`Cannes, France
`June 21-24, 2004
`
`Agenda item:
`
`6.3
`
`Source:
`
`Lucent
`
`Title:
`
`SHO Support for E-DCH
`
`Document for:
`
` Decision
`
`
`
`1
`
`Introduction
`
`
`
`R1-040706
`
`
`
`One of the objectives of allowing the NodeB control of the TFCS subset is to dynamically manage the UL
`radio resource usage and transmissions to optimize the overall system throughput. Having the NodeB in
`control of the subset reduces the latency that is experienced in the current R99/4/5 uplink RNC TFCS
`control. As in HSDPA, with NodeB control scheduling and control of the TFCS subset, better adaptation to
`the channel variation is possible. However, one disadvantage is that since the resource management entity for
`E-DCH is located at the Node B, it is difficult to coordinate the control between the different NodeBs during
`soft handover operation.
`
`SHO support for EDCH was agreed in WG1#37 in Montreal. In this document, the support of SHO for
`EDCH is evaluated from the point of view of different Scheduling and HARQ schemes.
`
`Handover Scenarios
`
` 2
`
`
`
`As the radio resource management and dynamic control of the UL transmission residing at the Node B, the
`behaviors of the supported enhancements for the EDCH feature, such as scheduling and HARQ, needs to be
`specified clearly during the handover. One NodeB control function in the R99/4/5 DCH is the inner loop
`power control. The R99/4/5 UL DCH supports soft handover with DL TPC combining strategy for the UL
`inner loop power control. The R5 HSDPA feature is designed to schedule user in time and code domain
`based on the DL radio channel condition feedback from the UE. In HSDPA, the support of soft handover is
`not provided mainly due to the difficulty of the scheduling control function coordination between Node B in
`short interval and the delay tolerance of the best effort service class defined for the HSDPA. Thus, for EDCH
`to support SHO, similar difficulties need to be overcome.
`
`
`
`2.1
`
`Different Scheduling Modes
`
`Many issues related to scheduler operation in SHO exist e.g. the selection of the scheduler entity in SHO.
`However, they will not be discussed here. Instead, in this section, the support of SHO by the type of
`scheduler: Rate and Time or Rate Scheduling is discussed.
`
`With Rate and Time scheduling, the Node B controls the UE transmission data rate and its transmission
`duration to optimize the system capacity. Since the UL system capacity is interference limited, the control of
`an UE transmission data rate is also relative to the interference it creates to other UEs. . The rate scheduling
`
`NAC1011
`Page 1
`
`
`
`technique manipulates individual UE’s UL data rate based on its radio channel condition and the system
`load. However, regardless of the scheduling modes being implemented, the common feature presents during
`handover period is that all Node Bs in the active set will receive the UL transmission from the UE. Hence:
`
`-
`
`-
`
`In order that the RoT for a specific Cell not to be adversely affected due to transmission from a UE
`in SHO, the NodeBs in the Active Set controlling the specific Cell needs to support scheduling
`operations for UE in SHO;
`
`If the NodeB supports SHO scheduling, it would also attempt to receive and decode the uplink
`transmission from the SHO UE. Hence, the gain from selection combining at the RNC is possible.
`However, if the NodeB support of SHO scheduling is removed, the NodeB may decide not to decode
`the uplink transmission, resulting in loss of combining gain;
`
`Rate Scheduling: With rate scheduling, the Node B dynamically sets the TFCS subset at each scheduling
`period time interval. All Node Bs in the active set during the SHO would operates the handover
`independently and would send independent comments to control the transmission rate of the UE in SHO. It
`would also evaluate its own system load. This behavior is similar to that of UL power control during the
`soft handover. Thus, rate scheduling technique should support soft handover with the addition of the
`combining strategy of the received rate control comments from each Node B.
`
`Rate and Time Scheduling: With rate and time scheduling, the Node B selects a sub-group of UEs to transmit
`at their given data rate at each time interval. In a given radio resource, the rate and time scheduling tends to
`distribute the UL data transmission in time domain with higher data rate. The higher data rate transmission
`potentially could create larger interference to other user, compared to Rate scheduling due to the un-
`synchronized NodeBs in SHO. Thus, the performance gains of rate and time scheduling in SHO should be
`investigated further and should be kept as the working assumption now.
`
`
`
`2.2
`
`HARQ Modes
`
`In the previous meeting, both Incremental redundancy and Chase combining have been agreed. Within these
`two modes, the options of operating with synchronous or asynchronous modes exist [1]. Between these two
`modes, the effects of link imbalance are different. To support SHO for Edch implies that multiple HARQ
`entities are needed for the UE in SHO. Due to link imbalance, the state machines across the different HARQ
`entities across the active set could be unsynchronized e.g. different ACK/NACK error transition probabilities
`on the different SHO links.
`
`Because of the different link performances, multiple error scenarios could potentially exist among the
`different HARQ entities. A key task is then to optimize the ‘synchronization’ of the different HARQ entities
`so that the performance degradation is minimized. Between Chase and IR, the latter would require more
`uplink signaling bits and between synchronous and asynchronous, the latter would require larger signaling
`bits. While the benefits of asynchronous HARQ are discussed in [1], the gains of Chase over IR for SHO are
`less obvious.
`
` Summary
`
` 3
`
`
`
`It is proposed that the following working way forward be considered:
`
`NAC1011
`Page 2
`
`
`
`1. Soft handover should be supported for Rate scheduling to enable the soft handover gain,
`
`2. The support of SHO for Rate and Time scheduling be kept as an option until its performance gains or
`degradation is shown through simulation studeis,
`
`3. Chase combining be supported for SHO and IR be kept as an option until its performance gains in
`SHO is verified.
`
`References
`
` 4
`
`
`
`[1] R1-Async, “Asynchronous/Synchronous HARQ” Lucent.
`
`NAC1011
`Page 3
`
`