RECONFIGURABLE ANTENNAS FOR SDR AND COGNITIVE RADIO K. R. Boyle*, P. G. Steeneken†, Z. Liu*, Y. Sun°, A. Simin°, T. Huang*, E. Spits°, O. Kuijken°, T. Roedle°, F. van Straten° *NXP Semiconductors, Cross Oak Lane, Redhill, Surrey, RH1 5HA, United Kingdom. †NXP Semiconductors, High Tech Campus, 5656 AE Eindhoven, The Netherlands. °NXP Semiconductors, Gerstweg 2, 6534 AE, Nijmegen, The Netherlands. e-mail: [email protected]

Keywords: Reconfigurable, cognitive, antenna, PIFA.

Abstract Mobile phones will be required to operate in more frequency bands. This paper presents appropriate reconfigurable and adaptive antenna systems. Two approaches are explored based on utilising RF MEMS and on the co-design of the antenna(s) and RF front-end circuitry. Both approaches yield overall improvements in performance and/or size and are likely to be used in future SDRs and cognitive radios.

1 Introduction Mobile phones are continually being required to operate in more frequency bands using an increasing number of communication protocols. Up until the end of the 1980s, mobile phones operated using single analogue systems such as the Advanced Mobile Phone System (AMPS) system in the USA and the Total Access Communication System (TACS) in Europe. These “first generation” (1G) systems operated in a single frequency band. The 1990s saw the introduction of digital “second generation” (2G) systems, such as the Global System for Mobile Communications (GSM), and the beginning of multi-band operation: for example, the introduction of the 17101880MHz band to accompany the existing 880-960MHz band in Europe. These frequency bands were not available worldwide. In particular, a different band (1850-1980MHz) was utilised in the USA, which led to the introduction and eventual widespread adoption of tri-band GSM phones. More recently, quad-band GSM phones have emerged, as the frequency bands associated with AMPS in the USA (824894MHz) have been re-allocated to GSM (amongst other systems). The first decade of the new Millennium has seen further major advances. “Third generation” (3G) systems were introduced, first in Japan in 2001, then in Europe in 2003 and more recently in several other countries worldwide.

When second generation systems were introduced, they largely replaced first generation systems. However, 3G systems were introduced to sit along side existing 2G systems. For the first time, this required that phones became capable of operation in two modes. At around the same time Bluetooth radios were introduced into phones to allow shortrange communications with other devices. In effect, this meant that some phones became tri-mode. Since the different modes operate in different frequency bands, phones also became increasingly multi-band. This trend towards the inclusion of more systems operating in more bands has continued with the introduction of the Global Positioning System (GPS), WiFi and new systems such as mobile television (several standards and bands are proposed) and WiMAX (again, several bands are proposed). Many of these systems will be required to operate simultaneously. Multi-mode, multi-band operation presents a formidable challenge to mobile phone designers, particularly for the RF parts. Of these, the antennas occupy the largest volume and, hence, have the biggest impact on the - commercially crucial styling of the device. It is well-known that antenna bandwidth is proportional to volume. However, the space allocated to the antenna(s) is not increasing: on the contrary, it is tending to decrease due to the demand for thinner, more highly stylized devices. To alleviate this problem, reconfigurable antennas have been proposed that are consistent with a general move towards Software Defined Radios (SDRs): radios that can change parameters – such as the operating band – in an optimal way, using software control that is invisible to the end user. This paper presents two reconfigurable antenna systems, both capable of operation in five cellular radio bands. The first antenna system utilises a new technology: radio frequency microelectromechanical systems (RF MEMS) devices. The MEMS are used to switch a modified planar inverted F antenna (PIFA) to operate in different frequency bands. Their low loss characteristics allow more bands to be covered by a physically small antenna. Control of the

antenna (switches) is of a feed-forward nature, as directed by the cellular network, based on the required band of operation. The second antenna system consists of two self-diplexing PIFAs that are co-designed with an antenna interface module (AIM) that contains switches, filters, matching and interconnects realised in conventional technologies. The codesign allows a reconfigurable system with more optimal antenna matching to be achieved (over a number of bands), which again results in a structure that can cover more bands with reduced dimensions. Co-design is also shown to reduce losses. Again, feed-forward control – consistent with SDR of the antenna and AIM is used to select the operating band. Due to the inclusion of the AIM, it is also used to select the operating mode (for example, GSM or UMTS). In addition, it is shown that feed-back control can be utilised to give the antenna and AIM some rudimentary cognisance, independent of the network: adaptive compensation of the antenna detuning effects that occur when a mobile phone is held is shown to be possible.

