3GPP TSG RAN WG1 NR Ad Hoc Meeting
`Qingdao, China, 27 - 30 June 2017
`
`Agenda Item: 5.1.7
`Source:
`Huawei, HiSilicon
`Title:
`Overview of wider bandwidth operations
`Document for: Discussion and decision
`
`R1-1709972
`
` 1
`
`•
`
` Introduction
`In RAN1#89 meeting, the following agreements on bandwidth part and co-existence of different UE types
`in the wideband carrier are achieved [1].
`Agreement:
`• Confirm the WA of RAN1#88bis.
`•
`Each bandwidth part is associated with a specific numerology (sub-carrier spacing, CP type)
` FFS: slot duration indication if RAN1 decides to not to downselect between 7 symbol and 14
`symbols for NR slot duration
`• UE expects at least one DL bandwidth part and one UL bandwidth part being active among the set of
`configured bandwidth parts for a given time instant.
` A UE is only assumed to receive/transmit within active DL/UL bandwidth part(s) using the
`associated numerology
` At least PDSCH and/or PDCCH for DL and PUCCH and/or PUSCH for UL
` FFS: down selection of combinations
` FFS if multiple bandwidth parts with same or different numerologies can be active for a UE
`simultaneously
`
`It does not imply that it is required for UE to support different numerologies at the same
`instance.
` FFS: TB to bandwidth part mapping
`The active DL/UL bandwidth part is not assumed to span a frequency range larger than the DL/UL
`bandwidth capability of the UE in a component carrier.
`•
`Specify necessary mechanism to enable UE RF retuning for bandwidth part switching
`Agreement:
`•
`Support single and multiple SS block transmissions in wideband CC in the frequency domain
`– For non CA UE with a smaller BW capability and potentially for CA UE, the
`measurement gap for RRM measurement and potentially other purposes (e.g., path loss
`measurement for UL power control) using SS block is supported (if it is agreed that there
`is no SS block in the active BW part(s))
`– UE can be informed of the presence/parameters of the SS block(s) and parameters
`necessary for RRM measurement
`• FFS: via either RMSI, other system information, or RRC signaling
`– FFS: number of SS blocks in wideband
`– FFS: number of SS blocks for RRM measurement
`– FFS: Details of measurement configuration
`Agreement:
`•
`Same PRB grid structure for a given numerology is assumed for narrow band UEs, CA UEs and
`wideband UEs within a wideband NR carrier
`• FFS: PRB indexing
`In this contribution, we discuss the remaining issues on wideband operation, including configuration of
`bandwidth part, scheduling schemes of active bandwidth parts, mechanisms of bandwidth adaptation,
`initial access for wideband carrier and co-existence of wideband UEs and CA UEs. More details could be
`found in our companion papers [2][3][4][5].
`
` Ex. 1012
`APPLE INC. / Page 1 of 9
`
`

`

`2 Configuration of bandwidth part
`2.1 Types of BWP
`It was agreed in RAN1#89 that single and multiple SS block transmissions are supported in wideband CC.
`For some SS blocks, there may be no corresponding RMSI. Therefore, three types of BWPs could be
`considered:
`o BWP with SS block and corresponding RMSI, e.g. BWP1 in Figure 1.
`o BWP with SS block and no corresponding RMSI, e.g. BWP3 in Figure 1.
`o BWP without SS block and corresponding RMSI, e.g. BWP2 in Figure 1.
`
`BWP1
`
`SS block
`
`RMSI
`
`BWP2
`
`BWP3
`
`SS block
`
`
`Figure 1 Three types of bandwidth part
`2.2 The number of configured BWPs
`For the number of configured BWPs, at least two and three BWPs for one UE should be supported. The
`scenarios are:
`o As shown in Figure 2(a), the UE receives both DL control information and DL data within the 1st
`bandwidth part during slot n while the UE switches to the 2nd bandwidth part during slot n+1 for
`DL control information and DL data.
`o As shown in Figure 2(b), the UE receives DL control information within the 1st bandwidth part and
`is scheduled to the 2nd bandwidth part for DL data reception during slot n.
`o As shown in Figure 2(c), , the UE receives DL control information within the 1st bandwidth part in
`both slots n and n+k and is scheduled to the 2nd and the 3rd bandwidth parts respectively during
`slot n and slot n+k for DL data reception.
`
`2nd BP
`
`3rd BP
`
`…
`
`1st BP
`
`2nd BP
`
`1st BP
`
`DL Ctrl
`
`DL data
`
`2nd BP
`
`1st BP
`
`Slot n
`
`Slot n+1
`
`Slot n
`
`Slot n
`
`Slot n+k
`
`(a)
`
`(b)
`
`
`Figure 2 Potential use cases for bandwidth adaptation among multiple bandwidth parts
`2.3 The number of active BWPs
`Multiple active BWPs with same or different numerologies for one UE at a given time instant may be
`needed in NR. The scenarios are:
`o Multiple active BWPs with same numerology
`
`(c)
`
` Ex. 1012
`APPLE INC. / Page 2 of 9
`
`

