1
`
`R2-071762
`
`3GPP TSG-RAN WG2 #58
`7-11 May 2007
`Kobe, Japan
`
`Agenda item:
`4.5
`QUALCOMM Europe
`Source:
`Scheduling of D-BCH
`Title:
`Document for: Discussion
`
`Introduction
`1.
`At RAN2#57, it was agreed that the system information is divided into scheduling units, with SU-1 carrying scheduling
`information for the other SUs. This document attempts to take some steps towards understanding how this agreement
`can be implemented in practice.
`
`2. Discussion
`
`2.1. Simple approaches to scheduling
`Once the UE has read the primary (and secondary, if applicable) BCH and camped on a cell, it needs to receive the rest
`of the system information from the D-BCH. It is fairly clear how this happens in general terms: TS 36.300 specifies
`that SU-1 includes “[s]cheduling information of the other Scheduling Units”, so presumably the UE monitors the
`scheduling channel for an indicator of SU-1, then reads further scheduling information from there.
`
`In the simplest interpretation, this text could mean that SU-1 contains explicit pointers to the resource blocks that
`contain the other SUs. In this case, a separate L1/2 control channel for the D-BCH might not actually be needed (with
`the consequence that when any SU changed, all affected UEs would need to read the P-BCH and then SU-1 to find the
`updated version). This situation is shown in Figure 1.
`
`Figure 1: Explicit scheduling in SU-1
`However, it should be remembered that some SUs, such as the container for the positioning information, could be
`delivered over a very long period (e.g. tens of seconds to minutes), and it is not clear that it would be realistic to expect
`the scheduler to know this far in advance exactly how resource blocks would be assigned. Therefore it could be
`impractical to expect SU-1 to carry a complete set of explicit scheduling for all the other SUs.
`
`On the other hand, if such a complete set is not present in SU-1, the question arises of how exactly the UE does receive
`scheduling for such a long-period SU. The simplest approach is to assume that the “scheduling information” in SU-1
`
`1
`
`APPLE 1007
`
`

`

`
`
`just consists of a pointer to a control channel for the D-BCH, and other SUs are scheduled independently using that
`control channel, as shown in Figure 2. (A “hybrid” approach could also be used, where some SUs were indicated
`explicitly in SU-1, but longer-period ones whose scheduling could not be known in advance would be scheduled
`independently using the control channel.)
`
`2
`
`
`
`Figure 2: Waiting for long-period system information
`Obviously, this mode of scheduling is not optimal; as the figure indicates, it forces the UE to “busy-wait” on the
`scheduling channel for a long period for the scheduling of SU-2. It also raises the question of how the UE distinguishes
`the scheduling for different SUs; does each SU have its own BCCH-RNTI?
`
`Neither of the straightforward approaches considered here is really satisfactory; either the scheduler must have the
`ability to reserve resource blocks outlandishly far in advance, or the UE has an unreasonable duty cycle while it waits
`for all SUs to arrive. We attempt to address these deficiencies in the next section.
`
`2.2. Possible solutions
`In order to take advantage of the flexibility of the DL-SCH structure, and to avoid the need for rigid advance
`scheduling, we assume that the D-BCH has an associated control channel, which is used to schedule the SUs. (SU-1
`could be an exception, scheduled only on the P-BCH; the figures below make this assumption for simplicity.) The
`“scheduling information” referred to in TS 36.300 then indicates not the exact location of the SUs themselves, but the
`point where they are scheduled (Figure 3).
`
`P-BCH
`
`SDCCH
`
`SU-1
`
`SU-2
`
`Figure 3: SU-1 indicates scheduling of SU-2
`This approach may be acceptable. However, in the case of a very-long-period SU, it still requires advance information
`about the scheduling. If the SUs can be scheduled according to rigid repetition cycles, this information is easy for the
`
`
`
`2
`
`

`

`
`
`3
`
`network to determine, but then the SUs must be scheduled compatibly to prevent collisions in the SDCCH slots, and if
`the number of SUs is large this could present a combinatorial problem. Therefore it might be desirable to allow a more
`flexible transmission schedule, e.g., by allowing SU-1 to indicate that a later instance of SU-1 will give the scheduling
`for SU-2 (Figure 4).
`
`P-BCH
`
`SDCCH
`
`SU-1
`
`SU-2
`
`Rx activity
`
`SU-1 received
`
`SU-2 received
`
`Scheduling received
`for SU-2
`
`
`
`Figure 4: SU-1 indicates when scheduling of SU-2 will be available
`Here the scheduling of SU-2 is slightly irregular as compared to the SU-1 cycle; the second instance of SU-2 is
`scheduled one slot “late”, and the location of that scheduling information is included in the fourth instance (only) of
`SU-1. The second and third instances of SU-1 just contain pointers into the future, to the fourth instance where the
`scheduling information will be available.
`
`The first alternative is conceptually straightforward, but raises some practical complexities. Although the constraint
`problem is not as bad as with explicit scheduling of SU-2 within SU-1, the scheduler still needs to be able to guarantee
`in advance that the timing of the scheduling for different SUs does not collide. To some extent this problem could be
`addressed by having the timing sent as a lower bound (“SU-2 will be scheduled 23873 slots from now, or shortly
`thereafter”) rather than as an exact offset (“SU-2 will be scheduled exactly 23873 slots from now”) or a fixed cycle
`(“SU-2 is scheduled every 24000 slots, starting 23873 slots from now”).
`
`The second approach does not have this minor complexity, but it is a rather roundabout approach to the ultimate goal of
`reading SU-2, and it may be preferable to use a logically simpler solution and find a way to resolve any issues that arise.
`
`In both cases, the control channel needs to have the ability to indicate which SU is being scheduled. This could be
`achieved either with extra information in the scheduling channel itself (space permitting), or by using multiple BCCH-
`RNTIs.
`
`In practice, the differences between these two approaches to scheduling do not appear to be very great; none of the
`issues noted are showstoppers, and either approach will work. Indeed, if the long-term constraints on the scheduler
`discussed in Section 2.1 are acceptable, the problem does not need to be solved, since SU-1 can simply carry an
`extremely foresighted schedule for SU-2 itself. It would be premature to take a decision at this level of detail, but it
`may be a good time to consider whether there is a group preference.
`
`3. Conclusion
`This document has identified some potential issues with the scheduling of the D-BCH and possible solutions. While a
`final decision is not needed at this stage, we suggest that RAN2 should begin discussion of these issues.
`
`
`
`3
`
`