Nikkei Electronics Asia -- June 2009
Insights
Evolution of 4G Networks Using Remote Radio Heads

Distributed base stations with remote radio head (RRH) capability assist mobile operators in resolving cost, performance, and efficiency challenges when deploying new base stations that are rapidly moving toward fully-developed 4G networks. Multi-mode radios capable of operating according to GSM, HSPA, LTE, and WiMAX standards and advanced software configurability are key features in the deployment of more flexible and energy-efficient radio networks. This article describes the key market and technology requirements for RRHs and how WiMAX/LTE RRH and intellectual property (IP) core solutions, combined with the latest FPGA technology help enable compact, environmentally-friendly/green, and full-featured applications for mobile network solutions.

Wireless and mobile network operators have been facing the continuing challenge of building networks that effectively manage high data-traffic growth rates. Mobility and increasing levels of multimedia content demanded by end users require end-to-end network adaptations that support both new services and the increased demand for broadband and flat-rate Internet access. In addition, network operators must consider the most cost-effective evolution of the networks towards 4G.

Wireless and mobile technology standards are evolving towards higher bandwidth requirements for both peak rates and cell-throughput growth. The latest standards supporting these are HSPA+, WiMAX, and LTE. The network upgrades required to deploy networks based on these standards must balance the limited availability of new spectrum, leverage existing spectrum, and ensure operation of all desired standards.

Distributed open base station architecture concepts have evolved in parallel with the evolution of the standards to provide a flexible, cheaper, and more scalable modular environment for managing the radio access evolution. For example, the Open Base Station Architecture Initiative (OBSAI) and the Common Public Radio Interface (CPRI) standards introduced standardized interfaces separating the base station server and the RRH portion of a base station by an optical fiber.

RRH System Requirements

The RRH concept constitutes a fundamental part of an advanced base station architecture. RRH-based system implementation is driven by the need to reduce both CAPEX and OPEX consistently, which allows a more optimized, energy-efficient, and greener base deployment. Fig 1 illustrates an architecture where a 2G/3G/4G base station is connected to RRHs over optical fibers. Either CPRI or OBSAI can be used to carry RF data to the RRH to cover a three-sector cell.

The most recent OFDMA-based standards - WiMAX and LTE - include 20MHz wideband radio channels for both time division duplexing (TDD) and frequency division duplexing (FDD) operational modes. The throughput is enhanced by multiple-in/multiple-out (MIMO) antenna techniques. Both 2x2 and 4x4 MIMO systems are attractive solutions which meet the small footprint requirements of the radio module.

Multi-Rate DSP Applications

Supporting multiple radio channel bandwidths per channel using only pure software reconfiguration is highly desirable, as is supporting multiple channels concurrently in a multi-carrier system fashion. To achieve this, a designer must choose between using a different clock domain for each rate and using one single-clock domain with software flexibility. The single-clock domain technique is more elegant and takes advantage of advanced digital signal processing (DSP) techniques for converting each rate into a common system rate. Together these techniques are called sample rate conversion (SRC), and when combined with multi-channel digital upconversion (DUC) and digital downconversion (DDC), multiple sample rates to be processed concurrently in the same application and in the same clock domain of the digital-to-analog converter (DAC) or analog-to-digital converter (ADC) are enabled. This method allows flexibility by supporting both single-clock and multiple-clock techniques simultaneously.

Crest Factor Reduction

An OFDM signal has a large peak-to-average envelope power ratio, which results in significant distortion when passed through a non-linear device such as a power amplifier (PA).

The objective of the crest factor reduction (CFR) technique is to reduce the peaks of the OFDM signal to a satisfactory level to ensure better usage of the PA. Due to the signal's reduced dynamic range, it is possible to operate at higher average power and hence closer to the saturation point of the PA when combined with digital pre-distortion (DPD). An effective CFR technique is the peak windowing approach, where power peaks exceeding a certain threshold are clipped and smoothed by filtering. The filtering mitigates unwanted out-of-band spectral products caused by the sharp signal edges of conventional clipping. Peak windowing approaches may offer several decibels of improvement, with 3 to 7dBs of improvement to the complementary cumulative distribution function (CCDF) demonstrated.

