Adapting to a multi-band future
Mobile data services are experiencing a period of exponential growth, driven by enthusiastic adoption of smartphones and tablets in the consumer market, and expansion of applications for mobile usage, such as social networking, mobile video and video calling. Some estimates anticipate that 1.3 exabytes of data traffic will pass through the network in 2012 (‘exa’ denotes a multiplier of 1018, equivalent to 1million terabytes), with no sign of that growth slowing; some estimates predict up to 10.8 exabytes per year by 2016.
Efficient modulation and coding schemes have already played a pivotal role in delivering increased data capacity; attention is also turning to closely-related – although more fundamental – parameters of the radio link, such as more spectrum allocation, and maintaining higher-quality connections over the air interface.
Spectrum is being added as legacy services are phased out and new bands made available, but provision is fragmented in terms of frequency and geography. Examples include the 700MHz bands in North America, the 2600MHz band in Europe, and 2300MHz band in China. This points to a major challenge for designers of next-generation handsets: today’s 3G handset typically covers three or four bands, which will soon grow to five, and as many as eight to 12 bands in the newest high-end 4G smartphones.
Inevitably, operators will need to improve radio link quality, and part of the solution will require lower-loss performance from RF components and an optimised antenna interface. The antenna is greatly affected by the addition of more frequency bands because it is very difficult to develop an antenna that covers two octaves yet maintains a form factor acceptable for the handset. Simply maintaining present levels of performance while adding more bands will require tunable handset antennae.
Operators are also reassigning spectrum from 2/2.5G technology to next-generation wireless standards such as High-Speed Packet Access (HSPA), Evolved High-Speed Packet Access (HSPA+) and Long Term Evolution (LTE). These standards support higher data rates and are more spectrally efficient, but they use more complicated modulation schemes which ultimately demand better signal quality from the terminals. To achieve the target peak data rates of LTE, a signal-to-noise ratio (SNR) of more than 30dB is needed . For comparison, a typical WCDMA system can operate with a signal-to-noise ratio (SNR) of only a few dB.
RF front-end limitations
In a conventional RF front-end (RFFE) architecture (Figure 1) all of the signals pass through a single broadband signal path, yielding acceptable performance for voice services and where just a few bands are in use.
Figure 1: Traditional RF Front-end Architecture
As already noted, broadband antennae that are also small are very difficult to realise; in particular, they do not present a uniform impedance (by convention, 50Ω) across a wide frequency range. Typical passive penta-band (824-960MHz and 1710-2170MHz) handset antennae are designed for a voltage standing wave ratio (VSWR) of 3:1 in order to get acceptable performance across the band. A VSWR of 3:1 leads to 1.25 dB of impedance mismatch loss. Further losses occur in the duplexer and power amplifier (PA) due to the mismatch.
The addition of the LTE mobile standard extends frequencies downwards into the 700MHz range, which impacts antenna performance by pushing the typical VSWR beyond 6:1 and increasing the mismatch loss by >4dB. Lower frequencies imply physically larger antennae; space-constrained handset configurations perform less well and exhibit narrow bandwidth at low frequencies.
The relationship known as Wheeler's equation defines the minimum volume an antenna must have to operate efficiently, at a given frequency. For 700 to 960MHz it yields a figure of over 3.5 cm3; in reality, the available volume is under 3.0 cm3 and even wider bandwidth is needed.
Additional bands and co-existence
Handling multiple signal bands and modes in the RFFE will require extensive switching of signals among circuit blocks, with complex signal routing via an ever-increasing array of higher-throw-count RF switches. As the complexity of the RF signal chain increases, maintaining acceptable levels of insertion loss, isolation and harmonic performance from that switch matrix becomes more challenging. Typically, gains in one RF metric come at the expense of another.
Additional services introduce more complex co-existence challenges. An LTE system designed to coexist with a global positioning system (GPS) receiver will use LTE Band 13 at 777-787MHz, Band 14 at 788-798MHz. The GPS signal is at 1575.4 MHz; the second harmonic (2fo) of the LTE transmit signal can fall into the GPS band causing degradation in GPS sensitivity. The high-throw-count RF switch must deliver better than +130dBm second-order intercept point (IP2). This is a formidable task for even the industry’s best switch solutions.
For the terminal, radiated power is measured using total radiated power (TRP), and terminal receiver performance is measured in total isotropic sensitivity (TIS). When combined with other factors, such as base station Tx/Rx performance, propagation losses and fading allowance, these metrics define the signal range of the phone and how much data can be carried. This is also known as the radio link budget.
One way to expand data capacity is to improve spectral efficiency. Doubling spectral efficiency requires approximately 7dB improvement in the SNR of the radio link. Every 1dB improvement in the radio link can reduce green-field basestation deployment by up to 10%. Some of the efficiency gains are offset by the introduction of the higher-order modulation schemes that require SNRs approaching 30dB (Figure 2).
