Linearity Looms Large for Next Generation RF Systems
The emergence of Long Term Evolution (LTE) as the next generation mobile wireless standard beyond 3G renews the importance of broadband, highly linear RF systems and components. The deployment of LTE in the 700 MHz spectrum in the United States only increases an already crowded spectrum, raising the bar even higher for systems engineers to ensure that co-existence and co-location requirements are met, adjacent channel bands are not corrupted and the signal bands of interest have sufficient dynamic range.
Figure 1 Worldwide LTE infrastructure revenue forecast.
The gradual transition from lower data rate 3G to 4G is highlighted by the TD-LTE rollout expected in China in 2012. Infonetics Research predicts that the LTE infrastructure market will exceed $11 B in 2014 (see Figure 1).1 The demand for faster, more efficient devices is driving this steep growth curve in LTE as consumers want to access more data on-demand from their laptops, smartphones, tablets and other portable devices.
LTE extends the performance of 3G to much higher data rates by using two different access schemes. Single-carrier frequency-division multiple access (SC-FDMA) is used on the uplink (base station receive, mobile station transmit) and orthogonal frequency-division multiple access (OFDMA) is used on the downlink (base station transmit, mobile station receive). In addition, data throughput is increased by the use of Multiple Input, Multiple Output (MIMO) technology.
From a linearity standpoint, the implementation of LTE poses several challenges in the system design. The use of an adaptive modulation scheme that varies from QPSK to 64 QAM necessitates reducing the amplitude and phase distortion of the modulated RF signal. The transmitter is required to meet specific spurious emissions requirements and minimize adjacent channel leakage. On the flip side, the receiver has to account for a worst-case sensitivity degradation from the mixing of an out-of-band interferer with its own transmit signal or a transmit signal from a separate antenna. Coupled with the use of closely spaced (15 kHz wide) orthogonal subcarriers and the limited transmit band to receive band spacing, LTE raises the complexity of the overall system linearity requirements.
For most components such as switches and digital step attenuators used in mobile wireless and wireless infrastructure systems, the out-of-band distortion products are quantified in terms of Harmonic Distortion (HD) and Intermodulation Distortion (IMD). To better analyze the impact of IMD with respect to the modulated RF carrier, second- and third-order distortion terms IMD2 and IMD3 are specified as second- and third-order intercept points IP2 and IP3. In addition, Cross Modulation Distortion (CMD) is a critical in-band consideration when multiple transmitters and receivers co-exist in the same geographic area.
Examining one scenario in LTE, the second harmonic of the ‘C’ Block (777 to 787 MHz) uplink signal will fall into the GPS ‘L1’ band of 1575.2 MHz, making HD2 an important consideration. In another scenario in W-CDMA, the operating band I (1950 MHz) uplink signal will intermodulate with a GSM1800 out-of-band interferer at 1760 MHz, causing a de-sense of its own receiver.2 In this case the component IIP3 would be the critical specification.
Designing highly efficient, highly linear systems has always been a challenge. Even with current 3G networks, the systems used in the back-end infrastructure typically use some form of linearization, be it analog pre-distortion or digital pre-distortion (DPD). For example, remote-radio heads, a cost-effective way to extend cell coverage without the addition of more base stations, often contain a multi-carrier power amplifier embedded in a DPD loop. The main goals of digital pre-distortion are to meet the stringent adjacent channel power ratio (ACPR) requirements by reducing odd order intermodulation distortion and to maximize power amplifier efficiency.
In wideband, multicarrier systems such as W-CDMA, there is also the concept of UTRAN sharing where two network operators actively share the same network infrastructure to reduce capital and operating expenses.3 For the systems engineer, that can translate to a more complex receive front-end that has to meet the sensitivity requirements of two co-located receivers operating simultaneously within the same band.
Sorting through the standard specifications can be challenging enough besides having to select components that enable the designer to meet the stringent system requirements. In most cases, simply meeting the technical requirements may not be enough; it has to be done at a low cost and with as little power consumption as possible.
UltraCMOS™ Technology Addresses Market Demands
Systems engineers are offered an extensive menu of amplifiers, switches, mixers and attenuators in various competing technologies from CMOS to GaAs to Silicon-On-Insulator (SOI). UltraCMOS™ technology, a patented Silicon-On-Sapphire (SOS) technology, has demonstrated the capability to provide broadband, highly-integrated and highly linear RF components. One of the main advantages of UltraCMOS technology is that by using an insulating sapphire substrate, there are no nonlinear, voltage dependent parasitic bulk capacitances that commonly plague other technologies. Coupled with HaRP™ technology enhancements, UltraCMOS components have consistently demonstrated the capability to meet the market requirements for broadband, high linearity at a low power consumption, size and cost.
With the progression of mainly voice communication (2G networks) to data intense communication requiring different multiplexing and modulation schemes (3G and beyond), the market requirements for a high linearity RF front-end (RFFE) switch have evolved. For example, W-CDMA required a high linearity RFFE as some or all W-CDMA bands had to be routed through a specially designed multi-mode antenna switch. The antenna switch, typically connected directly to the antenna port without any filtering, had to be linear enough to cope with any unwanted interferes without degrading the receiver sensitivity.
Figure 2 Evolving IIP3 market requirements for handset antenna switches.
Figure 2 shows the mobile handset antenna switch Input IP3 requirements as a function of time. For each technology transition, as the linearity requirements increased, an Ultra-CMOS switch solution was delivered to the market. One of the first devices to be released on the HaRP-enhanced UltraCMOS process was a SP7T switch for quad-band GSM and GSM/W-CDMA handset applications, featuring an IIP3 of +67 dBm.