to avoid problems associated with routing DC actuation voltages on to the antenna itself and because the antenna technology is often incompatible with circuit manufacturing techniques. The antenna and supporting PCB are shown in Fig. 2. The antenna has dimensions 40 x 12 x 8mm and is fabricated from a polyimide flexible PCB that is bonded to a rigid GETEK PCB and folded over a Rohacell block. The PCB has dimensions 40 x 100 x 0.8mm and is metalized on the back surface to provide an RF ground. The antenna/PCB combination is fed via a coaxial cable at a central point on the PCB to avoid excessive perturbation from the feeding cables [1]. A microstrip line runs from this point to the antenna feed.

PCB feed

DC bias lines

2 A Five-band MEMS Reconfigurable Antenna Future mobile phones will be required to operate in the five bands shown in Fig. 1. D. Feed Receive

Transmit

GSM850 UTRA V

C. Switched connection (for impedance matching) GSM1900 UTRA II

B. Short to ground

USA

Figure 2: MEMS switched PIFA and PCB

824 869 to to 849 894

1850 to 1910

GSM900 UTRA VIII

GSM1800 UTRA III

1930 to 1990

UTRA I (UMTS FDD)

Europe 880 925 to to 915 960

Low band

A. Switched connection (for frequency tuning)

1710 to 1785

1805 to 1880

1920 to 1980

2110 to 2170

High band

Figure 1: Typical cellular frequency bands used in Europe and the USA (MHz). Operation is only required in one band at a time, assuming that “compressed mode” is used for handover between GSM and UMTS. Hence, an electrically small antenna may be switched to operate over a number of narrow bands. The bandwidth should be enough to cover the widest of the bands under consideration: in this case UMTS has the widest fractional bandwidth of approximately 12%. Switching is required over a total bandwidth of approximately one octave. To achieve this without significantly reducing efficiency requires low loss switches. Hence, a planar inverted F antenna (PIFA) is switched using capacitive microelectromechanical systems (MEMS) switches fabricated in the industrialized NXP Semiconductors PASSITM process [1]. Switching is performed “off-antenna”,

The antenna has a single slot that is located in a position that is unlikely to interact with the user’s hand when the phone is held [2], [3]. The antenna is connected to the RF circuitry at points A, B, C and D. The antenna is fed at point D and shorted to ground at point B. The impedance at point A controls the antenna resonant frequency. With a high impedance at this point (i.e., a MEMS device in the OFF state), the slot in the antenna acts as an inductor, reducing the resonant frequency and allowing operation in the low frequency band, 824-960MHz. In this condition the antenna resistance is transformed up by the provision of a low impedance at point C (i.e., a MEMS device in the ON state connected to ground). A double-tuning capacitor is also introduced at the feed when point C is switched to ground in order to tune out the shunt inductance of the feed and shorting tabs and, hence, to extend the bandwidth. With a low impedance at point A, the antenna resonates at a high frequency. In this mode the slot provides an upwards impedance transformation, hence point C is switched to a high impedance. In the high frequency mode, the resonant frequency can be shifted slightly higher with capacitive loading (i.e., additional MEMS devices) at point A. The antenna has four operational modes, the simulated impedances of which are shown in Fig. 3.

Figure 3: Simulated S11 in the 824-960MHz (black, solid), 1710-1880MHz (grey, solid), 1850-1990MHz (black, dashed) and 1920-2170MHz (grey, dashed) bands. All modes, with the exception of the 1920-2170MHz band, have an S11 of –6dB or better (referred to 50 Ohms). The resonant frequency of the 1920-2170MHz mode is deliberately designed to be too high, to allow DC tuning to a lower frequency (when an S11 of –6dB is achievable).

