`
`ZTE Corporation and ZTE (USA) Inc.
`
`
`
`Inv. No. 337-TA-868
`
`RX-3305
`
`ETSI SMG2 UMTS L1 Expert Group
`
`Tdoc SMG2 UMTS—Ll 683/98
`
`Espoo, Finland
`14-18 December 1998
`
`Source: Motorola
`
`Mechanisms for managing uplink
`interference and bandwidth.
`
`Abstract
`
`This paper introduces the concept ofdynamic persistence as a backoff algorithm for the RACH reducing the risk
`of outages and improving RACH performance. Also, an uplink shared channel
`is proposed for providing
`scheduling flexibility to better support mixed traffic types.
`Finally, the size and structure of a composite
`downlink common control channel is discussed supporting feedback and control for both uplink and downlink
`packet data transfers.
`
`1.0 Introduction
`
`Recent papers regarding the RACH have concentrated on remedies for open-loop power control suggesting both
`faster feedback mechanisms and power ramping as solutions [l][2][3]. Power control is an important factor that
`should not be neglected, however, the impact of congestion and the interaction of mixed traffic types has not
`received adequate attention. This Tdoc proposes two strategies to aid interference and bandwidth management
`on both the RACH and during b11lk transfers. The congestion and mixed traffic may impact the RACH chamiel
`in the following ways.
`
`0
`
`0
`
`Periods of high load will cause congestion on the RACH. Although CDMA greatly reduces the likelihood
`of an irresolvable collision,
`the aggregate interference power produced by a score of subscribers can
`potentially exceed the allowable noise rise at the Node B. As a result, all users in the cell would experience
`an outage that the Node B cannot mitigate.
`
`The interaction of mixed traffic may reduce the average quality of service by favoring users making bulk
`transfers on the uplink while relegating the shorter interactive messages to a congested RACH. With the
`currently defined mechanisms, file transfers or large emails would be assigned DCHs and could effectively
`consume the majority of the uplink packet data capacity during periods of high loading. Moderate and small
`transfers would be relegated to a congested RACH and suffer from undue delay.
`
`2.0 Congestion Management of RACH channel
`Local Area Networks [LANs) carrying packet data have traditionally relied on distributed backoff algorithms for
`managing congestion on the system. Likewise, cellular systems employ distributed backoff algorithms for the
`common access channel e.g. GSM’s RACH. For FDMA systems, the impact of congestion is tolerable. A
`congested access channel will
`increase the call set-up time by escalating the backoff delay of backlogged
`subscribers until congestion is eventually resolved (even o11 an ill tuned system).
`In an FDMA system,
`subscribers assigned channels will be not be affected by the congestion because the quality of their circuit is
`protected by the frequency separation.
`In a CDMA system congestion on the RACH will increase the noise rise
`at the Node B and, at times, may produce outages that effect all users in the system, both backlogged users and
`those assigned dedicated channels.
`As a precaution,
`the parameters of the backoff algorithm may be
`conservatively assigned to reduce the likelihood of destructive congestion, however, the cost would be an overall
`reduction in performance. What’s more, the variety of data services and the potential for new yet unknown
`services may make it
`impossible to identify the appropriate parameters for a well-tuned RACH backoff
`algorithm. A more sensible approach would take advantage of the centralized architecture of a cellular system
`
`NK868|TCO17317124
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1006-00001
`
`
`
`and allow the system to dynamically manage the backoff of all subscribers based on estimates of the congestion
`in the cell.
`In essence, a closed loop backoff algorithm could be formed where the system proactively prevents
`destructive interference from occurring during times of high load without compromising the mean performance
`of the system.
`
`A dynamic persistence algorithm is proposed for managing interference and minimizing delay by controlling
`access to the RACH channel of the system. On a common downlink channel,
`the system will publish a
`persistence value, p, the value being dependent on the estimated backlog of users in the system [4]. The
`persistence value would be broadcast at regular intervals e. g. once per frame. A UE in an active data session
`would first read the current persistence value and transmit on the next available RACH slot with a probability, p.