Figure 2: SNR requirements increase with increasing band support
One approach is to implement RFFE tuning, to reduce mismatch loss reduction and optimise interface impedances in order to improve antenna radiation efficiencies.
Tunable dual-feed antenna
Optimising the RFFE calls for new system architectures and high-performance components. Traditionally, handsets have used an antenna structure with all bands routed to a single antenna feed point (Figure 1). Because it is physically impossible to design a single-feed handset antenna that covers a continuous frequency band between 700MHz and 2.7GHz, the antenna actually has two or three narrower resonances that correspond to the required frequency bands. This approach becomes progressively more difficult as more bands are added.
One solution is a dual-feed antenna (Figure 3), which splits the antenna into two separate radiators: one radiator optimised for the low-band frequencies and the other for high-band frequencies. Because components can be optimised for the band being used, and not the full range, this also improves the performance of the RFFE.
Figure 3: Tunable Dual-Feed antenna architecture
In a dual-feed architecture, switch performance can improve because the switch capacitance is reduced by using two devices, each assigned to a separate band. For example, a single-pole, ten-throw (SP10T) switch requirement can be implemented as two single-pole, five-throw (SP5T+SP5T) devices (Figure 4).
Narrower operating bandwidth allows improved narrow-band resonant matching by creating a tank circuit with the shunt L (which is used for electrostatic discharge protection). This reduces the components count and the insertion loss of the switch, by approximately 0.3dB.
In addition to the performance gains realised with a dual-feed antenna approach, the architecture in Figure 4 is also capable of handling dual-carrier signals (an option in the HSPA+ mobile standard) through simultaneous support for low band and high band.
Figure 4: Comparing the single- and dual-feed antenna approaches
Tunable matching networks
A new approach to improve performance in next generation RFFEs is the introduction of tunable matching networks – resonant circuits comprising tunable capacitors and inductors. These networks reduce VSWR values conditions through impedance matching. In the dual-feed antenna architecture, tunable matching networks can be optimised for high band and low band independently. This allows the RFFE tuners to have wide bandwidth, low loss, and wide Smith Chart coverage.
Figure 5: Physical implementation of an RFFE tuner
Figure 5 shows a prototype RFFE tuner module with a dual-feed architecture for LTE/WCDMA/GSM bands. A monolithic die containing four digitally-tunable capacitors (DTCs) and two fixed capacitors is placed on a small laminate with four surface-mount inductors. The module size is 3.5 x 3.5mm2.
The tuner has been optimised for very wide-band performance, for full-duplex systems with wide spacing between Tx and Rx frequencies, such as WCDMA and LTE. Using a coupled resonator topology it provides wide impedance coverage over 698 – 960MHz and 1710 – 2170MHz. Insertion loss (IL) and return loss (RL) performance at 50Ω demonstrate excellent broadband coverage. The broad Smith Chart coverage shown in Figure 6 demonstrates the RFFE tuner’s capability to effectively match a wide range of antenna impedances to 50 Ω.
Figure 6: Measured Smith Chart coverage (points that can be matched back to 50Ω) of Dual-Feed RFFE tuner
Closing the loop
A further development of the tunable dual-feed antenna is shown in Figure 7. The RFFE tuner is located between the dual-feed antennas and the output of the RF engine, reducing the mismatch seen by the dual-feed RF engine. This significantly improves RFFE performance and minimises PA current consumption and complexity. The RFFE tuner is equipped with a 50Ω-in-50Ω-out state, providing a 50Ω connection point for conducted test of the handset in production.
Figure 7: Tunable Dual-Feed antenna approach
The directional coupler at the output of the FEM is part of the PA power-control loop, maintaining output power into mismatch conditions. The outputs of the directional couplers are connected to the transceiver where forward and reverse power detectors are located, for delivered power and mismatch calculation. In the dual-feed architecture, the existing hardware is also used for mismatch detection for the RFFE tuner, eliminating dedicated mismatch detection.
The preferred location for the tuning algorithm (which adjusts capacitance inside the RFFE tuner to minimise impedance mismatch seen by the RF engine) is in the baseband processor. The tuning algorithm can be implemented as a part of the PA power-control loop algorithm, and also can use information on mode, band, frequency, and modulation that is available in the baseband to optimise the tuning algorithm for a given use case.
RF front-end performance must improve to meet the needs of forthcoming smartphones that will use many more bands than today’s designs. Dual-feed antennae together with the appropriate tuning technology can meet the required specifications, and the necessary components already exist.
By Tero Ranta, Duncan Pilgrim & Richard Whatley, Peregrine Semiconductor