LTE-A Release 12 transmitter architecture: analog integration
Part 1 was on Technology Evolution and examined market forces driving global adoption of the LTE standard and trends in fourth-generation (4G) radio access technology.
This article explores the analog integration challenges in 4G base stations. Rel-12 features, such as wideband downlink CA, downlink multiple-input multiple-out (MIMO) spatial multiplexing, and AAS with embedded RF, present new design challenges in next-generation eNodeB radios. A bits-to-RF solution can help engineers shape alternative radio transmitter architectures (an example is given). The discussion focuses on novel RF digital-to-analog converter (RF-DAC) technology that yields a single-chip, wideband RF transmitter. Topics include system-level applications of RF-DAC and the integration benefits that it delivers to eNodeB radio design.
LTE is recognized as the fastest growing mobile broadband technology and becoming the most widely adopted cellular standard. The popularity of LTE is mainly due to its high spectral efficiency and high peak data rates, low-latency IP-based network, and evolutionary roadmap. But LTE is not “true 4G” service and is technically still considered 3.9G.
The “true 4G” radio communication standard, known as International Mobile Telecommunications-Advanced (IMT-Advanced), must meet the requirements set forth by the International Telecommunication Union Radio Sector (ITU-R). IMT-Advanced defines 4G as a service that delivers 100Mbps peak data rates for high-mobility users, and 1Gbps peak data rates for low-mobility clients. To comply with the IMT-Advanced vision, the 3GPP has developed many enhancements since the initial LTE Rel-8 standard published in 2008.
In Rel-10 the 3GPP introduced LTE-Advanced as “true 4G” service to meet or exceed the IMT-Advanced requirements. Presently, Rel-12 is close to introduction with a functional freeze date planned for March 2015. Figure 1 illustrates LTE development timelines, where it can be seen that theoretical peak downlink (DL) and uplink (UL) data rates have increased about 10x and 20x, respectively, from DL = 300Mbps/UL = 75Mbps in Rel-8 to DL = 3Gbps/UL = 1.5Gbps in Rel-10. The extraordinary increase in peak data rates is due, in part, to wideband CA complimented by multilayer spatial multiplexing introduced in Rel-10 and now an important part of Rel-12 enhancements.
Rel-12 enhancements will significantly impact how evolved NodeB (eNodeB) radios are designed. Some of the important Rel-12 items include new combinations of carrier aggregation, spatial multiplexing enhancements with downlink MIMO, and RF requirements needed in AAS. Figure 2 summarizes some of the Rel-12 items with respective features and benefits. The Rel-12 feature enhancements bring many benefits to the LTE ecosystem, along with new radio design and radio architecture challenges. (For more detailed background information, see Part 1).
The RF Transmitter Integration Challenge in Macro Base Stations
In 4G macro cell base stations the adoption of downlink MIMO with multilayer spatial multiplexing or the deployment of AAS with embedded RF can increase transmitter channel density up to a factor of eight or sixteen, respectively, when compared to 2G/3G base stations. In the near future with multicolumn 3D antenna arrays, the radio channel count might increase to 32 or more per sector. Compounding the eNodeB channel density trend is a requirement for wideband frequency-agile transmitters. A common hardware eNodeB radio platform must support LTE band coverage from 450MHz to 2.2GHz with a roadmap for Band-7/Band-41 coverage, and support up to 100MHz of CA bandwidth. Clearly, a technology disruption is needed to facilitate integration of multichannel, high-performance radio transmitters into space-constrained, power-limited, and cost-sensitive applications like remote radio units, integrated antenna radios, AAS, and conventional base transceiver station (BTS) line cards.
The RF DAC Transmitter
Direct digital-to-RF conversion with a high-speed RF DAC is a technology disruption for eNode transmitters. In such a device, the RF-DAC partition uses direct digital synthesis (DDS) to move the quadrature modulator, agile local oscillator, and signal-filtering analog functions into the digital domain (see Figure 3). The RF DAC with DDS partition capitalizes on the fact that digital processes scale better than analog in terms of lower power consumption, faster speed, smaller die area, and lower cost. However, Moore’s Law can only be realized if direct signal conversion, from the digital domain to analog domain, is achievable. The RF DAC is the enabling technology that makes this possible because it bridges the digital-to-analog domain. An RF DAC is generally characterized as a mixed-signal device that operates in multiple Nyquist zones with conversion rates above 1.5Gsps to perform direct bits-to-RF signal synthesis. The RF DAC shown in Figure 3 synthesizes output signals of at least 500MHz signal bandwidth at carrier frequencies of 2.0GHz or higher.
Benefits of the RF DAC Transmitter
Compared to conventional RF transmitter architectures (see Figure 4) like zero intermediate frequency (zero-IF), complex-IF and real-IF, an RF-DAC architecture occupies less printed circuit board (PCB) area with fewer bill-of-material (BOM) components. It operates at lower power and can deliver excellent dynamic performance.
In terms of RF performance, an RF DAC has significant benefits versus other topologies. The digital upconversion (DUC) with digital filtering implemented in DDS eliminates gain-phase errors and achieves perfect carrier suppression with no LO leakage. The result is excellent EVM performance when transmitting high-order modulation like QAM64. The quadrature NCO makes the RF DAC an agile transmitter capable of tuning across the entire spectrum of LTE bands. For example, a single device can synthesize multicarrier, multiband, and multistandard signals including GSM, WDMA, and LTE.
Another benefit with direct-conversion RF-DAC technology is that it allows engineers to use a lower cost digital predistortion (DPD) observation receiver. Macro cell base stations use DPD techniques for RF power amplifier (PA) linearization. This requires a PA observation receiver channel, like that shown in Figure 3, to monitor the PA output. The observation receiver detects PA distortion products and works with a predistorter to compensate for intermodulation and adjacent-channel leakage power. Typically, DPD bandwidth expansion requires the DPD observation receiver bandwidth to be five times the data bandwidth. In 100MHz CA applications, this means that the DPD bandwidth must be at least 500MHz. Also, the observation receiver cannot add impairments to the observed signal because it cannot be discerned from the main TX path impairments. Consequently, the DPD observation path must have excellent linearity (which adds cost and circuit complexity). Conversely, if the main TX path has negligible impairments, then the DPD path impairments can be corrected.
Recall that the RF-DAC does not introduce gain or phase errors, thus resulting in negligible TX path impairments. Therefore, a low-cost and lower performance DPD receiver like a zero intermediate frequency (ZIF) receiver can be used. There are three reasons why the ZIF architecture is lower cost compared to high-IF or direct-RF sampling: (1) quadrature demodulation enables a lower conversion rate, baseband-sampling, dual-channel high-speed ADC because it only needs to quantize one-half the DPD expansion bandwidth; (2) the ADC samples baseband signals, not high-IF or direct RF, which means that the ADC does not need pico or femto second aperture jitter; and (3) the baseband I/Q anti-alias filters are lower cost and easier to design compared to IF or RF filters. In summary, an RF-DAC transmitter relaxes DPD receiver signal-path performance requirements, thus reducing system cost and design complexity.