`

`o In the scenario that a LTE SCell is activated/deactivated dynamically within a NR carrier, as
`shown in Figure 3(a), non-contiguous BWPs with the same numerology should be activated
`simultaneously to efficiently use NR resources.
`o For the two BWPs around the SS block in the frequency domain, as illustrated in Figure 3(b),
`if the numerology of the two BWPs is the same, and is different with that of SS block, the
`two BWPs should be aggregated.
`
`f
`
`f
`
`BWP1
`
`NR
`
`BWP1
`
`30kHz
`
`LTE SCell on
`
`SS block/ RMSI (60kHz)
`
`BWP2
`
`NR
`
`BWP2
`
`30kHz
`
`(a)
`
`(b)
`
`
`Figure 3 Scenarios of multiple active BWPs with same numerology
`o Multiple active BWPs with different numerologies
`o For simultaneously transmissions of eMBB and URLLC services for one UE, multiple
`BWPs with different numerologies should be activated. In addition, to schedule different
`traffic types dynamically for one UE, multiple active BWPs with different numerologies
`could be considered to achieve fast numerology adaptation.
`Proposal 1: Multiple active bandwidth parts with same or different numerologies for one UE at a given
`time instant should be supported.
`3 Scheduling schemes BWPs
`3.1 TB mapping
`Different TB mapping schemes could be considered for multiple active BWPs with same and different
`numerologies, details could be found in our companion paper [2]:
`In case of multiple active BWPs with same numerology, if the multiple active BWPs are associated
`o
`to the same CORESET, it is beneficial for reducing of signalling overhead that only one TB is
`scheduled and mapped to the multiple BWPs. If multiple active BWPs are associated to different
`CORESETs, one or multiple TBs can be transmitted over the multiple BWPs.
`In case of multiple active BWPs with different numerologies, to facilitate product implementation,
`per-BWP TB mapping should be supported.
`Proposal 2: Per-bandwidth part and cross-bandwidth part TB mapping should be supported for multiple
`active bandwidth parts with same numerology; Per-bandwidth part TB mapping should be supported for
`multiple active bandwidth parts with different numerologies.
`3.2 Dynamic resource allocation of different numerologies
`In RAN1#87, it was agreed that NR strives for efficient support of dynamic resource allocation of different
`numerologies in FDM/TDM fashion. One feasible approach is illustrated in Figure 4, only one CSI-RS
`numerology is supported in a carrier in one frequency band, data channel and control channel could have
`multiple numerologies and different numerologies are FDMed. Since the numerology of CSI-RS and data
`channel is independently configured, the channel state information is gotten irrespectively of the
`numerology of the BWP.
`
`o
`
` Ex. 1012
`APPLE INC. / Page 3 of 9
`
`

