Spatial topology and dynamics in CDMA cellular radio

SPATIAL TOPOLOGY AND DYNAMICS IN CDMA CELLULAR RADIO Joseph Shapira and Roberto Padovani QUALCOMM Incorporated 10555 Sorrento Valley Road San Diego, CA 92121-1617 1. CDMA Network Layout Overview i Cellular radio architecture has been devised in order to increase the information transfer capacity for many users in a common service area. It relies on the isolation caused by the attenuation of the transmission that propagates in the terrestrial path, which allows a reuse of the same channel beyond a certain guard distance. Systems relying on frequency channel allocation for each user (FDMA, or FDMA/TDMA) require such a guard zone and their cells are therefore arranged in lattices, where the channels are reused beyond these guard distances, sufficient to attenuate the signal below a discernible interference. The reuse lattice may consist of 12, 9, 7 or 4 cells, depending on the isolation margin C/I required to each system in the given fluctuating channel [1,2]. The cellular CDMA system differs from the above architecture. The same frequency band is shared by all users, and is reused in each cell. Codes are used in spreading the signal over the band, and provide the processing gain for each user. These codes are in abundance and do not require reuse. The cell arrangement is therefore not bound to any structured lattice, but designed so as to bound the interference accumulated from the other transmitting users, in the same cell and in the surrounding cells. The interference neighborhood affecting the cell design and capacity is much smaller and hardly extends beyond one tier of cells around it. The interference is controlled in the CDMA system by: Control of the power transmitted by users in the cell so as to minimize interference. b. Control of the cell-site receptivity to interference by spatial isolation, angular and radial. j Eb No is the interference introduced by surrounding cells/ sectors, referred to the self interference in the cell/ sector, is jamming power, referred to the system noise N0 W, is the energy per bit, is the noise spectral density The power control of all the transmitters in the cell causes the interference to change proportionally to the change of q, the signalto-noise ratio. In an environment where all the cells are equally loaded and controlled to the same SNR q, the interference iis also proportional to q. When solved for q [4] q= h= 1+j 1+j = W No W No Rb• Eb -m(1+i) Rb• Eb (1-h) m N o W Rb• Eb (1+i) (2) a. The transmission power of each user in each cell is controlled so as to be received by the cell-site with equal signal to interference ratio, minimal as necessary to maintain a required level of service. The angular isolation is achieved by directive antennas. The capacity gain by sectorization in CDMA is proportional to the (azimuthal) gain of the cell-site antennas, and equals almost the number of sectors. the capacity bound in each cell appears as a pole in the power-set . equation in Figure 1 2. The Power Control Equation The CDMA capacity equation (3) for the reverse (mobile to cell-site) link may be recast into W qR Eb b = 1+m(1+i)q+j No Figure 1: Required SNR at the Cell-site as a function of the load. Here h is the fractional load of the cell, while (1) where W Rb is the CDMA bandwidth, is the data rate, q is the signal received at the cell-site from a single user, referred to the system noise N0 W, m is the number of equivalent full-time active radio channels, W No 1 mpole [h=1] = R • E (1+i) b b (3) is the upper bound on equivalent active radio channels. The actual number of calls in process n is derived therefrom, using the voice activity factor d : n= m / d A convenient nominal capacity is set at h= 0.9, in which case each user is received at the cell-site with SNR=-4.2 dB. This nominal setting is a trade off of capacity vs. sensitivity of the power control to changes in the load. The required SNR per user at the cell-site, when the cell is only half loaded, h=.5, is SNR= -11.2dB, which translates into larger coverage, as depicted in Figure 2. R R nominal load α =3 α =4 η Figure 2: The radius of the cell coverage vs. load, for propagation exponents 3 and 4 The total received power at the cell-site, Ptotal h sT = N W = m• q (1+i) = 1-h o (4) depicted in Figure 3, reaches 9.5 dB above the noise at the nominal load. s T [dB] η Figure 3: The total received