Major considerations are the window and filter selection, as well as the output power-clipping threshold. It is necessary to look for the trade-off between acceptable error vector magnitude (EVM) levels and extra dBs of linear improvement obtained via CFR. The EVM requirement for WiMAX is as low as 2.8%, for LTE it is 8% for 64 QAM modulation.

Peak windowing works in the time domain and uses less logical resource than spectral analysis. A spectral analysis uses the more costly fast Fourier transform (FFT) and inverse FFT (IFFT) functions to benefit from working directly in the spectrum. Peak windowing uses around one-tenth of the hardware signal processing resources compared to spectral analysis.

CFR techniques often are combined with a DPD block to achieve high performance in the RF PA. The CFR block first cancels the peaks, thereby preventing the high peak signal from driving the PA into the non-linear operation region of the RF PA. In contrast, the DPD block extends the linear operation region and the RF PA operates close to the saturation levels.

Digital Pre-Distortion

Aligning a non-linear device such as a transmitter PA is a challenge. With DPD, the PA operates at its highest efficiency region resulting in higher transmission power levels. Pre-distortion techniques have been developed for different domains of the transmission chain between baseband and the RF/analog domains, each of which has a number of advantages and restrictions. An effective DPD compensates for the non-linearity of the transmitter PA, also taking into consideration any transistor memory effects and system temperature variations.

DPD techniques can be classified into two categories, feed-forward and feedback. In feed-forward techniques, a static look-up table (LUT) is calculated and stored in FPGA memory. It is not updated at any point of the system's operation.

This method is simple but lacks flexibility due to the absence of a feedback path. Once a feedback path is available, the LUT can be updated and adaptation algorithms can monitor and adjust the table entries.

FPGA-Based RF/RRH SoC

Feedback techniques shown in Fig 2 are suitable for developing modules in a low footprint compared to RF pre-distortion techniques, which require hardware changes and additional RF components. Moreover, they are suitable for previously built systems because the feedback path RF-ADC-FPGA exists in most SDR implementations. 

For feedback DPD, the typical choices are a table, a polynomial approach, and a mixed approach. The DPD corrects the nonlinearity in the PA, which causes intermodulation distortion and leads to spectral re-growth. DPD also enables efficient use of the PA transistors since it extends the linear region, taking advantage of the normally prohibited higher power nonlinear region. A mixed approach using both polynomial and LUT methods is proposed. In implementation, the approximation that a polynomial can achieve is better suited to deal with wideband PAs due to memory effects. The effectiveness of the polynomial approach is limited by its order. Increasing the polynomial order also increases complexity and hardware usage. For fine resolution, when LUT sizes grow, and the adaptation time slows. Typically, the LUT approach has been used in narrowband memory for less nonlinear solutions. However, the limited range of a LUT can be useful, especially for the higher output power regions that must achieve the highest levels of linearity. For certain parts of the nonlinear transfer function, it is possible that a polynomial of a rather high degree is needed. The benefits of both techniques are exploited by combining the LUT in which the polynomial linearity reaches its correction limit due to lack of orders. The hardware resources of memory and multipliers therefore are kept as low as possible for a high degree of linearity. These techniques are used to correct both amplitude and phase of baseband input IQ samples. 

Transceiver-based FPGAs (Fig 3), combined with software tools and IP, offer a variety of SoC implementation options for RRH applications. With channel bandwidth up to 20MHz, MIMO antenna configurations, and multi-mode support, implementation of the signal processing functionality in the RRH requires a highly integrated, flexible, and power-efficient silicon platform.

By Christian Lanzani,
co-founder and Senior Product Manager,
Radiocomp ApS; and Deepak Boppana, Strategic Marketing Manager,
Altera Corp

MOST POPULAR ARTICLES

24 Hours | 7 Days