PN Offset Planning Strategies For Non
Transcription
PN Offset Planning Strategies For Non
PN OFFSET PLANNING STRATEGIES FOR NON-UNIFORM CDMA NETWORKS Chu Rui Chang, Jane Zhen Wan and Meng F. Yee NORTEL Wireless Engineering Services Richardson, TX 75083-3805 Abstract - This paper presents a novel methodology in planning the PN-offsets of a highly non-uniform CDMA network and discusses the major related issues. The strategy is applicable to any real life PCS and cellular networks. It is shown that by optimally choosing/ arranging parameters, it is not only possible to mitigate the CO- and adjacent PN-offset interferences, sufficient PN-offset values can also be reserved for new cells to be added into existing clusters for future network development. I. INTRODUCTION In a CDMA system, pilots from all sectors are spreaded by the same PN Short Code, and a mobile distinguishes the pilots via their distinct time shifts of the basic sequence (PN-offsets). Therefore, a careful PN-offset planning should be designed to avoid pilot confusion for a CDMA network. In order to distinguish a pilot from a remote BTS from a multipath component of the home pilot, enough separation between the adjacent E"-offsets must be provided to avoid adjacent PN-offset confusion. In cases where PN-offset values must be re-used, the re-used distance must be large enough to avoid PN-offset confusion. For a uniform network where all cells having similar radii, the PN-offset planning is relatively straight forward since the Short Code should provide sufficient number of distinct PN-offsets to accommodate a large PN re-use pattern, while maintaining a reasonably large adjacent PN separations. However, the virtual PCS/Cellular network is seldom uniform. A typical situation is that the cells in rural and highway areas are much larger than cells in urban areas, as shown in Fig. 1. For extremely non-uniform networks, the task of I"offset planning becomes much more challenging. In the real network, the co-offset confusion is the major concern for small cell clusters, but the adjacent offset 0-7803-3659-3/97 $ 1 0.000 1 997 IEEE confusion is more likely to happen in large cells. If the small cell clusters and large cell clusters are forced to use the same re-use pattern, then it is possible that a sufficient reuse distance cannot be maintained, particularly in between of small cell clusters and large cell clusters, as shown in Fig. 2. Therefore, a more sophisticated planning strategy is needed. In this paper we will derive the criterion for avoiding the CO- and adjacent PN confusions, exam several possible PN-related problems that can happen in a real system and present a practical planning strategy for a non-uniform CDMA network. 11. CRITERION FOR AVOIDING ADJACENT AND CO-PN CONFUSIONS The major task of PN-offset planning is to avoid the PN confusion problem. There is a significant difference between a "I" confusion" and an ordinary "interference".The interference caused by a PN confusion is 19.3 dB worse than an ordinary interference, assuming a 13.3 kbits vocoder is used. A CDMA signal that is not inside the Active Set Search Window (SRCH-WIN-A) cannot be despreaded by the fingers of the RAKE receiver, and will only contribute to the background noise. On the other hand, a CDMA signal in the search window despreaded by a finger will obtain a 19.3dB processing gain. If a remote CDMA signal which does not belong to the home cell but falls into the SRCH-WIN-A and becomes one of the three strongest components, the mobile will treat this remote signal as one of the multipath components of the home signal, and will perform coherent combining on the two unrelated signals, resulting in strong interference (Fig. 3). It is the task of the PN offset planning to avoid the I" confusion. However, the ordinary interference can only be controlled, but not avoided. 