Flexible Duplexing: No More Self-Interference
This article is from the CW Journal archive.
Ten years ago a smartphone would typically support just a few standards operating in perhaps four GSM bands, and maybe a couple for WCDMA or CDMA2000. Since there were only a small number of frequency bands to choose from, some degree of global harmonisation was achieved via a “quad-band” GSM phone, which used the 850/900/1800/1900MHz bands, and worked anywhere in the world (well, just about).
This was a huge benefit for travellers, and also brought with it a huge economy of scale for device manufacturers, who only needed to make a few models (or possibly just one) for the entire global market. Fast forward to today, and GSM remains the only radio access technology for which global roaming has become a reality. By the way, just in case you didn’t know, GSM is being phased out.
Any smartphone worth the name needs to support 4G, 3G and 2G access, with many and varied RF front-end requirements for bandwidth, transmit power, receiver sensitivity, and a host of other parameters.
Moreover, because of the fragmented availability of spectrum across the globe the 4G standard covers a very large number of bands in order to allow operators to deploy it in the frequencies available in any particular region – currently a total of 50 bands, as defined in the LTE standards1. A true “world phone” would have to be capable of operating in all of these.
A key problem that any cellular radio device has to address is “duplexing”. When we talk, we listen at the same time. Early radio systems used “push to talk” (and some still do), but when we talk on the phone we expect to hear the other party as they interrupt. The first (analogue) generation of cellular devices used “diplexing filters” (or duplexers) to receive the downlink without being “deafened” by transmitting the uplink at a different frequency.
Making these filters small and cheap was a major problem for early handset makers. When GSM came along the design of the protocol allowed the transceiver to be “half duplex”.
This was a pretty clever way to eliminate the duplexer, and one important factor helping GSM to become a cheap, high volume technology able to dominate the industry (and change the way humans communicate in the process).
BARE ESSENTIALS
Android OS inventor, Andy Rubin’s Essential phone has the very latest in connectivity, ranging from Bluetooth 5.0LE with a wide range of GSM/LTE and WiFi antennas hidden in its titanium frame.
Sadly, that technological problem-solving lesson was soon forgotten in the techno-political wars of the early days of 3G, which in its now dominant frequency-division duplexing (FDD) form needs a duplexer for each operating FDD band. Without doubt the rush to LTE has imposed an ever greater element of cost.
While there are bands in which time-division duplexing, or TDD (where the radio switches quickly back and forth between transmit and receive) can be used, these are smaller in number. Most operators, those in Asia largely being the exception, favour the FDD bands, of which there are over 30.
The legacy of TDD and FDD spectrum, the difficulty of liberating truly global bands, and the coming of 5G with even more bands, makes the duplexing problem even harder. Promising techniques being investigated include novel filter-based designs and the possibility of cancelling self-interference.
The latter also leads to the possibility of “division free” duplex (or “In-band Full-duplex”) that has some promise. In the 5G future of mobile communication we will perhaps need to look at not just FDD and TDD but flexible duplexing based on these new techniques.
Researchers from the University of Aalborg, Denmark, have developed a “Smart Antenna Front End” (SAFE) architecture2-3, which uses (see inset on page 18) separate antennas for transmitting and receiving, combining these with (relaxed performance) tunable filtering to achieve the required transmit-to-receive isolation.
While the performance is impressive, the need for two antennas is a substantial drawback. As phones become thinner and sleeker, the space allocated for the antennas gets ever smaller.
Multiple antennas are also needed in mobile devices for spatial multiplexing (MIMO). A 2x2 MIMO-capable phone based on the SAFE architecture would require a total of four antennas. Moreover, the tuning range of these filters and antennas is limited.
A world phone would therefore also require the duplication of this front-end architecture in order to cover all LTE bands (which range from 450MHz to 3600MHz), requiring yet more antennas, more antenna tuners, and more filters, and bringing us back to the familiar problem of achieving multi-band operation through component duplication.
Whilst it may be possible to fit a larger number of antennas within a tablet or laptop, further advances in tunability and/or miniaturisation will be needed for this technique to be suitable for smartphones.
Electrical balance duplexing has been used since the early days of wired telephony17. In a telephone system, the microphone and earpiece must both be connected to the telephone line, but isolated from one-another to prevent the user’s own speech deafening them to the much weaker incoming audio signal. This was achieved in pre-electronic telephones using a hybrid transformer.