`When the system detects congestion on the RACH, the published persistence value would be reduced decreasing
`the number of users transmitting on the RACH in any given frame. During periods of inactivity the probability
`would be increased in order to improve the overall access delay. The elegance of the approach is that the
`backoff is handled deterministically based on a noise rise at the Node B. Whats more, the system may take
`multiple factors into account when calculating the persistence value such as data traffic patterns, recent history,
`and adjacent cell loading. Furthermore, having the system compute the dynamic persistence rather than the UE,
`provides the system operator greater flexibility in managing the RACH.
`
`The cost of publishing a dynamic persistence value would be small in terms of both complexity and capacity.
`Previously, a common channel for providing fast power control feedback to UEs accessing the RACH has been
`discussed [5]. A persistence value could be conveyed by using a few bits in a common downlink channel.
`Furthermore, the power assigned to this common channel could be conservative. UEs at the fringe of coverage
`pose a smaller threat of causing an outage and therefore the system will tolerate the diminished integrity of the
`persistence information. Although UEs near the center of the cell would pose a greater risk for causing outages,
`the risk would be mitigated by a more reliable downlink common channel carrying persistence information.
`
`Variations on dynamic persistence may provide other benefits such as a solution for managing different QoS
`data streams. UEs with a higher QoS might augment the persistence value they employ by an assigned QoS
`factor and transmit with a higher probability than the bulk-rate traffic.
`
`3.0 Managing mixed traffic types
`An earlier Tdoc [6] presented analysis and simulation that showed that “Internet” users will receive the optimum
`perceived quality of service when packets are scheduled using the broadest pipe possible.
`Several Tdocs
`[7][8][9][l0] presented methods for providing a broad pipe on the downlink using a shared channel. On the
`uplink a similar mechanism is required to allow scheduling of the uplink resource using the broad pipe.
`
`Currently in the specified mechanisms for managing the uplink resource, it is only possible to react to changes in
`user requirements very slowly (especially if time-outs are employed). This results because of the need for
`signaling negotiation between cell-specific RRC in the CRNC(s) and user specific RRC in the SRNC every time
`that the power assigned to a user is to be changed.
`It is important that the percentage time spent releasing and re-
`assigning resource is kept to a miiiiiiiuiii in order for the uplink capacity to be well utilized. Therefore, the above
`observation implies that in order to minimize the percentage time spent in re-assignment of capacity, users
`should hold the channel for a relatively long period of time (in other words transmit at a low rate).
`In addition,
`the current documents state a UE is autonomously allowed to decide at what bit rate it will transmit at (within
`the allowed TFCS). This tends to imply a reliance on a statistical averaging of bandwidth requirements between
`the transmissions of different packet users. Reliance on statistical averaging is only possible when the number
`of users is great. Clearly, the current mechanisms for managing access to the uplink resource implicitly imply
`the use of multiple thin pipes. To summarize, it is clear that there are a number of problems with the current
`proposals for managing access to the uplink resource for packet users:
`
`1. Lack of fast access control (assignment and re-assignment) will result in poor utilization of uplink
`capacity.
`
`2. The currently defined mechanisms are biased toward the use of multiple thin pipes
`
`0 Resulting on average in a longer delay for packet transmissions as perceived by the user (see
`companion paper [6]).
`
`0 Resulting in a loss of flexibility in prioritizing the transmission of packets from different users.
`
`The loss of flexibility in prioritizing the transmission of packets from different users occurs because current
`uplink packet data procedures tend to favor bulk transfers over infrequent acknowledgements or requests. The
`problem is exacerbated when the system is under a condition of high loading. As an illustration consider the
`
`NK868|TCO1731 7125
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1006-00002
`
`
`
`case where one UE, referred to as UEemai1, is transferring large email on a heavily loaded system and where
`several other UEs have moderate-sized infrequent traffic (moderate-size is greater than a few RACH bursts).
`UESW,-1 would typically be assigned a DCH for the transfer.
`In the absence of other traffic, the system should
`assign the broadest pipe possible to expedite the transfer. Learning the interference margin at the base, users
`making infrequent transfers (which may be of higher priority) would be relegated to the lowest data-rate
`available for the duration of Uiemaifs transfer. Of course the system could reserve more capacity for the
`infrequent transfers, but this would be sub-optimum solution since capacity would go unused prolonging
`UESIW-1’s transfer. Overall system performance could be improved by providing a scheduling mechanism that
`allowed uplink capacity to reallocate in an expeditious manner.