In addition to operation over a wide overall frequency range, the SAR of the antenna can be significantly less than that of conventional antennas in the high frequency modes. This is because the entire antenna is used in all bands, whereas for conventional antennas an additional local resonance – such as that produced by a parasitic element - is often employed to give extended high frequency bandwidth. This is illustrated in Fig. 4 (a), where the SAR of a tri-band (880-960MHz and 1710-1990MHz) conventional antenna of dimensions 40 x 22.2 x 8mm is simulated using the procedure outlined in [4]. It can be seen that the SAR contours are localized between the driven antenna and the parasitic element (in the top left corner of the PCB). Fig. 4 (b) shows contours for the MEMS switched antenna. Clearly, the SAR contours are less localized and the maximum SAR is significantly lower. The maximum SARs are 13.4W/kg and 9.3W/kg for the conventional and MEMS switched antennas respectively, despite the fact that the MEMS switched antenna has a volume that is approximately only 55% of that occupied by the conventional antenna. The SAR in the low-band is approximately the same for both designs. Measured results, with some capacitive and inductive matching applied in simulation in the low- and high-bands respectively (which mainly compensate for un-simulated effects of bond wires) are shown in Figure 5.

(a) (b) Figure 4: Simulated SAR at 1845MHz (W/kg with a 1cm3 averaging volume and an input power of 1W): (a) conventional antenna, (b) MEMS switched antenna

0

S11 (dB)

are 1840-2151MHz, 1849-2156MHz and 1901-2185MHz for the UTRA bands III, II and I respectively (fractionally, approximately 15%). These bandwidths are slightly wider than those simulated. Resonant frequency shifts are clearly observed in the high frequency modes, though the magnitudes of the shifts are less than simulated.

-5

-10

The differences between simulation and measurement are attributed predominantly to uncertainties in the capacitances of the MEMS devices and the long (un-simulated) bond wires used.

-15

-20

-25

3 A Five-band, Seven-mode Reconfigurable Antenna and Antenna Interface Module 0.8

1

1.2

1.4

1.6

1.8

2

2.2

Frequency (GHz)

Figure 5: Measured S11 after matching (shown below –6dB) for band V/VIII (black, solid), band III (dark grey, solid), band II (black, dotted) and band I (light grey, dotted). In the low-band the antenna is matched to an S11 of better than –6dB from 748-912MHz. This represents a fractional bandwidth of 19.9%. Another identical antenna with MEMS having slightly different capacitance values achieved 765950MHz (fractionally, 22%) without any matching, showing that a wide bandwidth resonance is obtained. The centre frequencies are somewhat lower than simulated.

In this concept the 50 ohm interface that normally exists between the antenna and the rest of the RF circuit is removed and the antenna and passive RF circuitry are co-designed in an optimal way. The basic structure of the antennas and handset is shown in Figure 6. Diplexing is achieved by using two antennas – one for the low-band group and another for the high-band group (shown in Figure 1). The antennas have no slots on their top surfaces to minimize the detuning effect of the user’s hand [3]. The functionality of such slots is implemented in a more optimal way in the AIM, allowing the antenna to be reconfigured using band-specific matching.

For the high frequency bands, the resonant frequencies are higher than simulated. With matching, the –6dB bandwidths

Antenna feed Shorting pin with via to ground

AIM

High-band antenna

Bottom PCB with ground plane on the back

Low dielectric constant, low loss spacer

Figure 6: Basic handset incorporating two antennas and an AIM

Flex PCB extending on to top and bottom PCBs Copper antenna formed on flex PCB (low-band)

Top PCB (to represent a handset cover)

Antenna Interface Module (AIM)