The duplexing circuit shown in A below matches the impedance of the transmission line with a resistor of the same value, such that current from the microphone is split when entering the transformer, flowing through the primary coil in opposite directions. The magnetic fluxes effectively cancel out, inducing no net current in the secondary coil, which is therefore isolated from the microphone.
However, the signals from the microphone are still coupled to the telephone line (albeit with some loss), and incoming signals on the telephone line are still coupled to the speaker (also with some loss), thus allowing bi-directional communication through the same wire.
The wireless electrical balance duplexer is analogous to the telephone duplexer, but replaces the microphone, earpiece, and telephone line with the transmitter, receiver, and antenna respectively, as is clearly shown in B.
A third way to isolate the transmitter from receiver is to cancel out the self-interference (SI), thus subtracting the transmitted signal from the received signal. Cancellation techniques have been used in radar and broadcasting for several decades.
For example, in the early 1980s, Plessey developed and sold a product based on SI cancellation called “Groundsat” for range extension of half-duplex analog FM military communication networks4-5.
The system acted as a full-duplex on-frequency repeater, extending the effective range of the half duplex radios in use across the area of operation.
More recently, largely due to the trend towards short-range (cellular and Wi-Fi) communication, which makes the problem of SI cancellation easier to deal with due to lower transmit and higher receive powers, there has been interest in self-interference cancellation for use in consumer wireless access and backhaul applications6-8.
ADVANTAGE APPLE
Apple's iPhones (with a little help from Qualcomm) arguably have the world's best wireless and LTE connectivity with support for 16 LTE bands on a single chipset. This means it only has to make two SKUs to cover GSM and CDMA markets
In division-free duplex applications, self-interference cancellation could increase spectral efficiency by allowing use of the same spectral resources for uplink and downlink9,10. Self-interference cancellation techniques could also be used to create a tunable duplexer for FDD.
The cancellation itself generally has multiple stages. A directional network between the antenna and the transceiver gives a first level of separation between the transmit and receive signals. Second, additional analogue and digital signal processing is used to cancel out the remaining self-interference in the received signal. The first stage can use separate antennas (as in SAFE); a hybrid transformer (as described below); or a circulator.
The problems with separate antennas have been described already. Circulators tend to be narrow band as they use a ferromagnetic resonance in a crystal. The hybrid – or “Electrical Balance Isolation” (EBI) – is a promising technique that can be wideband and could potentially be integrated on-chip.
The smart antenna front-end design, depicted below, uses two narrowband tunable antennas: one for transmitting and one for receiving, and a pair of lower performance but tunable duplexing filters. The separate antennas not only provide some passive isolation through the propagation loss between them, but are also designed to have a limited (but tuneable) instantaneous bandwidth.
The transmitting antenna operates efficiently only in the transmit band, and the receiving antenna operates efficiently only in the receive band. By doing this the antennas themselves also act as filters, with the out-of-band Tx emissions being attenuated by the transmitting antenna, and the self-interference in the Tx band being attenuated by the receiving antenna.
This architecture therefore requires that the antennas are tunable, which is achieved through the use of antenna tuning networks. The antenna tuning network has some unavoidable insertion loss. However, recent advances in MEMS tunable capacitors18 have substantially increased the quality factor of these devices, thereby reducing losses. The Rx insertion loss is around 3dB, which is comparable to the loss of SAW duplexer and switches combined.
The antenna-based isolation is then supplemented by the tunable filters, also based on MEMs tunable capacitors3, aiming to achieving 25dB of isolation from the antennas and 25dB of isolation from the filters. Prototypes have demonstrated this is achievable.
Several research groups within academia and industry are studying the use of hybrids for duplexing11-16. These circuits cancel SI passively, allowing for simultaneous transmission and reception from a single antenna but isolating the transmitter and receiver. They are inherently broadband, and can also be implemented on-chip, making it an attractive choice for implementing frequency duplexing in mobile devices.
Recent advances demonstrate that FDD transceivers using EBI can be made in CMOS (complementary metal-oxide-semiconductor) with insertion loss, noise figure, receiver linearity, and blocker rejection characteristics suitable for cellular applications11,12,13. However, there is a fundamental limitation which has affected the duplex isolation, as can be seen in many examples in academic and scientific literature.
The impedance of a radio antenna is not fixed but varies with operating frequency (because antennas are resonant) and time (because of inter-action with the changing environment). This means the balancing impedance has to adapt to track impedance variations, and the isolation is band-limited, because of the frequency domain variation13 (see figure 1).
Our work at the University of Bristol has focussed on investigating and mitigating these performance limitations in order to demonstrate that the required transmit-to-receive isolation and bandwidth can be achieved in realistic use cases.