`

`For UEs with different numerology requirement, e.g. UE1 and UE2 in Figure 4, the configured BWPs for
`the two UEs could be overlapped in the frequency domain. Since the CSI information over the BWPs is
`known at gNB side by the reporting of the two UEs, dynamic resource allocation with different
`numerologies in a FDM manner could be achieved by gNB scheduling. Furthermore, same approach could
`be used for SRS numerology configuration for dynamic UL resource allocation of different numerologies.
`
`CSI-RS, 60 kHz
`
`f
`
`UE1 Data
`
`SCS=15 kHz
`
`UE2 Data
`
`SCS=60 kHz
`
`UE2 BWP
`
`UE1 BWP
`
`t
`
`
`
`Figure 4 Dynamic resource allocation of different numerologies
`Proposal 3: CSI-RS/SRS numerology should be independently configured for the bandwidth part, which
`may be different with the associated numerology for this bandwidth part for data and control.
`4 Bandwidth adaptation
`In our companion paper [3], mechanisms of bandwidth adaptation, e.g. indication of BWP activation, RF
`retuning and fallback mechanism, are discussed. Compared with MAC-CE, DCI is preferred to indicate the
`activation of BWPs due to its low latency. In addition, the time pattern option is necessary for both
`receiving the periodic common information and supporting the fall back mechanism.
`Proposal 4: Support both DCI and time pattern to indicate the activation of bandwidth part, e.g., explicit
`UE-specific or group common DCI.
`If DCI is adopted to indicate the BWP activation/deactivation, there is a possibility that the gNB and UE
`may not be aligned in the current active BWPs. To deal with this problem, a fall back mechanism can be
`considered. One example is illustrated in Figure 5.
`
`
`BWP2
`
`gNB
`
`BWP2
`
`gNB
`
`UE
`
`lost
`
`DCI
`
`BWP1
`
`BWP3
`
`UE
`
`lost
`
`DCI
`
`BWP1
`
`(a)
`
`
`
`(b)
`Figure 5 Example of BWP activation fall back
`Proposal 5: Support fall back mechanism for bandwidth part activation/de-activation.
`The impact of RF retuning should be considered in bandwidth adaptation. If the time for retuning is not
`consistent for gNB and UE, the gNB may schedule the UE when the UE is retuning. Consequently, it is
`necessary for the gNB and the UE to get aligned on the location of the retuning time. Both predefined and
`explicit indication of the guard period for RF retuning could be considered.
`Proposal 6: The starting/ending position of the guard period for UE RF retuning should be predefined
`and/or signalling in DCI.
`5 Initial access for wideband carrier
`For multiple RMSIs in a wideband carrier, the PRACH configuration in different RMSIs should be
`excluded, if a PRACH configuration is configured by two RMSI in a wideband carrier, the gNB will be
`confused about the frequency location of the control for RAR as illustrated in the Figure 6, and the
`
` Ex. 1012
`APPLE INC. / Page 4 of 9
`
`

`

`frequency location of control for RAR may be same as the control for RMSI which is indicated in the
`PBCH.
`
`SS block
`
`SS block
`
`SS block
`
`PRACH
`
`PRACH
`
`PRACH
`
`PRACH
`
`PRACH
`
`RMSI
`
`RAR
`
`？
`
`RMSI
`
`RAR
`
`？
`
`
`Figure 6 PRACH configuration in wideband CC
`Proposal 7: When multiple RMSIs are present in one wideband CC, each RMSI shall contain its unique
`PRACH configuration.
`It has been agreed in previous meeting that the maximum bandwidth for CORESET for RMSI scheduling
`and NR-PDSCH carrying RMSI should be equal to or smaller than a certain DL bandwidth. To avoid
`frequency retuning during initial access, SS block, CORESET for RMSI and NR-PDSCH carrying RMSI
`should be confined within a certain DL bandwidth, which could be regarded as common bandwidth part,
`and the bandwidth of the common bandwidth part should be supported by all UEs [4].
`Proposal 8: SS block, CORESET for RMSI scheduling and NR-PDSCH carrying RMSI should be confined
`within a certain DL bandwidth that is equal to or smaller than minimum UE receive bandwidth.
`6 Co-existence for different UE types for wideband carrier
`Let us define Type A UEs as those operating in intra-band contiguous CA mode and Type B UEs as those
`operating in single wideband mode. To improve the resource efficiency, flexible MU-MIMO multiplexing
`of Type A and Type B UEs should be supported. In LTE, the reference signal sequence is independently
`generated and mapped per CC. In a given time-frequency resource, if the sequence is not the same for Type
`A and Type B UEs, the orthogonality may be degraded and RS with OCC could not be utilized. To tackle
`this issue, as illustrated in Figure 7, in each CC within the wideband CC, Type A and Type B UEs should
`use the same sequence, so the orthogonality of whole RS within the wideband CC can be obtained.
`f
`
`Wideband CC
`
`CC1
`
`CC2
`
`CC3
`
`CC4
`
`Sequence 1
`
`Sequence 1
`
`Sequence 2
`
`Sequence 2
`
`Sequence 3
`
`Sequence 3
`
`Sequence 4
`
`Sequence 4
`
`Wideband
`
`4-CC CA
`
`
`
`Figure 7. Reference signal generation and mapping
`Proposal 9: MU-MIMO multiplexing of Type A and Type B UEs should be supported for both DL and UL.
`The same reference signal sequence should be mapped in the same time-frequency resource for Type A and
`Type B UEs.
`
` Ex. 1012
`APPLE INC. / Page 5 of 9
`
`