1543 the system, the large cells are more vulnerable to adjacent PN confusion than small cells. The adjacent PN-offset confusion can happen due to large differences in the propagation delay. Assuming two pilots with adjacent PN-offsets, when reaching the mobile, the pilot with an earlier phase propagates an extra distance that is large enough so that its phase shifts behind and it falls into the search window of the pilot with a later phase, the mobile will confuse these two pilots as two multipath components of the same pilot. Therefore a necessary condition for the adjacent PN confusion is a sufficient large difference in propagation delay between two signals. IS95 specifies that the minimum separation between two adjacent valid I"-offsets is 64 chips, which is larger than most of multipath spread. To further reduce the chance of the adjacent PN confusion, the minimum separation can be increased by assigning a global parameter PILOT-INC > 1, so that the minimum adjacent PN separation becomes 64*PILOT_INC [chips]. Since the RF wave propagates 244 meters per chip, if two pilots with adjacent PN-offset separation of 64*PILOT_INC chips are to have confusion, the pilot with the earlier phase will have to travel an extra distance of [ "I - AD[m]= PILOT INCX 64 -- x 244 (I) 2 where AD is the difference in propagation distance "1 [ half (2) separation must be: - 11 (4) If Cell A and Cell B have different radii, then R should be the maximum cell radius. To avoid the PN confusion, it is sufficient to require that either the PN separation S is bigger than the difference in propagation distance; or the remote signal is at least 21 dB weaker than the home signal. It can be proven that the minimum adjacent PN L the minimum required physical separation between two BTS that reuse the same PN-offsets to be: D>%+2R 2 Note that of the window size, W A / ~ ,is used since the mobile always centers its Active Set Search Window to the earliest usable multipath component of the arriving signal. S >Rx ( D - d , ) - d , >-WA 2 Since d, I R , from the above formula we obtain rrr between the home pilot and the remote pilot, and WA is SRCH-WIN-A of the mobile. In (l), the effective PN-offset separation S is expressed as: S[chip]= PILOT- INC x 64 - 2 Next we derive the criterion to avoid the CO-PN confusion. In Fig. 4, Cell A and Cell B both uses the same PN-offset (omni-cells are the worst case) and CO-I" confusion problem will happen if the pilots from two different cells fall into the same SRCH-WIN-A of the mobile, and both become one of the three strongest signal components. Note that if one pilot signal propagates much longer than the other, so that the difference in propagation delay causes the remote pilot to "fall out" of the mobile's SRCH-WIN-A, the mobile's search finger will never find the remote pilot and the CO-E" confusion can be avoided. In Fig. 4, assuming the distance between Cell A and Cell B is D. Further assume the worst case scenario that a mobile is located between these two cells and its distance to the home Cell A (the cell in which the mobile establishes its time reference from) is d b and its distance to the remote Cell B is (D - dH). To guarantee the remote pilot "fall out" of the Active Set Search Window, one needs (3) 2 Where, cx is the propagation path loss exponent. From (3) it is clear that the minimum required PN separation is proportional to the cell radius R. If using the same adjacent PN separation through out Assuming the number of cells in a PN reuse cluster is K, then the reuse distance is approximately D=R& (5) which reduces proportional to the cell radius R. Since the search window WA should be independent of R, the required CO-PNdistance, WA D =+2R, 2 does not decrease proportionally with R. Therefore small cells will be much more vulnerable to CO-PN confusion since they tend to have a very small reuse distance. A third type of PN-confusion is the so-called confused handoff. Assume two sectors A and B having the same PN-offsets, Sector A is close to the mobile and is in the Neighbor List, and Sector B is not. However, due to different antenna orientations 1544 or terrain conditions, the arriving signal from Sector B is strong (>T-add), but the arriving signal from Sector A is weak. If the Pilot B falls into the mobile's Neighbor Set Search Window W,, the mobile will confuse Pilot B as Pilot A and will handoff to Sector A, while despreading signal from Sector B. This results in strong forward link interference (Fig. 5). Proper antenna down tilting which reduces the spilled RF energy to remote cells, is an effective way to reduce the likelihood of confused handoff. Increasing the reuse distance, and proper allocation of the PN-offsets also helps to mitigate the problem. 