Advances in Base Station Receiver Integration

Faced with the continued rise in demand for mobile phone data services, base station designers are constantly being forced to increase bandwidth and reduce costs. Many factors affect the overall cost of installing and operating additional base stations to meet the growing demand. In addition to helping reduce initial costs, using smaller, lower-power electronic components in macrocell base stations also helps reduce the current land and building lease costs and power consumption of base station towers. New architectures such as remote radio heads (RRHs) have the potential to further reduce costs. Very small microcells and femtocell base stations extend data services to areas not covered by larger macrocell base stations. To achieve these benefits, base station designers need new components with extremely high levels of integration, but they cannot sacrifice performance.

Integration of the RF portion of the radio circuit is particularly challenging due to performance requirements. More than 10 years ago, a typical base station architecture required several circuit stages, including low-noise amplification, down-conversion to an intermediate frequency (IF), filtering, and further amplification. Today, higher performance mixers, amplifiers, and higher dynamic range analog-to-digital converters (ADCs) with higher sampling rates allow designers to eliminate the down-conversion stage (integrated into a single IF stage). However, component integration is still somewhat limited. Mixers provide buffered IF outputs, integrated baluns, LO switches, and dividers. Devices with a mixer and a PLL for the LO represent the latest level of integration. Devices with dual mixers and dual amplifiers are available. Until now, there have been no devices that integrate any part of the RF chain with the ADC on the same chip. This is primarily because each component must be produced using a unique semiconductor process. For the application, the performance trade-offs resulting from choosing a common process are unacceptable.

At the same time, mobile phone RF circuits have evolved into highly integrated baseband and transceiver ICs and integrated RF front-end modules (FEMs). The RF functional blocks between the transceiver and the antenna include filtering, amplification, and switching (and impedance matching functional blocks between components when necessary). The transceiver integrates the receive ADC, transmit DAC, and associated RF functional blocks. Here, the performance requirements are at a level that allows the use of a common process. FEMs use system-in-package (SiP) technology to integrate different ICs and passive components, including multi-mode filters and RF switches for transmission and reception. Although a common process is not suitable here, integration is still required.

The performance requirements of RF/IF, ADC, and DAC components in picocell and femtocell base stations are often much lower than those of macrocell base stations due to the lack of coverage, power output, and number of users per service area. In some cases, modified versions of cell phone components can be used in picocell or femtocell base stations and provide the necessary integration, low power, and low cost. Here, a common semiconductor process provides sufficient performance levels for all functional blocks in the signal chain. How can this level of integration be achieved in macrocell base stations?

The traditional approach to integration is to integrate more and more functions on a monolithic chip, often in "groups". Smaller geometry semiconductor processes become feasible, production-worthy, and produce sufficient performance for base station applications. The resulting power reduction, often by an order of magnitude, enables the integration of many previously separate functional components. This allows the integration of many new functions, and soon a group of highly integrated products will be available. In the mobile phone field, this happened with the advent of baseband and transceiver ICs. There is often a long lag between two major process geometry improvements. When the traditional approach to integration is blocked, other alternatives emerge.

Just as in cell phones (i.e., monolithic integration continues all the way to the RF front end, with modules providing the rest of the integration), a new class of modules is providing a new level of integration for base station applications. A recent example is the LTM9004 and LTM9005 μModule® receivers from Linear Technology, which integrate a high-speed ADC with the RF signal chain. The LTM9004 uses a direct-conversion architecture with an I/Q demodulator, low-pass filters, and a dual-channel ADC (Figure 1). The LTM9005 uses an IF sampling architecture with a downconverting mixer, SAW filters, and a single-channel ADC (Figure 2). Both devices are available in a 22mm x 15mm LGA package, which reduces board space by about 75% while integrating multiple ICs and dozens of passive components. This level of integration is not possible using traditional methods because the high-speed ADC uses fine-line CMOS processes, while the RF components use silicon germanium (SiGe) processes. Furthermore, many passive components simply cannot be integrated into silicon due to performance requirements.