`

`6.1 Frequency location indexing for reference signal generation
`To ensure all types of UE use the same reference signal sequence over the same time-frequency resource,
`the frequency location indexing of PRBs for RS sequence generation should be the same for all UE types
`over the wideband CC. One possible approach is to define a single reference point in the entire wideband
`CC for frequency location indexing for RS sequence generation. Once the reference point is obtained, the
`frequency location index of a PRB is determined by its distance, in number of PRBs, to the reference point.
`Based on the location of PRB, its index could be positive or negative.
`Depending on whether the same type or different types of sequence generation is used for remaining
`minimum system information (RMSI) and unicast data, two categories of alternatives can be considered for
`the frequency reference point (of default numerology):
`Alt 1. Same type of sequence generation for both RMSI and unicast data. In this alternative, the reference
`point is obtained by UE before decoding the RMSI. The reference point is then used to generate the RS
`sequences for both RMSI and all unicast data such as NR-PDSCH.
`Alt 1-1. The downlink DC subcarrier of wideband CC is taken as the reference point. In this case, the
`location of DC is fully signaled in PBCH. The location of DC can be represented as an offset with respect
`to a predefined point in the SS block or a predefined point in the bandwidth part (BWP) of RMSI and SS
`block, called common BWP.
`Alt 1-2. An SS block is taken as the reference SS block with respect to which the reference point is
`obtained. This alternative can be realized by any of the following two approaches:
`• Alt 1-2-1. Only one SS block out of the multiple SS blocks in frequency in wideband CC has a
`corresponding RMSI. This SS block is the reference SS block and a predefined point with respect
`to this SS block is taken as the reference point. All UEs in wideband CC need to detect this SS
`block to be able to obtain the reference point. The indication of this SS block can be done through
`NR-MIB in PBCH by indicating whether or not the SS block has a corresponding RMSI.
`• Alt 1-2-2. Multiple SS blocks in frequency have corresponding RMSIs. For example, each SS
`block in frequency in wideband CC has a corresponding RMSI. In this case, one approach is that
`one bit in NR-MIB in PBCH indicates whether the SS block is considered as the reference SS
`block [5]. If an SS block is indicated as reference SS block, a predefined point with respect to this
`SS block is taken as the reference point. Another approach with higher overhead is that the location
`of the reference point as an offset with respect to each SS block is included in NR-MIB of its
`PBCH. The offset values of different SS blocks are such that they all point to the same reference
`point, as illustrated in Figure 8.
`Alt 2. Two types of sequence generation for RMSI and unicast data. In this alternative, each UE needs to
`first decode RMSI to be able to obtain the reference point. Therefore, a specific RS is used for CORESET
`of RMSI and for RMSI itself in NR-PDSCH. After obtaining the reference point, the UE uses it to generate
`the RS sequences for all unicast data such as NR-PDSCH.
`Alt 2-1. The downlink DC subcarrier of wideband CC is taken as the reference point. In this case, part
`of DC info is signaled in PBCH, e.g. to indicate whether or not DC is present in common BWP. If DC is
`present in the common BWP, then the predefined location of the DC for common BWP is used, otherwise,
`RMSI indicates the location of DC. The location of DC is represented as an offset with respect to a
`predefined point in the SS block or a predefined point in the common BWP.
`Alt 2-2. Location of the reference point is signaled through RMSI or other system information. The
`location of reference point can be represented by a frequency offset, in number of PRBs of default
`numerology, with respect to the frequency location of the corresponding SS block. If there are more than
`one SS blocks in frequency in the wideband CC, the offset values transmitted with different SS blocks are
`such that they all point to the same reference point, as illustrated in Figure 8.
`
` Ex. 1012
`APPLE INC. / Page 6 of 9
`
`