111. I"-OFFSET ALLOCATION SCHEMES In l" planning, the first parameter to be determined is PILOT-INC. The setting tradeoff is that a large PILOT-INC will increase the adjacent PN-offset separation, but will reduce the number of valid PNoffsets, which in turn will reduce the reuse distance. A low setting will do just the opposite. Also there is a lower bound on PILOT-INC, that is PILOT- INC x 64[chips] > max(W,, W,> Where W, ,W, are the search window sizes for the Remaining and Neighbor Set. If this condition is violated, the two adjacent search windows will overlap and the measured value of the PN-offset of a pilot found in the overlapping region will have an ambiguity, as it belongs to both search windows. In the following, we present the I"-planning strategy via an example. Assume the maximum cell radius 5 15 km (= 61 chips), and further assume that for the worst case the path loss exponent a = 3.2, and (3)gives S = 216 chips. If the search window WA = 28 chips, from (2) the PILOT-INC = 4, since (4 x 64 - 14) = 242 > 216. The total number of valid PN-offsets is then 512/PILOT_INC = 128. Assuming 3-sectors/cell, 128 valid PN-offsets will yield a total of 42 cells with distinct PN-offsets. The maximum search window size that can be accommodated for PILOT-INC = 4 is 226 chips, which is more than sufficient for most cases. For uniform networks (ideal case), we may deploy 42 cells/cluster, or 37 cells/cluster plus some reserved PN-offsets for future use. It is desirable to reserve certain number of PN-offsets so that a new cell can be inserted into the existing cluster without disturbing the others. As mentioned previously (Fig. 1, 2), for a highly non-uniform network, it is often not desirable to force large cell clusters and small cell clusters into the same reuse pattern. Typically the shape of the small cell clusters and large cell clusters are different: small cells clusters usually cover an area (2-D), but large cells often cover the highways (1-D), as shown in Fig. 1. We propose to divide the total available PN-offsets into two disjoint sets, one for small cells, and one for large cells. In this way, the small cells and large cells no longer share the PNs from the same set, the reuse distance from large cells to small cells is no longer a concern. Also the small cell clusters can have a different reuse pattem from the large cell clusters. The number of cells per reuse cluster for small cells must be much higher than that for large cells, since small cells are more vulnerable to CO-PNconfusion. Depending on the actual cell configuration, we recommend K, = 27 - 32 cells per reuse cluster for small cells. For large cells, the number of cells per clusterKL can be much smaller, specially if the large cells are used only to cover the highways. Usually K , = 7 - 12 is sufficient. The resulting reused distance must satisfy (4), also K , K L I 42. + In frequency planning, the adjacent channels should not be allocated to the adjacent sectors. The situation is just the opposite for I" planning: the adjacent sectors are _least vulnerable to adjacent PN confusion. A large path difference is a necessary condition for adjacent pilot confusion and it is least likely for two pilots from the adjacent sectors to produce a large path difference. Since the two sectors are facing different directions and in order for two adjacent pilots starting from the same point (BTS tower), and ending at the same point (mobile), one pilot must go through at least one reflection and become much weaker, as shown in Fig. 6. Also the pilot with earlier phase has to travel an extra distance of about 59 km [(l) gives U = (4 x 64 - 14) x 244 = 59km1, which is highly unlikely. Therefore, one should allocate adjacent PNoffsets to adjacent sectors, and allocate non-adjacent PNs to remote sectors. 