To overcome the antenna impedance fluctuations (which have a severe impact on the isolation) our adaptation algorithms track the antenna impedance in real time, and tests have demonstrated that performance can be maintained across a range of dynamic environments, including interaction with the user’s hand, and in high-speed car and rail travel.
Furthermore, to overcome the limited matching of the antenna in the frequency domain, and thereby increase the bandwidth and the total isolation, we combine the electrical balance duplexer with further stages of active SI cancellation, using a second transmitter to generate a cancellation signal to further suppress self-interference (see figure 2).
Results from our testbed are promising, with the active technique providing substantial increases in transmit-to-receive isolation when combined with the EBD, as shown in figure 3.
Our latest lab setup incorporates low-cost mobile device components (handset power amplifiers and antennas), making it representative of a handset implementation. Furthermore, our measurements show that this type of two-stage, self-interference cancellation can achieve the required duplex isolation in both the uplink and downlink band, even when using low-cost commercial grade hardware.
A cellular device at maximum range needs to receive a signal power 12 orders of magnitude lower than it transmits. In time division duplexing (TDD), the duplexing circuit is simply a switch which connects the antenna to either the transmitter or the receiver so the duplexer in TDD is a trivial switch. In FDD, where the transmitter and receiver operate simultaneously, the duplexer uses filters to isolate the receiver from the high-powered transmitter signal.
Duplexers in cellular FDD front-ends provide >~50dB of isolation in the uplink band, to prevent the Tx signal from overloading the receiver, and in the downlink band, to prevent receiver desensitisation due to the out-of-band Tx emissions which fall in the Rx band, with minimal loss in the transmit and receive paths.
This low loss and high isolation requirement, at a frequency separation of just a few percent, requires high-Q filtering which can still only be implemented using surface acoustic wave (SAW) or bulk acoustic wave (BAW) devices.
Whilst this technology continues to develop, with advances resulting largely from the huge volumes of these devices needed, multi-band operation means a separate off-chip duplexing filter for each band, as shown in figure A. All the switches and routing also add extra loss and compromise performance.
A reasonably priced world phone based on the current technology would be too hard to make. The resulting radio architecture would be prohibitively large, lossy, and expensive. Manufacturers must make multiple product variants for the different band combinations needed in various regions, and this prevents unrestricted global roaming on LTE. And the scale economies that led to GSM’s dominance get increasingly hard to achieve.
The ever-increasing demand for high data-rate mobile services has led to 4G mobile networks being deployed across 50 bands, with more to come as 5G becomes fully defined and widely deployed. Covering all of these using the current filter-based technology is not feasible in one device because of the complexity of the RF front-end, so tunable and reconfigurable RF circuits will be needed.
Ideally a new approach to solving the duplexing problem is required, which might be based on tunable filters or self-interference cancellation, or some combination of the two.
Though we don’t yet have a single approach capable of meeting the many requirements of cost, size, performance and efficiency, maybe the pieces of the puzzle are coming together towards something that could be in your pocket not many years from now.
Techniques such as EBD with SI cancellation may open the possibility of using the same frequencies simultaneously in both directions, which could potentially give a significant increase in spectrum efficiency.
Maybe as we progress we won’t speak of FDD and TDD as separate modes but just expect that a radio will work flexibly, switching seamlessly from time to frequency to “division free” duplex to suit the spectrum available, network configuration, and the user’s activity, moment to moment.
One observation we can make based on this work is that such advances are only possible by taking a holistic approach to the system.
Ideally, the standards’ design needs to take more account of RF hardware issues; and product design has to combine novel circuit approaches with digital and analogue signal processing and software control.
The industry today has divided the system into conceptual black boxes, each with its defined inputs and outputs and performance parameters and generally tackled by a different design team.
This constrains thinking and inhibits innovation. The industry of tomorrow needs to reinvent itself. And the radio.
The author would like to thank his colleagues Prof. Mark Beach, Dr. Kevin Morris, Jack Zhang and Prof. John Haine at Bristol University for their cooperation in the research leading to this article, and to the University, EPSRC, and u-blox AG for supporting the work. Thanks also go to many colleagues in u-blox Melbourn and Cork for their support.
- 3GPP, “Evolved Universal Terrestrial Radio Access (E-UTRA); User Equipment (UE) radio transmission and reception,” 2017, no. 36.101 v14.3.0.