As next generation mobile phones integrate LTE with quad-band W-CDMA (850, 900, 1900, 2100) and quad-band GSM (850, 900, 1800, 1900), the antenna switch linearity requirements will only get more demanding. To support the transition to LTE and LTE-Advanced, multi-throw switches requiring an IIP3 greater than +80 dBm will likely become a reality.
Besides designing and manufacturing such a switch, the challenge extends to validating the high linearity in the lab. An +80 dBm IIP3 translates to an extremely low IMD3 level, with the IIP3 expressed as the RF input power plus half the difference between the desired fundamental output and undesired IMD3 output. The distortion produced from the test system itself would have to be at least 18 dB below the IMD3 output in order to prevent the system from limiting the measurement.4 That translates to a system IIP3 requirement of at least +89 dBm to accurately measure the true linearity of the switch.
Most engineers are familiar with the traditional two-tone test for measuring the intercept point, expressed as Input or Output IP3.5 In this setup, two CW tones (f1, f2) typically of equal amplitude spaced at a specific frequency offset are combined and driven into a nonlinear device under test, producing IMD3 products at the frequency offset below and above the fundamental tones. The output is then driven directly into a spectrum analyzer for intercept point measurement using the f1, f2 and 2f1 - f2 and 2f2 - f1 frequencies. In fact, many spectrum analyzers even have a third-order intercept point “TOI” measurement built-in.
The reality is such a conventional technique will be limited by the system dynamic range. To measure the distortion of a highly linear component, the two test tones would need to have large amplitudes to raise the IMD tones out of the measurement noise floor. To process the large desired output signals, the spectrum analyzer requires its RF input attenuation to be optimized to function in the linear range of operation. Doing so, however, raises its noise floor, increasing the minimum signal level that can be measured. This dynamic range limitation leaves the user with balancing the system distortion and system noise floor even if the signals are individually measured in a narrow span by optimizing the resolution and video bandwidth filter settings.
For example, with two tones at +18 dBm/tone, an +80 dBm IIP3 would place the IMD3 tones at -106 dBm, a dynamic range of 124 dB. Most high-performance spectrum analyzers today cannot meet this requirement, system limiting at 110 dB levels for third-order intercept measurement.4 If the requirement to keep the measurement uncertainly to less than 1 dB is applied, then the dynamic range is closer to 100 dB levels.
Overcoming System Limitations
An approach to work around this limitation is to measure the distortion at the sum frequencies rather than the difference frequencies. Since the distortion products at the sum frequencies are mathematically equivalent to the distortion products at the difference frequencies, the test system can focus on capturing the desired and undesired outputs separately and is no longer required to simultaneously process both sets of signals.
Figure 3 Two-tone IIP3 measurement setup for sum IMD3 measurement.
The sum IMD3 products occur at 2f1 + f2 and 2f2 + f1 frequencies and can be captured using a diplexer or triplexer at the DUT output, with each path capturing a specific frequency band. As shown in Figure 3, a dual-channel power meter is used to accurately measure the fundamental input and output power levels, with the spectrum analyzer measuring the sum IMD3 tones. All tones (fundamentals and distortion) are terminated to 50 Ω thru filters and attenuators, eliminating any sensitivity to phase angles. In addition, all setup losses and DUT losses at the sum IMD3 frequencies are included in the final IIP3 calculations.
Figure 4 UltraCMOS SPDT performance in the LTE band.
A new SPDT switch was measured in this setup with tones at +18 dBm/tone. It features a 50 W P1dB compression point and an insertion loss of < 0.4 dB below 1 GHz. As shown in Figure 4, it measures an IIP3 close to +80 dBm in the LTE band. Competing SOI and GaAs technologies have struggled to demonstrate this level of broadband linearity at these power levels. With GaAs technology, while the linearity may be acceptable for lower power levels, the distortion increases at high power due to the gate voltage being modulated.6 Finding a multi-throw antenna switch that can meet the high linearity, insertion loss and isolation requirements of an integrated LTE RFFE can be extremely challenging.
As the next generation RF systems continue to develop, one thing is for sure—high linearity will always be a premium. The co-existence of multiple bands and the sharing of existing network infrastructure to alleviate CapEx and OpEx costs will only increase the challenges for RF system designers. Coupled with the need for high linearity components arises the dilemma of how to accurately measure their performance as conventional techniques fall short.
- Infonetics Research, April 2010, LTE Infrastructure and Subscribers, http://www.infonetics.com/pr/2010/2H09-LTE-Infrastructure-Market-Highlights.asp.
- T. Ranta, J. Ella and H. Pohjonen, “Antenna Switch Linearity Requirements for GSM/W-CDMA Mobile Phone Front-ends,” 8th European Conference on Wireless Technology Proceedings, Paris, France, October 2005, pp. 23-26.
- “Network Sharing in LTE, Opportunity & Solutions,” Alcatel-Lucent Technology White Paper, pp. 1-2, http://lte.alcatel-lucent.com/locale/en_us/downloads/CMO1649091201_LTE_Network_Sharing_EN_TechWhitePaper.pdf.
- “Optimizing Dynamic Range for Distortion Measurements,” Agilent Technologies Product Note, 5980-3079EN, pp. 4-22, http://cp.literature.agilent.com/litweb/pdf/5980-3079EN.pdf.
- K. Kundert, “Accurate and Rapid Measurement of IP2 and IP3,” Designers Guide Consulting Application Note, pp. 2-4, http://www.designers-guide.org/analysis/intercept-point.pdf.
- “RF Switch Performance Advantages of UltraCMOS Technology Over GaAs Technology,” Peregrine Semiconductor Application Note AN18, pp. 1-2, http://www.psemi.com/pdf/app_notes/an18.pdf.