power at the cell-site, normalized to the noise vs. load 3. Cell Coverage and Signal Quality Coverage is ultimately determined by the quality of service achieved: voice quality, hand-off smoothness, and set up time. These depend not only on the link budget for the total power, but also on the channel dispersion and fluctuations, for which the different systems provide different means of mitigation. Only digital systems have the means to measure the service quality, e.g. the voice quality, through the error rate in the received message. The CDMA system further uses this measurement to control and maintain the service quality, as will be explained later. Analog systems, however, rely on the link budget only for their coverage predictions. Total power measurements in narrow-band systems are averaged over a few wavelengths, so as to average the detailed interference pattern produced by the multiple scattering [1,2]. This procedure differentiates between the forward propagation mechanisms, essentially including the first Fresnell zone, which is frequency insensitive [5], and the fast fade pattern. The latter is frequency sensitive and is averaged out when total power is measured for a signal bandwidth broader than the coherent bandwidth of that area [6]. The CDMA signal bandwidth of 1.23MHz exceeds by far the typical coherent bandwidth in the mobile radio channel [7] , and the measurement of its total power is thus equivalent to the above mentioned narrow-band measurements. These are used for comparative coverage measurements and planning. The CDMA system offers many diversity mechanisms to smooth the fluctuating nature of the mobile channel: • The soft hand-off is a two-cell diversity by which the mobile is received by two cells and the better transmission selected. This continuous process, carried over a predetermined power ratio between the two cells' pilots reception at the mobile, smoothes local shadowing and fading effects, otherwise causing alternate hand-offs or call drops in other systems. This allows the link budget required to the boundary of the cell to be lower by about 2 dB than that in an isolated cell. Systems implementing hard hand-off procedures require typically 6 to 10 dB excess power, to guarantee positive hand-off well beyond the cell boundary. This by itself gives the CDMA 8 to 12 dB advantage in coverage. • The Eb /N0 relates to the SNR q at the cell-site via equation (2) (see also Figures [1,3]), which depends on the load. This is now redrawn in Figure 6, from Equations (2,4) for different number of calls in an isolated cell, as verified by the major CDMA system test. The multipath delay spread of the mobile channel is used to benefit in the CDMA system. The long delayed peaks are each tracked separately by a dedicated correlator, and optimally combined. The bandwidth of the system is sufficient to resolve the delayed peaks and also to resolve the cluster adjoining the first peak to the extent that its fluctuation is tamed considerably in most cases [8]. The code and interleaver spread out and correct the error bursts in the remaining fades at high speed, while the power control loop tracks the fluctuations at low speed. This effect is exemplified in Figure 4. Figure 6: The observations on the relative cell coverage are summarized in Table 1. Table 1: Link Budget and Radius of Coverage Different Systems Fo rward FER i s 0 .4 % Frame Perfo rman ce v s. Sp eed Rev erse FER i s 0 .8 % Power received at the cell-site vs. Eb /N0 Validation point are marked. 2 0 M o b i l es p er Secto r for Portables of 2 .5 0 % ITEM CDMA CDMA FM TDMA CAPACITY PER CELL *10 *16 *1 *3 PORTABLE ERP (dBm) 23.00 23.00 28.00 28.00 SY STEM NOISE(dB) -105.00 -105.00 -122.76 -122.76 NOMINAL THRESHOLD,AWGN (dBm) -118.60 -117.00 -116.00 -110.00 THRESHOLD IN FADES (dBm) -116.00 -112.50 -110.00 -103.00 LINK BUDGET(dB) 139.00 135.50 138.00 131.00 LINK BUDGET, 10dB HANDOFF MARGIN (dB) 141.00 137.50 128.00 121.00 Cell radius, r^-4 at the margin 3.16 2.59 1.50 1.00 CEll radius, r^-3 at the margin 4.64 3.55 1.71 1.00 LINK BUDGET, 6dB HANDOFF MARGIN (dB) 141.00 137.50 132.00 125.00 Cell radius, r^-4 at the margin 2.51 2.05 1.50 1.00 CEll radius, r^-3 at the margin 3.41 2.61 1.71 1.00 Fo rward L i n k 2 .0 0 % Rev erse L i n k 1 .5 0 % 1 .0 0 % 0 .5 0 % 0 .0 0 % 0 -1 0 1 0 -2 0 2 0 -3 0 3 0 -4 0 4 0 -5 0 50+ Sp eed (mp h ) Notes: • The CDMA portable transmits