SW1/1 GSM850 TX 824-849 GSM900 TX 880-915

SW1/2

Low bands GSM900 RX 925-960

SW1/3

detector GSM850 RX 869-894 UTRA V RX 869-894

SW1/4

UTRA V TX 824-849

SW2/1 GSM1800 TX 1710-1785 GSM1900 TX 1850-1910

SW2/2

High bands

GSM1900 RX 1930-1990 UTRA II RX 1930-1990 SW2/3

UTRA II TX 1850-1910

detector

UTRA I RX 2110-2170

SW2/4

UTRA I TX 1920-1980

Figure 7: Overall system schematic showing The overall system schematic, including a simplified circuit diagram of the components within the AIM is shown in Figure 7. All measurements of the antennas and AIM are performed in a spherical near field chamber. Prior to measurement, the handset is covered with Rohacell foam in order to represent the plastic and air gap that are present in commercial phones. The measured efficiencies of the antenna and AIM combination - averaged over the frequency band specified are summarized in Table 1. These efficiencies include mismatch and imperfect isolation between ports (which are terminated in 50
loads). For user interaction, the VSWR is measured with the “worstcase” condition that the user’s hand completely covers the antennas. The performance is worst for the lowest frequencies of operation in the low and high frequency bands, i.e., GSM850 TX and GSM1800 TX, since the user’s hand inductively tunes the antenna to a lower frequency. In effect, the inductive matching provided within the AIM for these bands is no longer required. Opportunely, these bands have a shared input port with higher frequency modes: GSM900 TX and GSM1900 TX respectively. Hence, if adaptive switching can be achieved when the VSWR exceeds a certain threshold,

the worst-case VSWR can be kept below 4.2:1 for all Band

Frequency Range (MHz)

Free Space Efficiency Sim. Meas.

Max. VSWR with Hand 7.91 3.33 3.35 4.13

GSM850 TX 824-849 48 54 GSM850 TX * GSM900 TX 880-915 55 53 UTRA V TX 824-849 31 34 UTRA V RX 869-894 37 44.5 3.89 GSM850 RX GSM900 RX 925-960 50 45 2.84 GSM1800 TX 1710-1785 61 58.5 4.91 GSM1800 TX * 2.34 GSM1900 TX 1850-1910 59 65 2.89 UTRA II TX 1850-1910 41 40.5 3.92 UTRA II RX 1930-1990 36 38.5 2.07 GSM1900 RX UTRA I TX 1920-1980 40 46.5 3.77 UTRA I RX 2110-2170 30 32.5 3.54 Table 1: Average free space efficiency and VSWR when covered by a hand (* indicates bands that are adaptively switched)

bands/modes. This represents a significant improvement over conventional antennas, where the worst case is around 12:1 in the low-band and 6:1 in the high-band. This may improve or degrade when a conventional front-end module is added. The detection circuitry necessary for adaptation was not implemented in the measured demonstrator. However, simulations of the overall system performance with measurements of the phone held by 63 users and a selfcontained adaptive detection and control circuit were simulated. Typical results – for the GSM850 TX band - are indicated in Figure 8. j1

Conclusions Two reconfigurable antenna systems, both capable of operation in five cellular radio bands are presented. Results show that the two approaches explored - based on utilising new technologies (in this case RF MEMS) and on the codesign of the antenna(s) and RF front-end circuitry – yield overall improvements in performance and/or size. Combinations of such methods are likely to be used in future SDRs and cognitive radios.

References [1]

j0.5

j2

[2] j0.2

[3] 0.2

0.5

1

2

[4] -j0.2

-j0.5

-j2

-j1

(a) j1 j0.5

j2

j0.2

0.2

0.5

1

2

-j0.2

-j0.5

-j2

-j1

(b) Figure 8: S11 of the antennas and AIM in the GSM850 TX band. (a) without adaptation, (b) with adaptation

P.J. Massey and K.R. Boyle, “Controlling the effects of feed cable in small antenna measurements”, in Proc. 12th Int. Conf. on Antennas and Propagation, 2003, pp. 561 -564. K.R. Boyle and L.P. Ligthart, “Radiating and Balanced Mode Analysis of PIFA Antennas”, IEEE Trans. Antennas and Propagat., vol. 54, no. 1, pp. 231 – 237, Jan. 2006. K.R. Boyle and L.P. Ligthart, “Radiating and Balanced Mode Analysis of User Interaction with PIFAs”, in Proc. IEEE Antennas and Propagation Soc. Int. Symp., vol. 2B, 2005, pp. 511–514. P.J. Massey, “Finite element simulation of SAR”, presented at IEE Antenna Measurement and SAR Seminar, Loughborough University, 28-29 May, 2002.