- S. Caporal Del Barrio, A. Tatomirescu, G. F. Pedersen, and A. Morris, “Novel Architecture for LTE World-Phones,” IEEE Antennas Wirel. Propag. Lett., vol. 12, pp. 1676–1679, 2013.
- P. Bahramzy, P. Olesen, P. Madsen, J. Bojer, S. Barrio, A. Tatomirescu, P. Bundgaard, A. S. Morris III, and G. F. Pedersen, “A Tunable RF Front-End With Narrowband Antennas for Mobile Devices,” IEEE Trans. Microw. Theory Tech., vol. 63, no. 10, pp. 3300–3310, Oct. 2015.
- “Communications News: Plessey Breaks New Ground,” Electron. Power, vol. 24, no. 10, p. 715, 1978.
- C. K. Richardson, “Improvements in or relating to transmitters/receivers,” UK Patent, GB1577514, 1980.
- S. Rockman, “New radio tech could HALVE mobe operators’ bandwidth needs,” The Register. Feb 2014. http://www.theregister.co.uk/2014/02/27/mwc2014_kuma_networks_mobile_tech_halves_bandwidth/.
- Geoff Carey (Mimotech), “Air Division Duplexing doubles Transmission Capacity for Microwave Backhaul.” Presented to the CW Radio Technology SIG, July 2015, Bristol, U.K. https://www.mimotechnology.com/p_microwave_carrier_ethernet.htm.
- D. Bharadia, E. McMilin, and S. Katti, “Full Duplex Radios,” in Proc. 2013 ACM SIGCOMM, 2013.
- S. Chen, M. A. Beach, and J. P. McGeehan, “Division-free duplex for wireless applications,” Electron. Lett., vol. 34, no. 2, pp. 147–148, 1998.
- A. Sabharwal, P. Schniter, D. Guo, D. W. Bliss, S. Rangarajan, and R. Wichman, “In-Band Full-Duplex Wireless: Challenges and Opportunities,” IEEE J. Sel. Areas Commun., vol. 32, no. 9, pp. 1637–1652, Sep. 2014.
- M. Mikhemar, H. Darabi, and A. A. Abidi, “A Multiband RF Antenna Duplexer on CMOS: Design and Performance,” Solid-State Circuits, IEEE J., vol. 48, no. 9, pp. 2067–2077, 2013.
- S. H. Abdelhalem, P. S. Gudem, and L. E. Larson, “Hybrid Transformer-Based Tunable Differential Duplexer in a 90-nm CMOS Process,” Microw. Theory Tech. IEEE Trans., vol. 61, no. 3, pp. 1316–1326, 2013.
- L. Laughlin, M. A. Beach, K. A. Morris, and J. L. Haine, “Optimum Single Antenna Full Duplex Using Hybrid Junctions,” IEEE J. Sel. Areas Commun., vol. 32, no. 9, pp. 1653–1661, Sep. 2014.
- B. van Liempd, B. Hershberg, S. Ariumi, K. Raczkowski, K.-F. Bink, U. Karthaus, E. Martens, P. Wambacq, and J. Craninckx, “A +70-dBm IIP3 Electrical-Balance Duplexer for Highly Integrated Tunable Front-Ends,” IEEE Trans. Microw. Theory Tech., pp. 1–13, 2016.
- L. Laughlin, C. Zhang, M. A. Beach, K. A. Morris, and J. L. Haine, “Passive and Active Electrical Balance Duplexers,” IEEE Trans. Circuits Syst. II Express Briefs, vol. 63, no. 1, pp. 94–98, Jan. 2016.
- L. Laughlin, M. A. Beach, K. A. Morris, and J. L. Haine, “Electrical balance duplexing for small form factor realization of in-band full duplex,” IEEE Commun. Mag., vol. 53, no. 5, pp. 102–110, May 2015.
- G. A. Campbell and R. M. Foster, “Maximum Output Networks for Telephone Substation and Repeater Circuits,” Am. Inst. Electr. Eng. Trans., vol. XXXIX, no. 1, pp. 231–290, 1920.
- D. R. DeReus, S. Natarajan, S. J. Cunningham, and A. S. Morris, “Tunable capacitor series/shunt design for integrated tunable wireless front end applications,” in 2011 IEEE 24th International Conference on Micro Electro Mechanical Systems, 2011, pp. 805–808.
Leo’s award-winning research on signal cancellation technology forms the basis of Forefront RF’s product development. He was previously a Research Fellow at the University of Bristol, and has worked in the wireless industry on mobile chipset development. He has a Ph.D in wireless communication, has published in top academic and industry journals, and is named inventor on several patents.