Figure 1: Direct conversion architecture implemented in the LTM9004 μModule receiver

Figure 2: IF sampling architecture implemented in the LTM9005 μModule receiver

Why is this level of integration best suited to SiP technology? The IF sampling architecture provides a clear argument for this. Each component has unique and demanding requirements and must rely on an optimized manufacturing process. This has resulted in the current situation where there is no universal process. More importantly, the signal filtering at the IF frequency band must be very good to ensure out-of-band suppression of interfering signals that may affect the performance of the base station. Today, this filtering is accomplished using surface acoustic wave (SAW) filters in hermetic ceramic packages. Such filters are integrated in the LTM9005.

The LTM9005-AB can be used with an RF front end to build a complete UMTS band uplink receiver. The minimum performance of the receiver is detailed in the 3GPP TS25.104 V7.4.0 specification (Medium coverage area base station in operating band I, 4 carriers). The RF front end consists of a ceramic duplexer and one or more low noise amplifiers and ceramic bandpass filters. Here is an example of typical performance of such a front end:

 •Receive frequency range: 1920 to 1980MHz
 •RF gain: 17dB (max)
 •AGC range: 20dB
 •Noise figure: 1.6dB
 •IIP2: +50dBm
 •IIP3: 0dBm
 •P1dB: -9.5dBm
 •Rejection (at 20 MHz): 2dB
 •Rejection (in transmit band): 95dB

Although not as common as the IF sampling architecture, the direct conversion receiver architecture has some advantages over the traditional superheterodyne structure and offers the greatest potential for ultimate monolithic integration. Since it is not susceptible to image signals, the requirements for RF front-end bandpass filtering are relaxed. The RF bandpass filter only needs to attenuate strong out-of-band signals to prevent them from overloading the front end. In addition, direct conversion eliminates the need for additional IF amplifiers and bandpass filters, and directly converts the RF input signal to baseband.

However, direct conversion does present its own set of implementation issues. Since the receive LO signal is at the same frequency as the RF signal, it can easily radiate from the antenna and violate relevant regulatory standards. In any case, the LTM9004-AC can be used with an RF front end to form a similar UMTS band uplink receiver using the 3GPP TS25.104 V7.4.0 specification discussed earlier (medium coverage area base station in operating band I, 4 carriers). Again, the RF front end consists of a duplexer and one or more low noise amplifiers (LNAs) and bandpass filters. In this case, variable adjustments of the automatic gain control (AGC) are placed in the RF domain to minimize the gain or phase deviation between the in-phase and quadrature channels. As before, the typical performance of such a front end also meets the 3GPP standard:
• Receive frequency range: 1920 to 1980 MHz
 • RF gain: 23.5dB (maximum)
 • AGC range: 20dB
 • Noise figure: 1.6dB
 • IIP2: 50dBm
 • IIP3: 0dBm
 • P1dB: -9.5dBm
 • Suppression (at 20 MHz): 2dB
 • Suppression (in the transmit band): 96dB

Second-order nonlinearity in the receiver can also create unwanted baseband signals. Any frequency tone entering the receiver will cause a DC offset in the baseband circuitry. In addition, second-order nonlinearity in the receiver can allow a modulated signal (even a desired signal) to produce a pseudo-random energy block centered around DC. However, it is not only suitable for many applications today, but also has a very good future due to many reasons, including the great potential for integration.

Conclusion
It is not always possible to achieve significant integration using a traditional monolithic approach, especially when performance requirements are high. When semiconductor processes are incompatible, it is still possible to integrate functional components without sacrificing performance. The μModule receiver proves that macro base station performance can be achieved in a fully integrated and compact package. Perhaps in time, performance limitations will be overcome, allowing all functional components in the signal chain to use a common semiconductor process. Until that day comes, the pressure to integrate will continue, and SiP technology can provide clear advantages in performance and form factor.