1545 Fig. 2 shows that it is more likely for large cells with high antennas to cause interference to small cells, since signals from high antenna can propagate much farther. Therefore it is best to allocate the E" with later phases to large cells and those with earlier phases to small cells. This further reduces the likelihood of confusion because the propagation delay will tend to shift the phase behind, so a pilot from a large cell with later phase, after propagation delay, will appear with an even later phase and will be even less likely get confused with the pilot from small cells with early phases. The small cell pilots, with lower antennas, cannot propagate very far. The only exception is that because the Short Code is periodic with the period = 512*64 [chips], so a PNoffset of 512*64 is the same as I"-offset of 0. Thus it is not desirable to use PN offset of 0. We recommend the first I"-offset value starts at 4 [ x 64 chips], and the last one be 508 [ x 64 chips 1. The assignment of PILOTJ'Ns to each cell within a small cell reuse cluster or within a large cell cluster, should follow a consistent fashion. For example, if values are assigned to each cell within a cluster in a spiral fashion (Fig. 2), then the assignment for other clusters should follow the same fashion as well. This produces an approximately equal reuse distance for every cell in the cluster. However, the assignment fashion for large cell clusters and for small cell clusters do not need to be identical, since they use PNs from disjoint sets. IV. SUMMARY OF PN-OFFSET PLANNING STRATEGIES There are three types of I"-offset confusion problem: CO-PN, adjacent I" and confused handoff. I" confusion is more harmful than conventional interference. Propagation delay, search window sizes and path loss exponent are the three key factors in PN-offset planning. Sufficient reuse distance and I" separation, proper antenna down tilt, together with proper allocation of I"-offset values, reduces the likelihood of PN confusion. For non-uniform networks, reuse clusters for large cells and small cells may have different shapes and may use I"-offsets from two disjoint sets; small cell clusters using I"-offsets with smaller values and large cell clusters using larger offset values. The general expression for allocating PILOT-PN to each sector is: {a,p, y } - Sector's Offset (6) = ( P N , ) j + 12k - (8, 4, O} Where (PN, ) is the first I"-offset value from the sets, and k = l , 2, ..., Kj, K j is the number of cells in a reuse cluster. Equation (6) is applicable for both the small cell and large cell clusters. For small cell clusters, the subscript j = S in (6), and for large cell clusters, j = L . Larae Cells in RurallSub A n Small Cells Near the Center Figure 1 A typical situation where the urban areas are covered by large number of small cells and rural/Hwy are covered by much larger cells. Also large cell clusters have different shapes from small cell clusters. F i m e 2 An example of forcing cells with different sizes into the same reuse pattern. In this example K=7 and R=2r.The reuse distance within small cluster is 4.6r, and within large cluster is 4.6R.But from large cell cluster to small cell cluster is only 2.3R. The situation is much worse if R >> r. 1546 Short Code Length = 2 Cell A = 32768 Chips . . .. ... .. . , , , , .. , . ... . .... . . . . .. , 2!5473 Mobile Remote Pilot (phwe advanced) I '. .._ . . '. .. .._ I I I I Cell B ... I I I i Home Pilot (phase hehind) I ~- 1 I I * Pilot A Pilot B T h [ehlp] 4 - Active Set Scar ch Window e Delayed Remote 'lot Home Pilot Active S e t k i n d o w 1 Figure 3 Illustration of adjacent PN confusion, where a pilot with an earlier phase propagated an extra distance and its phase falls into the home pilot's SRCH-WIN-A. F i m e 4 Illustration of CO-PNconfusion: both cells use the same PN-offset and their pilots fall into the same SRCH-WIN-A. Note that the CO-PNconfusion will not happen if the remote pilot travels an extra distance and falls out of SRCH-WIN-A. 0 0 r*. I I I Figure 5 Illustration of confused handoff. Assume the a-sector of both Cell A and Cell B use the same PNoffset. A mobile located in Cell A's y-sector will have Cell A's a-sector as its neighbor. However, if the signal from Cell A's a-sector is weak but the signal from Cell B's a-sector is strong, and if the a-Pilot B falls into mobile's SRCH-WIN-N, the mobile will confuse a-Pilot B as a-Pilot A and will handoff to a-sector of Cell A. Figure 6 Illustration of allocating adjacent PNs to adjacent sectors. If a mobile receives two pilots from two adjacent sectors via a direct path, then there will be no path difference between them. For there to be an adjacent PN confusion, the signal with earlier phase has to go through at least one reflection and becomes much weaker; it also has to travel an extra 59 km. 1547