HSPA+ Advanced Smart Networks: Multipoint Transmission

Transcription

1 Qualcomm Incorporated February 2011

2 Table of Contents 1. Introduction Multipoint HSPA Description... 2 Single Frequency Multipoint HSPA... 2 Dual Frequency Multipoint HSPA Advantages... 4 Improvement of cell edge performance... 4 Load balancing across sectors and frequency carriers... 5 Leveraging RRU and distributed NodeB technology System Design Considerations Conclusion /2011 page i

3 1. Introduction Rapid wireless subscriber growth coupled with the necessity to provide desired levels of user experience for broadband multimedia applications has led to the continuous evolution of HSPA technology. Some of the initially developed HSPA+ features, such as Higher Level Modulation and MIMO, focused on increasing spectral efficiency by taking advantage of favorable channel conditions when available. More recently introduced features such as Dual-Carrier HSPA (DC-HSPA) and Multicarrier HSPA (MC-HSPA) exploit the multiplexing gain achievable when transmitting bursty data over multiple frequency carriers. All these features were enabled by enhancing radio transceiver chain design, both on the network side and in the mobile device (UE). Even with advanced radio transceivers in place, there is still scope for improving cell edge user experience. Even though many current HSPA networks are plagued with issues of capacity saturation and inadequate cell edge performance, neither the capacity nor the quality potential of the network as a whole is fully reached. Adjacent sectors and frequency carriers are often unevenly loaded; different topological layers in the network (e.g. macro, pico, femto) are sometimes unevenly loaded as well.most UEs with poor serving cell data rates can often receive signals from other cells which are yet fully exploited in HSPA+. The next step in the evolution of HSPA must take all this into consideration. Multipoint HSPA is a new feature currently under study in 3GPP with the objective of addressing some of the aforementioned issues while leveraging existing transceiver capabilities of the network and UEs. It offers advantages such as: - Improved user experience at the cell edge - Efficient and dynamic load balancing across sectors in single-carrier deployments - Efficient and dynamic load balancing across sectors / carriers in multicarrier deployments - Leverage DC-HSPA / MC-HSPA capabilities of the network and UEs by means of incremental hardware and software upgrades 2/2011 page 1

4 F1 HSPA+ Advanced Smart Networks: Multipoint Transmission 2. HSPA Multipoint Transmission Single Frequency Multipoint HSPA The simplest form of Multipoint HSPA, Single Frequency Dual Cell HSPA (SFDC-HSPA), can be seen as an extension to the existing DC-HSPA feature. While DC-HSPA allows scheduling of two independent transport blocks to the mobile device (UE) from one sector on two frequency carriers, SFDC-HSPA allows scheduling of two independent transport blocks to the UE from two different sectors on the same carrier. In other words, it allows for a primary and a secondary serving cell to simultaneously send different data to the UE. Therefore, the major difference between SFDC-HSPA and DC-HSPA operation is that the secondary transport block is scheduled to the UE from a different sector on the same frequency as the primary transport block. The UE also needs to have receive diversity (type 3i) to suppress interference from the other cell as it will receive data on the same frequecny from multiple serving cells.figure 1 llustrates the high-level concept of SFDC-HSPA. F1 F2 DC-HSPA Primary SFDC-HSPA Secondary Figure 1: DC-HSPA and SFDC-HSPA In the case where the two sectors involved in Multipoint HSPA transmission belong to the same NodeB (Intra-NodeB mode), as illustrated in Figure 2, there is only one transmission queue maintained at the NodeB and the RNC. The queue management and RLC layer operation is essentially the same as for DC-HSPA. RNC RLC F1 RLC (a) Intra-NodeB Multipoint HSDPA Figure 2: Intra-NodeB and Inter-NodeB Multipoint HSPA 2/2011 page 2

