RF technology and integration strategy of mobile phone RF components

JasonYoo

RF technology and integration strategy of mobile phone RF components [Copy link]

The integration of low-frequency and high-frequency RF wireless systems is very different. In the high-frequency band, CMOS technology is the preferred technology for RF circuits because it can achieve higher bandwidth than bipolar technology. Usually RF-CMOS is not integrated with digital CMOS on the same chip. The most important system in the low-frequency band is the cellular communication system. The focus of RF function integration in such systems is the integration of passive components. This article introduces the strategy of integrating passive components with RF active components through multi-chip packaging or modules.

The RF function plays an important role in the transmission of information between two points in a communication system. In such systems, the RF function is usually physically separated from other functions, and RF transmission and reception are generally implemented by different ICs. In order to reduce the size of the system and reduce costs, people continue to explore ways to integrate RF with other system functions, especially the development of DSP technology has a very important impact. In addition to this trend of RF and non-RF integration, there are other integration trends of RF devices themselves. These different trends are because different systems require different technologies to achieve the required RF functions. For example, before passing the received signal to the low noise amplifier (LNA), some systems require effective filtering of the signal, which requires the use of ceramic filters or surface acoustic wave (SAW) filters to filter the received signal, but these filters cannot be integrated into the receiver IC.

The difference between low frequency and high frequency systems

An important difference between low-frequency and high-frequency systems is that the latter can only achieve signal transmission when there is no obstruction between the transmitter and the receiver, while the low-frequency system does not have such a requirement and can therefore achieve a larger coverage area. There is no obvious dividing point between low-frequency and high-frequency. The transition frequency is between 2-5GHz and depends on system characteristics, such as transmitter output power and receiver sensitivity. This article uses 2.4GHz as the transition point between high and low frequencies. High-frequency systems can also be divided into long-range systems and short-range systems. Long-range systems such as radar, satellite links, base station links, fixed wireless broadband access (FWBA), etc., all require higher transmission power than short-range systems such as Bluetooth and 802.11a/b.

High frequency RF integration

The target market for short-range wireless communication systems is the consumer electronics market, which requires small size and low cost, and as the demand for applications that transmit video streams via data grows, data rates will continue to increase. These systems are basically portable battery-powered products that require long standby and talk time.

High-frequency systems (above 2.4 GHz) can achieve high bandwidth and moderate receiver selectivity because fewer transmitters operate at high frequencies. Similarly, the receiver's signal-to-noise ratio (S/N) is high, so the transmitter output power can be lower. For example, 802.11b has a bandwidth of 11 Mbps at 2.4 GHz, and 802.11a can reach up to 54 Mbps at 5 GHz. Using wider bands or more complex modulation methods requires more stringent signal linearity, which is closely related to the transmitter.

Figure 1 is a comparison of the operating frequency development that can be achieved by CMOS and BiCOS

The process technology used in the system is related to the achievable operating frequency. Figure 1 shows the development of the achievable operating frequency of CMOS and BiCOS. Assuming that fmax is directly related to the achievable operating frequency, it is obvious that CMOS is a better choice. In addition, CMOS can also meet the non-strict selectivity, signal-to-noise ratio and output power requirements, but the dynamic performance is reduced due to the low operating voltage. However, since many systems work in open frequency bands, there may be many transmitting devices interfering with each other between the transmitter and the receiver, such as microwave ovens interfering with Bluetooth communications is a typical example.

Although CMOS has these advantages at high frequencies, BiCMOS technology has the advantages of RF model and transistor parameter matching of bipolar technology, and BiCMOS design experience is richer. Size is not the main consideration in process selection, because the chip size of 0.18um CMOS or BiCMOS process to realize Bluetooth transceiver function is similar.

If CMOS technology is selected, standard digital CMOS will be the development trend. Since these digital CMOS already use multi-layer mask technology, no other options will be added. Digital functions will occupy the largest chip area, so the main cost will be generated in these digital function parts.

Does it still make sense to use mainstream CMOS technology to integrate digital circuits and RF functions on a single chip? This question needs to be considered from two aspects: from a technical point of view, it is possible to use standard CMOS modified for RF functions, such as high impedance substrate to reduce crosstalk through the substrate, and thick dielectric to achieve high quality factors of passive components; from an integration point of view, there is not much benefit in applying standard CMOS to RF and integrating digital and RF functions on one chip, because the models and libraries of digital and RF are fundamentally different. Digital circuits are often designed in VHDL/Verilog language, and digital libraries for CMOS technology are usually implemented before the emergence of new technologies. These digital libraries are used from generation to generation, so design engineers can perform digital designs before the next generation of processes are released.

For RF design, models and libraries are only possible after the emergence of technology, so RF devices have their own unique characteristics. Since there are generally no 1:1 reusable modules for RF functions, each new device must be developed almost from scratch. RF libraries are usually 1-2 years behind digital libraries, and using mainstream CMOS processes to implement RF functions means that the process will be one generation behind. Therefore, integrating digital and RF functions on a chip means that the digital functions will be implemented using the previous generation CMOS process, which is usually more expensive to implement. Moreover, passive components (inductors) and RF/analog functions cannot really develop in sync with CMOS process technology, so the area occupied by the RF part will increase with different technology generations relative to the digital part.