`

`WCC
`
`NCC
`
`Reference point
`
`SS blocks within WCC
`
`frequency
`
`
`
`Figure 8. Location of reference point as an offset with respect to the location of SS block.
`Alt 2-3. Location of the reference point is predefined with respect to the location of a reference SS
`block. If there are multiple SS blocks in WCC, one of them is defined as reference SS block, e.g. SS block
`#0. But UE may not know the location of the reference SS block, since it may have accessed the system
`through a different SS block. To inform the UE of the location of the reference SS block, one possible
`approach is that the relative frequency distances of the SS blocks is signaled through RMSI or other system
`information of each SS block together with the frequency index of the corresponding SS block, as
`illustrated in Figure 9. Each UE after accessing an SS block can obtain the locations and indexes of all SS
`blocks using this information. Then, it can determine the location of the reference point by using the
`predefined offset and the location of the reference SS block.
`
`SS block #0
`
`SS block #1
`
`SS block #2
`
`WCC
`
`2Δf
`1Δf
`Figure 9. Relative frequency distances of SS blocks and the frequency index of each SS block.
`
`frequency
`
`
`
`
`Proposal 10: Single frequency reference point is supported for frequency location indexing for RS
`sequence generation by all UE types in wideband CC.
`6.2 Frequency location indexing for mixed numerologies
`In the case of mixed numerologies, for each numerology, all types of UE should use the same RS sequence
`over the same time-frequency resource. Therefore, the frequency location indexing of PRBs of each
`numerology for RS sequence generation should be the same for all UE types over the wideband CC. In
`particular, since the PRB grid of a numerology only depends on its underlying subcarrier spacing (SCS),
`the frequency location indexing for RS sequence generation is SCS specific. In other words, for each SCS,
`a single reference point can be defined in wideband CC for frequency location indexing for RS sequence
`generation of any numerology with that SCS.
`Proposal 11: Frequency location indexing for RS sequence generation is subcarrier spacing specific.
`The reference point of a specific SCS can be defined based on the reference point of the default SCS by
`using a predefined rule. Once the reference point of an SCS is obtained, the frequency location index of a
`PRB of the SCS is determined by its distance, in number of PRBs of that SCS, to the reference point. Some
`examples of predefined rules for reference points of different SCSs are shown in Figure 10.
`Proposal 12: The reference point for frequency location indexing of each subcarrier spacing is obtained
`from the reference point of the default subcarrier spacing by a predefined rule.
`
` Ex. 1012
`APPLE INC. / Page 7 of 9
`
`

`

`(a) Right-alignment of reference PRBs
`
`Figure 10. Examples of predefined rules for frequency reference points in mixed numerologies
`
`(b) Left-alignment of reference PRBs
`
`
`
`
`
` 7
`
` Conclusion
`Proposal 1: Multiple active bandwidth parts with same or different numerologies for one UE at a given
`time instant should be supported.
`Proposal 2: Per-bandwidth part and cross-bandwidth part TB mapping should be supported for multiple
`active bandwidth parts with same numerology; Per-bandwidth part TB mapping should be supported for
`multiple active bandwidth parts with different numerologies.
`Proposal 3: CSI-RS/SRS numerology should be independently configured for the bandwidth part, which
`may be different with the associated numerology for this bandwidth part for data and control.
`Proposal 4: Support both DCI and time pattern to indicate the activation of bandwidth part, e.g., explicit
`UE-specific or group common DCI.
`Proposal 5: Support fall back mechanism for bandwidth part activation/de-activation.
`Proposal 6: The starting/ending position of the guard period for UE RF retuning should be predefined
`and/or signalling in DCI.
`Proposal 7: When multiple RMSIs are present in one wideband CC, each RMSI shall contain its unique
`PRACH configuration.
`Proposal 8: SS block, CORESET for RMSI scheduling and NR-PDSCH carrying RMSI should be confined
`within a certain DL bandwidth that is equal to or smaller than minimum UE receive bandwidth.
`Proposal 9: MU-MIMO multiplexing of Type A and Type B UEs should be supported for both DL and UL.
`The same reference signal sequence should be mapped in the same time-frequency resource for Type A and
`Type B UEs.
`Proposal 10: Single frequency reference point is supported for frequency location indexing for RS
`sequence generation by all UE types in wideband CC.
`
` Ex. 1012
`APPLE INC. / Page 8 of 9
`
`

`

`Proposal 11: Frequency location indexing for RS sequence generation is subcarrier spacing specific.
`Proposal 12: The reference point for frequency location indexing of each subcarrier spacing is obtained
`from the reference point of the default subcarrier spacing by a predefined rule.
`
`References
`[1] Chairman notes RAN1#89
`[2] R1-1709974, “Scheduling and resource allocation mechanism for active bandwidth parts”, Huawei, HiSilicon,
`Qingdao, China, 27-30 June, 2017.
`[3] R1-1711424, “On bandwidth adaptation”, Huawei, HiSilicon, Qingdao, China, 27-30 June, 2017.
`[4] R1-1709973, “On initial access for wideband carrier”, Huawei, HiSilicon, Qingdao, China, 27-30 June, 2017.
`[5] R1-1709914, “NR-PBCH contents and payload size”, Huawei, HiSilicon, Qingdao, China, 27-30 June, 2017.
`
`
`
`
`
`
`
`
` Ex. 1012
`APPLE INC. / Page 9 of 9
`
`