5 In the case where the two sectors belong to different NodeBs (Inter-NodeB mode), as illustrated in Figure 2, there is a separate transmission queue at each NodeB. RLC layer enhancements are needed at the RNC along with enhanced flow control on the Iub interface between RNC and NodeB in order to support Multipoint HSPA operation across NodeBs. These enhancements are discussed in more detail in Section 4. In both modes, combined feedback information (CQI and HARQ-ACK/ NAK) needs to be sent on the uplink for both data streams received from the serving cells. On the uplink, the UE sends CQIs seen on all sectors using the legacy channel structure, with timing aligned to the primary serving cell. Dual Frequency Multipoint HSPA. When two carriers are available in the network, there is an additional degree of freedom in the frequency domain. Dual Frequency Dual Cell HSPA (DFDC-HSPA) allows exploiting both frequency and spatial domains by scheduling two independent transport blocks to the UE from two different sectors on two different frequency carriers. For a DC-HSPA capable UE, this is equivalent to having independent serving cells on the two frequency carriers. In Figure 3, UE1 is in DC-HSPA mode, whereas UE2 is in DFDC-HSPA mode. F1 F1 UE1 F2 Sector 1 UE2 F2 Sector 2 Figure 3: DFDC-HSPA Dual Frequency Four-Cell HSPA (DF4C-HSPA) can be seen as a natural extension of DFDC-HSPA, suitable for networks with UEs having four receiver chains. DF4C- HSPA allows use of the four receiver chains by scheduling four independent transport blocks to the UE from two different sectors on two different frequency carriers. DF4C-HSPA is illustrated in Figure 4. Sector 1 UE2 Sector 2 Figure 4: DF4C-HSPA Like SFDC-HSPA; DFDC-HSPA and DF4C-HSPA can also be intra-nodeb or inter- NodeB, resulting in an impact on transmission queue management, Iub flow control and the RLC layer. 2/2011 page 3

6 Sector Switching In Sector Switching mode data is transmitted to the UE on only one of the serving cells, either the primary or the secondary. The choice of cell is made every TTI based on CQI value and sector loading conditions. Sector Switching can also operate in both modes Inter-NodeB and Intra-NodeB. In case of Inter-NodeB mode, tight synchronization between the serving NodeBs will also need to be maintained. 3. Advantages Cell Edge Performance Improvement Network operators have identified user experience at the cell edge as an issue to be addressed. With the advancement of heterogeneous networks, in which multiple topological network layers co-exist, cell edge areas and regions with overlapping coverage from multiple cells will increase in total area and number. This magnifies the above problem and at the same time provides opportunity for improving their performance using Multipoint transmission. Multipoint HSPA specifically addresses the performance of users at the cell edge by allowing them to receive data from two different cells in parallel. This concept is similar to downlink soft / softer handover (SHO) in Release 99, the difference being that in Multipoint HSPA, two independent data streams are received from the two cells, whereas in SHO, both cells transmit identical information. Figure 5 Multipoint HSPA Data rate improvement Figure 5 shows the average relative burst rate improvement obtained by the UE using SFDC-HSPA. One UE was randomly placed in the cell during each simulation run lasting 110 seconds and the results were averaged over multiple runs.the table in Figure 6 shows the simulation setup. 2/2011 page 4

7 The expected average improvement for low geometry (i.e., cell edge) users is between 30% and 50% when inter-nodeb mode is enabled and reaches 15% when only intra-nodeb mode is enabled. The greater benefits of the inter-nodeb mode stem from the significantly larger number of users in the soft handover region as compared to the ones in the softer handover region. It can be inferred that inter- NodeB is essential in order to exploit the full potential of Multipoint HSPA technology. Network Topology Inter Site Distance Receiver type Traffic model Power allocation Antenna pattern HS-DPCCH Decoding Channel Model 3GPP 57 cell wrap around 1 km LMMSE, Rx Diversity 3GPP bursty source with 1 Mb bursts Exponential burst inter-arrival times with 5 seconds mean 30% total overhead power (incl. 10% of C-PICH) 2D with 70deg 3dB beam width soft combining for intra-nb aggregation independently at both Node-Bs for inter-nb aggregation PA3 Figure 6: Simulation Setup Load balancing across sectors and frequency carriers In typical deployments, there is a non-uniform load distribution in the network, which often results in neighboring cells having significant load disparities. Multipoint HSPA allows UEs located within the softer / soft handover region between two Multipoint HSDPA unevenly loaded sectors to take advantage of the available resources in the lightly loaded cell and translate them into an aggregate throughput gain. Figure 7: CDF of the Burst Rate within a cell in an Unevenly Loaded Network Figure 7 shows the impact of Multipoint HSPA on the cumulative distribution of the user burst rate within a cell. The cell in question is one of the three center cells in the 57-cell network cluster used in the simulation setup (Figure 6). The three center cells were loaded with 24 active users with a corresponding TTI utilization in the cell of 2/2011 page 5