less power • The bandwidth of FM , TDMA is 30kHz • The thresholds for FM and TDMA are from the standard • The corresponding Eb /N0 for CDMA are 5.5 and 6.8 dB. Figure 4: Frame error rate vs. speed The required average bit energy, Eb /N0 is the measure of the channel quality in CDMA, similar to the C/I measure in narrow band systems. The powerful diversities and processing in the system lead 4. Cell Radio Design to low Eb /N0 requirements, that vary only a little over a variety of channel conditions. That is depicted in Figure 5. An exceptionally benign environment required Eb /N0 =5.5 db The capacity of any cellular network is determined by the isolation maintained between the desired link and the interference. The slope of the path-loss beyond the cell coverage is thus the major constraint on the cellular radio design. Asymptotic slopes of 30 to 40 dB per decade are typical to the outdoors environment. Obtaining minimum slope within the cell is a desirable feature, especially for portable units, in order to minimize their power consumption. Cell RX Eb/No Statistics 62 Mobiles in 3 Sectors 0.14 Mean = 0.4 [dB]; Std Dev = 1.1 [dB] 0.12 Average Eb/No = 6.8 [dB] 0.1 0.08 0.06 Mean = 0.5 [dB]; Std Dev = 1.7 [dB] 0.04 0.02 0 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 Received Eb/No - Setpoint Eb/No [dB] Only RX Full Rate All RX Rates Figure 5: Statistics of received Eb /N0 rate the cell-site. 62 mobiles in 3 sectors 6 The CDMA system differs from the FM or the TDMA systems in that it accumulates the interference from a multitude of transmitters, each disadvantaged by the code rejection, within its own cell and its neighborhood. The isolation design in this case is an integration process over an interference spatial distribution rather than the avoidance of a small number of interfering sources, roaming in distant cells. The interference neighborhood that affects the cell is much smaller in CDMA. More than 90% of the interference, in a power controlled CDMA environment, comes from the first tier of cells around the cell in question. The interference produced by each neighboring cell depends on its actual load at the moment, an effect further reducing the total interference and the importance of local interference peaks. Interference within the cell may be screened by using directive antennas at the cell-site. The amount of interference accumulated is directly proportional to the total beam area, or inversely proportional to the (azimuthal) antenna gain. Therefore the capacity is proportional to the antenna gain, and is insensitive to the detailed radiation pattern and sidelobe structure. Coverage overlap between antenna beams causes interference spill-over, however, which reduces the capacity. The computation of the sectorization gain for sector antennas with [sin(x)/x] patterns shows a sectorization efficiency of 86% or higher for 3 dB overlap between the beams in Figures 7. P [dB] NoW Pm NoW q2 q1 C el l -si te #1 C el l b o u n d ar y C el l -si te #2 b(dB) Figure 8: The cell boundary and its dynamics With Softer handoff A change in the threshold setting of either cell causes the boundary to shift so as to shrink the coverage of the cell with the higher q. Load is thus automatically shared throughout the system, and cells that are heavily loaded shrink and hand users off to the neighboring cells. This process is instantaneous, as both cells are engaged to the user while within the soft hand-off window. Single sectored cell in an omni cluster Uniformly sectored cluster Figure 7: Sectorization efficiency vs. b, the beam overlap level between sectors. An external source of interference, whether penetrating the cell or radiating from the outside in the CDMA band, is mitigated by raising the power of all users and cell-site accordingly . Users on the cell margin might reach their maximum possible or allowed transmission power, beyond which they loose the ability to communicate with the cell-site and drop out (if the cell is isolated) or hand-off to a neighbor cell . This has been demonstrated during the major CDMA test, when one sector in a three-sector cell was jammed. The cell was surrounded by other cells, each loaded to about 10 times the equivalent FM capacity. The jammer increased its power to be received at the cell-site at 30 dB above N0 W and then exercised Highly cluttered and obscured areas might not enable a reasonably uniform