`
`4.0 Proposed MAC based scheme for dynamic scheduling between users
`Before discussing the motivation for performing dynamic scheduling between users on the MAC layer it is first
`worth stating that it is assumed that the uplink shared channel (USCH) to be described will be used by packet
`users.
`It is an assumption (as is the case for the DSCH and as was assumed in the evaluation documents) that
`packet users can be operated efficiently in hard hand-off due to the extra delay budget available and the
`consequent capability to use ARQ.
`
`A method for allocating uplink capacity is proposed that mimics the features of the downlink shared channel
`which allow for a broad pipe and nimble scheduling. Unlike the downlink, there is no shortage of orthogonal
`codes to manage, however, there is finite allowable noise rise at the Node B.
`It is proposed that the system
`apportion interference power for uplink packet scheduling in a similar manner to how codes are assigned in the
`downlink. This differs from noise rise published on the BCCH in that the system budgets UE access based on
`the known UE queue depths in a deterministic rather than a probabilistic manor. Ideally, the UE would transmit
`at the maximum data rate that is commensurate with the apportioned noise rise at the Node B. The maximum
`data rate is a function of the apportioned noise rise and the total interference power incident at the Node B.
`From these variables, the system can calculate the target Eb/Nt on the uplink and with the knowledge of link
`performance calculate the maximum sustainable data-rate. The spreading factor could be assigned on a common
`downlink broadcast transport channel called the Access Control Channel (ACCH) which will be described in
`more detail later. Each allocation would be valid for the duration of one frame allowing for the flexible
`reassignment of resources. With this fast assignment mechanism in place, bandwidth could be managed more
`efficiently allowing a scheduler to multiplex a mix of traffic types.
`
`Table 1 examines the overhead associated with conveying the uplink resource assignments. An assignment may
`be made in as few as 12 bits consisting of a Temporary UE ID (TUEID), a Spreading Factor Assignment (SFA),
`and Power Control Position Assignment (PCPA). A 6-bit TUEID would provide resolution equivalent to the
`GPRS slot assignment plus USF. A TUEID would last the duration the UE was in the active state. The NRA
`(Noise Rise Apportionment) value would represent a quantized assignment of the noise rise apportioned to the
`UE for the duration of one frame. Finally, the PCPA assigns a position in a common power control channel for
`the duration that a NRA is valid [1 1].
`Table 1
`
`
`
`PCPA
`
`Total
`
`Bits
`
`6
`
`3
`
`3
`
`1 2
`
`Reference
`
`A 6-bit temporary ID providing equivalent resolution
`as GPRS (Slot+USF <=> 3 + 3) used to identify an
`allocation.
`
`Assigns the spreading factor for the next frame.
`
`Assigns a position in the common power control
`channel for the duration of the transfer.
`
`UEs with data to send would perform the following procedure:
`
`1) At the beginning of a session or packet call the UE would send an initial request on the RACH carrying a
`reservation request containing the queue depth. The exact encoding of the queue depth information and
`other information which may be required is FFS_
`
`2) A TUEID would be assigned
`
`NK868|TCO1731 7126
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1006-00003
`
`
`
`3) UEs would then monitor the ACCH for a SFA.
`
`4) Upon receiving an assignment, UEs would use measured information on the path loss and SFA to calculate
`the initial transmit power.
`
`5) The UEs would then transfer the data on the USCH while performing fast power control using a common
`power control channel as indicated on the PCPA. As part of the transfer the UE may also piggyback queue
`depth information, this and other potential techniques for uplink signaling are FFS.
`
`6) The UE would continue to monitor the ACCH for subsequent SFAs until the queues are empty.
`
`The use of the ACCH for managing multiple traffic types in the uplink provides many benefits to the UMTS
`system. Scheduling of traffic is centralized allowing the system to better manage utilization. System control
`allows the scheduling to be adjusted on a frame by frame basis accommodating new reservation requests with
`varying levels of QoS. UEs with a higher priority and lower delay tolerance could be apportioned power before
`users transferring bulk-rate traffic. Finally, the system is better able to manage varying traffic types during
`periods of congestion.
