Automatic Analog Beamforming Transceiver for 60 GHz Radios

Automatic Analog Beamforming Transceiver for 60 GHz Radios Shalabh Gupta University of Cali fornia, Los Angeles, CA 90095, USA Abstract — W e pr opose a transceiver architectur e for automatic beamforming and instantaneous setup of a multi- gigabit-per -second wireless link between two millimeter wave radios. The r etro-dir ective ar chitectur e eliminates necessity of slow and complex digital algorithms required for searching and tracking the directions of opposite end r adios. Simulations predict <5 micr o- seconds setup time for a 2- Gbps bidirectional 60-GHz communication link between two 10-meters apart radios. The radios h ave 4-element arrayed antennas, and use QPSK modul ation w ith 1.5 GHz analog bandwidth. Index terms–– Adaptive arrays, millimeter wave antenna arrays, phased arrays. I. I NTRODUCTION The demand for high speed wirel ess comm unication systems has made it necessary to move to millimeter wave frequencies. Device scaling in CMOS technol ogies has enabled low cost communications system s in the unlicensed 57-64 GHz frequenc y band [1-2]. However, the path loss at these frequencies becom es severe as the effective antenna area scales proportionally wi th the square of wavelength to maint ain omni-di rectional operation. Thus, use of arrayed antennas providing adaptive beamforming becom es very important. Millimeter wave frequencies highly favor use of beamforming because of small antenna elem ent sizes and absence of multi-path effects. Adaptive beamforming provides high link gains because of antenna directivity, and the radios can conform to different signal directions. However, adaptive beamform ing requires com plex digital algorithms for computing the direct ion of signal arrival and transmission to set up a com municat ion link [3]. Conventional algorithms are very slow, and interface between the digital and an alog circuitry also adds significant delays. Even when the radio directions are known, controlling phases of tr ansmit and receive signals to achieve prop er directionality becomes difficult. Retro- directive arrays can be used to overcome these lim itations [4-7]. In the proposed architecture, we exploit ret ro- directivity to achiev e instantaneou s link acquisition , in addition to much sim pler analog im plem entation and automati c phasing of the beamform ing arrays. A retro-directive antenna is an arrayed antenna that transmits in the direction of the in coming electromagnetic Fig. 1. Link setup in the retro-directive radios. After a few round- trips (RT), the signals from the two ra dios become directed towards the each other. signal without prior knowledge of it s direction. Ret ro- directive arrays work on the principl e of phase conjugation, as described in detail s in Section II. Based on the proposed retro-directive radio architecture (described fully in Section III), quick setup of an automati c beamform ing R F link is shown in Fi gure 1. In the absence of an incoming signal, most of the tran smitted power goes omni -directionally . When a signal is present at the receiver input, power is transmitted in the direction of the incoming si gnal. For setup of an R F link, bot h radios initially transmit omni-directionally (as there are n o input signals). Om ni-directional t ransmission i s made possible by random phases at different transm it elem ents. Because of omni-directiona l transm ission, the initial signal power received at the opposite end radios is very weak. However, the transmitters start building directivity towards opposite end radios rapidly in successive transmissions, as they star t receiving more and m ore power after every signal r ound trip (RT). This positive feedback leads to a quick, au tomatically tracking RF link between the two radios. This link setup process is analogous to target acquisition in the retro-di rective noise correlati on radar [8-9]. However, the thres hold power requirement s for setting up the link are much sm aller as compared to the retro-directive radar since there are active transmitters on both ends. Fig. 2. Retro-directive arrays: (a) Van- Atta array achieves retro- directivity by geometrical positioning of transmit and receive elements. (b) In heterodyne mixing approach , phase conjugation is achieved by mixing RF with LO and using pha ses of lower sideband IF for transmission. II. R ETRO - DIRECTIVE A RRAYS The retro-directive arrays, also known as self-phasing antennas, rely on the principle of phase-conjugation (or phase inversion) for their operation. Tradi tionally, t hey have been used as passive transponders, in which any incoming signal is reflected b ack to the source. When the incoming wave is phase conjugated, both the incoming and the transmitted waves hav e parallel wave fronts, b ut travel in opposite direct ions. Phase conjugation can be achieved by Van Atta array arrangement, or by using heterodyne mixing, as discussed in the following subsections. A. Van Atta Array In Van Atta configuration [4] , shown in Fig. 2(a), the signal received at one elemen t is transm itted by another one which is at a “conjugate” position t o it, after additi on of a constant phase delay due to i nterconnects and intermediate circuitry. In the figure, for example, signal received at