coverage by a single radiator. Such is the case in buildings and in other packed neighborhoods. Distributed radiators, which could certainly reduce the path-loss variation, are difficult to apply in narrow-band systems, unless extreme care is taken to avoid coverage overlap. The interference between the radiators in such areas produces a highly fading environment. This is altogether avoided in CDMA, when time delay is inserted between these radiators, enough to be resolved by the rake receivers. The overlap is thus used to advantage, as it adds a degree of diversity combining to the system. gating with 15 dB and rates of .1,1,10 and 100 Hz. The results are summarized in Figure 9, demonstrating both the robustness of the individual cells and the load sharing capability. 5. The CDMA Network Dynamics The setting of adjacent cells is interconnected via the hand-off condition , which determines the cell boundary. A mobile in the hand-off zone is controlled by both adjacent cells. Its power level is always set to match the lower power requirement between these cells. The boundary between the cells is thus defined by equal power requirements from both cells. Obviously this balance depends on the power setting in each cell, which in its turn, depends on the number of active users in the cell and on the additional interference from the neighboring cells . The boundary between the cells is defined as the equilibrium point, where the transmission power required from the user by both cells is the same, Pu 40 35 30 25 20 Jammer cycle time ( complete on/off period) 10 sec........1 sec...........0.1 sec..................0.01 sec............... 15 10 At the boundary where L1 , L2 are the path-loss from the user to each cell-site. Assuming Time into test (sec) alpha sector calls system call drops alpha jammer dB Figure 9: Cell reaction to jamming - test results 6. Heterogeneous Cell Clustering -a Li = Ai Ri where A, a are constants of the propagation model, and R is the cell radius - the boundary condition becomes a a q 1 R1 = q 2 R2 depicted in Figure 8. Cell size, and coverage vs. capacity are design parameters, optimized locally for a CDMA system, by proper location of the antenna, its height and its beam shape. The boundaries are defined at the intersection of their respective path-loss curves. This is shown in Figure 10, demonstrating the clustering of microcells with large cells. PCN cells, with lower power, may also interconnect with mobile cells, though while a mobile may roam through, a portable might not be able to reach the required power level to roam into the large cell. The soft hand-off, which extends over distance, allows the roaming of mobiles through microcells without dropping calls due to fast handoff, which appears to be a problem with other systems. 740 720 700 680 660 640 620 600 580 560 540 520 500 480 460 440 420 400 380 360 340 320 300 280 260 240 220 200 180 160 140 80 120 60 100 0 40 0 (5) 20 q1 q2 Pu L1 = L2 = N o W 5 I[d B ] Microcell R[m] b) Figure 10: Isolation profile and clustering of cells a) Isolation profiles b) Microcell within a large cell 7. Summary The CDMA system is adaptive to the channel and to the traffic load, and therefore offers flexibility in the network design and operation: there is no lattice arrangement of cells. • • • • The cell coverage is larger than that of other systems, and a smaller number of cells maybe used. Sectors are individually designed. The load is shared through the network. No frequency planning is necessary. References 1. 2. 3. 4. 5. 6. 7. a) William C. Jakes, Jr., Microwave Mobile Communications, John Wiley & Sons, New York, 1974. William C.Y. Lee: Mobile Cellular Communication Systems, McGraw Hill, 1989. Klein S. Gilhousen, Irwin M. Jacobs, Roberto Padovani, Andrew J. Viterbi, Lindsay A. Weaver, Jr., and Charles E. Wheatley :On The Capacity of a Cellular CDMA System, IEEE Trans. VT-40, pp.303-312, May 1991. Samir Soliman, Qualcomm internal memo,1991. William C.Y.Lee: Theory of Wideband Radio Propagation, Proceedings of the 41st VTS Conference, pp.285-287, St. Louis, May 1991. A.J. Giger, G.D. Alley, P.L. Sullivan, D.E. Major, "Time and Frequency Fluctuations of Microwave Interference Due to Terrain Scatter," Globecom '86, 43.5.1. Joseph Shapira: Channel Characteristics for the Land Cellular Radio, and its System's Implications Submitted for publication in the IEEE Trans. AP, July 1991.