Other difficulties in integrating digital and RF functions on a single chip include:
1. The crosstalk between the digital and RF parts through the substrate must be controlled;

2. The mask cost of advanced CMOS process is very high, and the integration of digital and RF will inevitably lead to many design iterations due to RF design, which will increase the cost;

3. RF IC production is usually determined by design, while digital IC is determined by parameters. Therefore, the production of integrated circuits integrating digital and RF functions will be lower than that of digital ICs.

4. The high pin inductance generated by digital CMOS packaging will reduce RF efficiency.

Technically, the best solution for short-distance high-frequency systems is to use multi-chip packaging and modules, in which digital and RF functions are implemented using independent ICs and BiCMOS processes. These solutions are feasible for manufacturers that have both design capabilities and production packaging capabilities, but multi-chip packaging, especially modules, is not easy to implement for fabless companies that rely on foundries. Therefore, these companies will likely move towards integrating digital and RF functions on a single chip.

Figure 2 shows the functional block diagram of the CDMA RF front end.

The wireless system also requires antennas and switching devices for band selection, Tx-Rx switching, and antenna diversity, as shown in the CDMA RF front-end functional block diagram in Figure 2. In order to embed these devices, multi-chip packaging is usually used instead of module integration.

Low frequency integration

For applications below 2.4 GHz, cellular systems are the most widespread and important applications. Cellular phones require low cost and small size, and require higher integration. In addition, cellular systems have strict performance and cost requirements, and use a wide variety of components.

The receiver side of a cellular system requires high sensitivity and selectivity, which is generally achieved by using a receiving filter such as a surface acoustic wave (SAW) filter; a low noise amplifier (LNA) is used to achieve a large signal-to-noise ratio, where the inductor is used in the transmitter to achieve the best balance between noise and gain matching, and this LNA function is usually integrated on a single-chip transceiver IC; the baseband function is always implemented in a mainstream CMOS IC; the transceiver function traditionally uses BiCMOS technology, but CMOS technology is attracting more and more attention. At the same time, multi-band/system integration is also developing.

Another challenge is the transmit (TX) path, where these omnidirectional, non-point-to-point transmission systems require high output powers of 24-33dBm. The technology of choice for the power amplifier (PA) function is silicon (Si) bipolar or GaAs HBT (Si LDMOS) for ease of use, efficiency, and performance. After the final amplifier stage, a low-loss output matching circuit is required, as this circuit is technically difficult to integrate. This function is often partially integrated with discrete surface mount devices or implemented with special low-cost passive integrated (PI) chips.

The technologies used for low-frequency integration include GaAs HBT for PA, Si BiCMOS for PA driver, and Si PI chip for output matching bias stage and power control loop. Today's mobile phones are multi-band and multi-mode, requiring a lot of switching and filtering functions between PA, receiving channel and antenna. Switching devices are usually implemented using GaAs pHEMT or pin diodes and RF-MEMS. Duplex filters (RX-TX separated), diplex filters for band selection and harmonic filters constitute the passive front-end part of the antenna. The front-end integrated TX-FE module after the multi-band PA module.

After the passive front end, it’s all about the wireless module, which adds the transceiver functionality. It’s extremely challenging to integrate all of these technologies into a cellular system in a cost-effective manner. The transceiver functionality (including the LNA) can be implemented as a system-on-chip, but the receive filter still needs to be placed off-chip, and the PA and RF front end cannot usually be placed on a single chip. In general, the challenges come from passive components and multi-technology packaging, and the module integration on LTCC or organic substrates is generally chosen.

A key technology to reduce passive components and promote passive integration is PASSI technology. This technology can achieve capacitor accuracy of 145pF/mm2 and 4% (3σ), and the Q factor of the inductor exceeds 50. This technology can also serve as a platform for lateral integration of pin diodes, high-density capacitors and future MEMS variable capacitors and switches. Another related technological development is bulk acoustic wave (BAW) technology, which can replace ceramic and SAW technology in filters. BAW technology can be implemented in several ways, one of which is shown in Figure 4.

The advantages of using SAW technology are performance, loss, thermal characteristics, size and cost, especially at frequencies above 1GHz, SAW technology requires the use of submicron lithography. Due to the use of submicron structure, the loss of SAW filters will increase rapidly above 2GHz, but BAW technology can be applied at frequencies up to 10GHz at least. Due to the cost issues related to the addition of additional masks and qualified rates, the use of BAW technology on BiCMOS processes may not have too many advantages.

Outsourcing RF functions and complete system solutions is becoming a new business model. The front-end integration trend mentioned above will develop further. In the future, it will involve baseband and power control loops, matching, RF switching and filters, etc., to provide a complete RF system solution. When these functions are fully mature and OEM manufacturers accept this product, this complete system solution will be widely used. The development trend of front-end integration mentioned above will also extend to the baseband and power management fields.

Author: F. van Straten
Senior Director, RF Product Development
Philips Semiconductors
Email: freek.van.straten@philips.com