8 Change in Slot Utilization HSPA+ Advanced Smart Networks: Multipoint Transmission around 80%, while the adjacent cells have eight users and consequently have significant spare capacity. The graph, infigure 7, illustrates that the majority of users in the loaded cell benefit from traffic offload created by Multipoint HSPA. Moreover, the users that benefit most are those with lower data rates, most likely to be cell edge users. For example, less than 5% of users experienced burst rates of less than 2 Mbps with Multipoint HSPA compared to 20% of users without Multipoint HSPA. 25% 20% 15% 10% 5% 0% -5% -10% Offloading to Adjacent Sector Center Sector Adjacent Sector Users in center sector Figure 8: Load Balancing Impact of Multipoint HSPA Figure 8 shows the change in TTI utilization in the center cells and neighboring cells. For the case when the number of UEs in the core cells is 24, corresponding to 80% TTI utilization, more than 10% of TTI utilization in the neighboring cells is due to offloaded traffic from the center cells, which demonstrates the load balancing benefits of Multipoint HSPA. In addition to unevenly loaded neighboring cells on the same frequency carrier, different frequency carriers can also be unevenly loaded. Dual Frequency Multipoint HSPA provides load balancing benefits between frequency carriers similar to those described above for the single frequency case. Leveraging RRU and distributed NodeB technology Remote Radio Units (RRU) technology has increased in popularity recently as it allows operators to deploy multi-sector sites with improved link budget along with greater flexibility for designing cell sites with complex coverage objectives. RRU technology is complimentary to Multipoint HSPA because it allows for sites with more sectors and larger softer handover regions, which is suitable for Multipoint HSPA operation in intra-nodeb mode. 2/2011 page 6

9 4. System Design Considerations For simplicity, this section addresses design considerations for SFDC-HSPA. The concepts are extendable to Dual Frequency Multipoint HSPA. Intra-NodeB Mode For intra-nodeb mode of Multipoint HSPA, all downlink transmissions are controlled by one NodeB. Therefore, there needs to be only one transmission queue in the NodeB as well as only one MAC-ehs entity in NodeB and UE, respectively. The HARQ entities are per data stream with the same timing between the streams. A new requirement on the NodeB is the capability of scheduling data to the UE across two cells. On the UE side, there is a counterpart requirement of handling HSPA transmission from two cells simultaneously. As discussed previously, advanced UEs with multiple receive chains compatible with 3GPP Release 8 or later may already have multiple receive chains. A new requirement on the RNC is the capability of turning on / off Multipoint HSPA operation and handling the UE mobility via RRC signaling. In summary, intra-nodeb mode of Multipoint HSPA requires relatively minor enhancements in the network and the UE. The UE also needs to have a type 3i capable receiver. Inter-NodeB Mode In inter-nodeb mode of Multipoint HSPA, some additional issues need to be addressed. First, there is a separate transmission queue and a MAC-ehs entity in each NodeB, as illustrated in Figure 9. Figure 9: Design considerations for inter-nodeb Multipoint HSPA This requires the RNC to split the downlink data stream into two separate MAC-ehs flows. This is called Multi-link RLC (ML-RLC). Since the packets sent by the RNC across separate flows may experience different amounts of delays, there is a possibility that the UE receives RLC packets out of sequence and therefore sends RLC acknowledgements out of sequence as well. 2/2011 page 7

10 Enhanced RLC implementation in the RNC is needed in order to differentiate missing acknowledgments caused by genuine packet loss from those that are due to out-ofsequence RLC packet delivery; this is referred to as Skew management. Second, enhanced flow control over the Iub interface between NodeB and RNC along with enhanced NodeB buffer management may need to be implemented in order to minimize the differential packet delay between the flows and optimize TTI utilization at each NodeB without causing NodeB under-runs. Enhanced flow control should be based on the current transmission queue status, achievable data rate for HDSPA transmission (e.g., based on UE CQI reports) and cell load conditions. These enhancements are limited to upper layer software upgrades in the NodeB and RNC. On the UE side, in addition to the requirement to receive HSPA transmissions from two cells sectors simultaneously, inter-nodeb mode creates the potential for an additional requirement on the processing time for HARQ acknowledgments. This is because downlink transmissions are not synchronized between different NodeBs and uplink HARQ acknowledgments are paired at the UE with their transmission timing aligned to the primary serving cell. This in turn reduces the processing time available for generation of uplink HARQ ACK / NACK to one of the two NodeBs. Currently allowed processing time is 7.5 ms. It could be reduced to 6 ms for the average case and 4.5 ms in the worst case scenario. An option to the above technique would be for the NodeB to absorb this potential 3 ms reduction in HARQ ACK / NAK processing time. This could potentially reduce the amount of time the NodeB has to determine if it needs to send new data or retransmit old data from the current 4.5ms to a worst case of 1.5ms. Alternatively, this burden can be shared by both, the UE and the NodeB. In all cases the timing alignment information will need to be communicated by the RNC to the UE and the NodeBs involved. 5. Conclusion Multipoint HSPA improves the performance of cell edge users and helps balance the load disparity across neighboring cells. It leverages advanced receiver technology already available in mobile devices compatible with Release 8 and beyond to achieve this. The system impact of Multipoint HSPA on the network side is primarily limited to software upgrades affecting the upper layers (RLC and RRC). 2/2011 page 8