chip also provide more possibilities for future audio applications.

Jiang Bo introduced that as the first Bluetooth audio SoC to integrate an NPU unit, the MariSilicon Y's high-performance NPU unit has reached an advanced 590 GOPS computing power. In comparison, the high-performance DSP computing power integrated in the MariSilicon Y is 25 GOPS, which is also one of the strongest DSPs in the industry (the computing power of the best-selling earphone chip in the world is 9 GOPS).

It is understood that this 590 GOPS high-performance NPU has three major advantages: high energy efficiency (compared to traditional DSP units, the NPU is more efficient when running AI algorithms, achieving more with less effort), independent computing power (the independent NPU design does not rely on mobile phone computing power. Calculations can be completed in real-time within the audio device, reducing latency caused by data transmission between the earphones and the mobile phone), and high compatibility (the high computing power of 590 GOPS can continue to provide ample support for the rapid development of AI audio applications in the future, and can continue to provide upgradable functions).From Jiang Bo's introduction, we learned that the current AI processing in the audio field is essentially handled by DSP, with relatively limited computing power or network complexity. The algorithms based on DSP are also very popular at present, and existing products should have good support for this. However, when the complexity of algorithms increases and higher processing capabilities for audio are required in the future, the combination of DSP and NPU may become a mainstream trend, and OPPO is one step ahead in this regard.

"Superb computing power combined with advanced AI algorithms, Mariana® MariSilicon Y is expected to create new computing value," said Jiang Bo. In his view, the reason why this chip can achieve such excellent performance is not only because of its unique design but also because the N6RF process is the most important determining factor.

Data shows that N6RF is TSMC's latest advanced RF CMOS semiconductor technology, which enables customers to maximize battery life using its outstanding RF functions and digital PPA advantages. Compared with TSMC's own 16FFC RF technology, N6RF offers significant improvements.

According to Jiang Bo, currently, only Apple's H2 chip, the GPS module in Apple's S8 chip, and OPPO's Mariana® MariSilicon Y chip have adopted TSMC's N6RF advanced process technology worldwide. In other words, Mariana® MariSilicon Y has become one of the most advanced manufacturing process Bluetooth SoC chips to date, leading the industry by at least two generations.

With the three major technologies of this chip, OPPO has respectively solved the problems of high-resolution lossless audio transmission, insufficient terminal computing power, and chip performance power optimization, providing new guidance for the industry and offering consumers better content and product experiences. Based on its powerful NPU, Mariana® MariSilicon Y has finally achieved sound separation technology on the audio terminal for the first time. Using this technology, OPPO's new chip can provide users with the next generation of personalized (such as custom panoramic sound, universal panoramic sound) listening solutions.

Because this chip has extremely high compatibility, it can support Bluetooth 5.3 and LE Audio's LC3 codec, which can be well compatible with all next-generation devices that support LE Audio; at the same time, Mariana® MariSilicon Y also supports LHDC, LDAC high-definition codec, and is compatible with traditional OPPO products and other brand products, still obtaining excellent sound quality; in addition, Mariana® MariSilicon Y is also compatible with traditional SBC and AAC codecs, which can achieve full coverage of traditional Bluetooth devices.

This means that no matter what device you use, you can get excellent sound quality with Mariana® MariSilicon Y.

"Making chips, never hoping for miracles"

Recently, an internal speech by OPPO founder and CEO Chen Mingyong was reported by the media. In this speech, he first revealed the two original intentions of OPPO making chips: one is to provide users with a better experience; the second is to have their own technical moat."We make chips, never hoping for miracles. If miracles could be achieved overnight, anyone could do it. It is precisely because good chips are difficult to make that we can form a long-term advantage in user experience by making them. This process is not easy, it takes time, and there will be a lot of doubts along the way, but we must maintain a normal mind, respect objective laws, and take solid steps one at a time. At the same time, maintain long-term optimism and short-term caution. No matter what the future holds, OPPO's chip-making is very meaningful," Chen Mingyong emphasized in his speech.

Jiang Bo also told reporters that OPPO's self-developed chips are not intended to replace anyone. The company insists on continuous exploration, hoping to bring more valuable things to everyone. OPPO has now established a chip team of more than 2,000 people, with the mentality of sharpening a sword for ten years to make chips.

In his view, the evolution from MariSilicon X to MariSilicon Y not only reflects the in-depth understanding of terminal applications by the company's chip team but also allows chip team members to accumulate more technology, making them better prepared for the launch of the next generation of chips.

Taking the N6RF process adopted on the Mariana® MariSilicon Y as an example, on the basis of the previous 6nm, OPPO has further expanded to a more advanced process, enabling the company to have expertise in chip design under advanced processes. This process and chip also represent OPPO's mastery of applying advanced process technology to the RF system, giving the company the confidence to create more possibilities in the future.

"Developing this chip is an exploration of RF, which is a very important starting point for us. The combination of the baseband part, digital part, and RF part of this chip does bring some design challenges, but this challenge is exactly what we need to overcome when making other chips in the future, and we have overcome it well," Jiang Bo said.

He also pointed out that starting from this chip, OPPO has expanded the vertical integration field centered on self-developed chips. On the basis of the image integration of the previous generation of self-developed chips, it has added the integration of Bluetooth connection, laying the necessary underlying connection capabilities for the integration of all things.

"We always maintain a gradual and normal mind, respect objective laws on the road of self-developed chips, isolate external pressures, do not fantasize about overtaking on the bend, and firmly believe in long-termism, and are prepared to sharpen a sword for ten years. We don't need to consider costs like traditional chip manufacturers, our self-developed chips are not limited by revenue pressure, and we can even invest in technology research and development and chip stacking without considering the cost, to create real surprises and value for users," Jiang Bo emphasized.

In their view, this is the starting point of OPPO's self-developed chips and also the best endpoint.