the first element is retransmitted by the fourth element, that received at th e second element is transmitted by the third elem ent, and so on. Starting from the source and returning back to the source, t he signal undergoes equal phase delays for different paths, as L1 + L4 = L2 + L3, assuming adjacent elem ents are equidistant to each other. Hence, there i s constructive interference of signals transmitted by all elements at the source resulting in directivity towards it. The sign al can be transmitted passively or there can be active com ponents in the signal path to modulate and/ or amplify the signal before transmission. B. Retro-directivit y by heterodyne mixing Use of heterodyne mi xing technique to achieve phase conjugation was first proposed in [5] . As in Figure 2(b), when the wave-front from the source is incident at non- zero angle, each array elemen t receives the RF with a different phase due to path length difference. Elem ent i is at distance L i away from the source and the signal obtained at the i th antenna element becomes cos( ω RF t − β L i ) . This signal is mixed with a local o scillator (LO) frequency ω LO such that ω LO > ω RF , and low pass filtered to obtain the transmi t signal at di fference frequency ω ' RF = ω LO −ω RF , as cos( ) cos( ) cos( ' ) LPF R Fi L O R F i tL t tL ωβ ω ω β −× ⎯ ⎯ ⎯ →+ (1) where, β = ω RF /c, c = velocity of light. The resultant signal, when transmitted back b y element i , has propagation constant β '= ω ' RF /c , and becomes cos( ω ' RF t + β L i − β 'L i ) due to the free space propagation delay. When the receive and transmit frequencies are roughly equal, i.e., when ω LO ≈ 2 ω RF , the two propagation constants also becom e roughly same, i.e., β≈β '. As a result, back at source, the signal phase becom es independent of path differences from differe nt elem ents, resulting in constructive interference, an d hence, directivity towards the source. To achieve retro-directivit y, other frequency plans can be used [10], wi th the goal being that the excess phases added in the receive paths to the elements are subtracted from the transmit paths for the direction of interest. If the Fig. 3. Proposed 60 GHz radio architecture with retro-dir ective antennas. upper side-band signal is used afte r heterodyne m ixing, the transmit array elem ents should be placed in the same sequence as the receive elements, and if lower sideband is used, the sequence must be reversed. It is import ant that the receiver and the transmitter frequencies should be close (if inter-element spacing is same for both transmit and receive signals) to avoid any beam pointing inaccuracies. Also, any mism at ches in phase delays in different channels m ust be minim ized or calibrat ed out. III. R ADIO A RCHITECTURE Figure 3 shows the architecture of the proposed 60GHz retro-directive radio. As an exam ple, the radio shown in the figure receives signals at 58-GHz and transmits at 62- GHz. On the other hand, t he opposite end radio transm its at 58-GHz and receives at 62-GHz. The incoming phase modulated 58GHz RF signal is received by different antenna element s of the radio and is down converted to a convenient 2.5GHz IF frequency in each of the receive chains. After band pass filte ring, phase demodulator (PDM) is used to extract the RF phase for each channel. This phase contains b oth the transmitted data ( x in (t) ) as the high frequency component , and the direction phase φ (t) as the low frequency component which is used for phase conjugation and retro-directivit y. The phase modulat ing digital data should be encoded such t hat it doesn’t have low frequency components, t o avoid interference with the phasing components φ i (t) . For transmitter beam forming, low pass filtered phase sign als are sent to the transmit channels, where the data to b e transmitted is added to them. These signals m odulate the phase of 1.5GHz IF (intermedi ate frequency), which is t hen up converted t o 62GHz RF and transmitted by individual an tenna elements. Combining the hi gh pass data signals together results in receiver beam forming as the noise components are scaled down due to averaging. For i th channel phase, instead of using the low pass filtered phase φ i (t) , Δφ i (t) is used, where Δφ i (t) = φ i (t)- φ 1 (t), for i = 1, 2, .. n. This simple difference operation ensures the there is no drift in the transmit frequency, even if th e demodulated input phases are dr ifting together because of clock offsets between the two radios. This difference operation gives a significant im provem ent over the retro- directive architecture used in [10], an d a lot of instability concerns [11] for a retro-directive link becom e unimporta nt. A key feature of this architecture is that in the absence of an input signal, the power goes omni-direct ionally since the phases to the transm it elements are uncorrelated. Initial omni-directional transmission is crucial as it provides the cuing signal for opposite end radi os to start build up of directivi ty. The transmit and receive frequencies are intentionally kept unequal to ensure that there is a si gnificant isolati on between the transmitter and the receiver, as both have to operate simul taneously. The frequency plans can be varied a lot dependin g on availability o f the circuit level choices. Different modulati on schemes can be used for high bandwidth data modul ation, but low bandwidth phase modulation and demodulati on are still required for phasing the antennas to achieve direct ivity. Low pass bandwidth of the phase demodulator dict ates the link setup time. Sm aller bandwidth result s in longer locking times but m ore stable operati on. IV. S IMULATIONS AND R ESULTS The radio system discussed in previous section is simulated using Simulink so ftware (from Mathworks), with discrete time si mulations and fixed ti me steps of 5- ps. One of the radio trans ceivers receives signal at 58- GHz and transmits at 62-GHz . The other radio receives signal at 62-GHz and transmits at 58-GHz. Each of the transceivers has 4 receive antenna elements and 4 transmit elements, aligned linearly at λ /2 mutual spacing. Both the radios are placed at 42º angles to the broadside direction of the arrays. The path loss is calculat ed based on a distance of 10 meters which corresponds to a 33-ns propagation delay between the radi os (a propagation delay of 35-ns was used in the simulations t o account for additional group delay of el ectronics). For simplici ty, the signals are modulated and demodulated di rectly at the carrier frequencies instead of first convert ing them to IF. Fig. 4. Signal amplitudes at the input of each receive element as a function of time for the two radios are shown. The stable envelopes beyond 3 micro-seconds indicate that the two radios have locked retro- directively. Fig. 5. Eye diagram of the recei ved 2 Gbps QPSK signal obtained after averaging the four phase demodulator channel outputs. The power transmitted by each an tenna element is set as 0-dBm. Since 60-GHz radios assume direct line of sight transmission, use of completely o mni-directional antenn as is useless. Therefore, the an tenna elements are assumed to have a gain of 4-dBi, which lim its the operation of the system to a solid angle of about 1.6 π steradians. This gives a path loss of 75-dB for a 60-GHz carri er and element-to-element transmissi on. The receiver is assumed to have a noise figure of 3-dB and 1.5-GHz band pass FIR filters are used before phase demodulation. Low pass filter used at the phase demodulator output has a 2-MHz bandwidth. Figure 4 shows the received signal amplitudes at the inputs of antenna elements of the two radios. Because of retro-directivity, t he signal strength builds up very quickly (within 3 micro-seconds). The link acquisition times vary depending upon the si gnal strengths, and at extremel y low power levels, the directionality lock ing between the two radios cannot be achieved. Figure 5 shows the eye diagram of the phase demodulated 2-Gbps data signal. This signal is obtained by combining the demodulat or outputs of indivi dual channels and improves the SNR by √ N, where N is the number of elements. The improvement corresponds to the receiver side beamforming gain. V. C ONCLUSION The proposed radio architecture can be very useful in instantaneous setup and tracking of multi-gigab it-per- second communication link at millimeter wave frequencies. A 60-GHz 2-Gbps l ink is demonstrated usi ng computer simulations for t wo radios at a distance of 10-m and total transmit power of 6-dB m (combined for all 4 array channels) using QPSK modulation. The li nk acquisition tim e is found to be 3 microseconds. R EFERENCES [1] C. H. Doan, S. Emam i, A. M. Niknejad, and R. W. Brodersen, "Millimeter-wave CMOS design," IEEE J. Solid-State Circuits , vol. 40, pp. 144-155, Jan. 2005. [2] S. Reynolds, et. al, "A silicon 60GHz receiver and transmitter chipset for broadband com munications," IEEE J. Solid-State Circuits , vol. 41, no. 12, pp. 2820- 2830, Dec. 2006. [3] B. D. V. Veen and K. M. Buckley, “Beamforming: A versatile approach to spatial filtering,” IEEE ASSP Magazine , vol. 5, no. 2, pp. 4-24, Apr. 1988. [4] L. C. Van Atta, “ Electromagnetic Reflector,” U. S. Patent No. 2,908,002; October 6, 1959. [5] C. Y. Pon, “Retrodirective arra y using the heterodyne technique,” IEEE Trans. Antennas and Propagation , Mar. 1964, pp. 176- 180. [6] K. M. K. H. Leong and T. I toh, “Full-duplex retrodirective array using mutually-exclusive up link and downlink modulation schemes,” in 2004 IEEE MTT-S International Microwave Symposium Digest , June 2004, vol. 3, pp.1695- 1698. [7] S. Lim and T. Itoh, “60-GHz Retrodirective Array Sy stem with Efficient Power Managem ent,” Proc. 2006 36 th European Microwave Conference , pp. 314-317, Sept. 2006. [8] E. R. Brown, A. C. Cotler, S. Gupta and A. Um ali, “ First demonstration of a retr odirective noise-correlating ra dar in S band”, in 2004 IEEE MTT-S International Microwave Symposium Digest , June 2004, vol. 2, pp.751-754. [9] S. Gupta and E. R. Brown, “ Noise-correlating radar based on retrodirective antennas,” IEEE Transactions on Aerospace and Electronic Systems , vol. 43, no. 2, pp. 472- 479, April 2007. [10] L. D. DiDome nico, and G. M. Rebeiz, “Digital com munications using self-phased arrays,” IEEE Transactions on Microwa ve Theory & Techniques , vol.49, no.4, pt. 1, April 2001, pp.677-84. [11] L. D. DiDomenico, and G. M. Rebeiz, “Frequency stability in adaptive retrodirective arrays,” IEEE Transactions on Aero space & Electronic Systems , vol.36, no.4, Oct. 2000, pp.1219-31.