`
`5.0 Sizing the Downlink Shared Control Channel
`It
`is proposed to aggregate the functions of power control, dynamic persistence, downlink OVSF code
`assignment and uplink noise rise apportionment onto the downlink access control channel (ACCH) .
`It is
`proposed that the ACCH transport channel be carried over a third common control physical channel referred to
`as the Tertiary Common Control Channel Physical Channel. This tertiary CCPCH is efficient in terms of code
`usage and provides scheduling flexibility.
`
`The tertiary CCPCH would carry Common
`illustrates a proposed structure for a Tertiary CCPCH.
`Figure 1
`Transmit Power Control (CTPC) information, a Dynamic Persistence Indication (DPI), Common OVSF Code
`Assignments [12] and uplink spreading factor assignments (SFA). For flexibility, it is assumed that the tertiary
`CPCCH may operate at multiple spreading factors and be sized appropriately depending on tie traffic load in the
`cell (the rate and spreading factor of the tertiary CCPCH would be broadcast on the BCH). Each Power Control
`Group (PCG) slot will contain pilot information and 8 power control feedback bits for each of a maximum of 8
`individual UEs transmitting on the uplink.
`The remaining bits per PCG may be encoded with rate l/3
`convolution code and interleaved over the 10 ms frame to enhance the reliability of the DPI, and the DSCH and
`USCH assignment information. The tertiary CCPCH is not power controlled and is transmitted over the entire
`cell.
`
`‘NPi1Ot* NTPC <eND,,,T>
`TPC
`TPC
`TPC
`TPC
`TPC
`TPC
`TPC
`TPC
`.
`.
`Pilot
`
`
`
`#0
`#1
`#2
`#3
`#4
`#5
`#6
`#7
`Coded Assignment Information
` 0.625 ins, 2o*2k bits (k:O..6)
`
`
`
`Frame #72
`
`Figure l The structure of the Tertiary CCPCH
`
`Starting with a
`Table 2 examines the number of UEs that can be serviced for varying spreading factors.
`spreading factor and considering the structure of the tertiary CPCCH illustrated in Figure l, the table first
`calculates the available coded information bits per PCG slot, Ndata_ The fields Npilot and NTPC represent the
`number of bits assigned to pilot and power control feedback, respectively. Next, the maximum number of
`available information bits per frame is calculated based on the frame rate, CRC, tail, and coding rate. Finally,
`the number of assignments per frame is calculated by dividing the difference of information bits minus DPI by
`
`NK868|TCO1731 7127
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1006-00004
`
`
`
`the sum of one DOCA and UPA assignment. The Downlink OVSF code Assignment per user is defined in [12].
`The UPA was defined earlier in Table 1.
`
`At a minimum the tertiary CCPCH will require a spreading factor of 128 accommodating 3 UEs in both direction
`or 6 UEs in one direction. A spreading factor of 64 should be large enough to service most needs
`accommodating 11 UEs. As an alternative, a rate 1/2 code is considered providing an assignment of 6 UEs in
`both directions or maximum of 12 in either.
`
`Table 2 Calculation of the number of UEs supported for various spreading factors and coding rates.
`
`S"’ea"'“9
`Factor
`256
`128
`64
`128
`
`N
`
`pm
`8
`8
`8
`8
`
`N
`
`WC
`8
`8
`8
`8
`
`Ndata
`4
`24
`64
`24
`
`Re itition
`p
`1
`3
`1
`0
`
`9
`
`Codin
`1/3
`1/3
`1/3
`1/2
`
`(ms)
`10
`10
`10
`10
`
`CRC
`16
`16
`16
`16
`
`Tail
`8
`8
`8
`8
`
`'”f° MS
`available
`.3
`103
`317
`168
`
`DPI
`10
`10
`10
`10
`
`DOCA/user UPA/user
`13
`12
`13
`12
`13
`12
`13
`12
`
`Users
`o
`3
`12
`6
`
`6.0 Conclusion
`
`Several mechanisms for managing bandwidth and interference have been proposed. A dynamic persistence
`mode offers a sensible approach for controlling congestion on the RACH taking advantage of the centralized
`architecture of a cellular system and allowing the network to dynamically manage the backoff of all subscribers
`based on estimates of the congestion in the cell. A centralized uplink packet scheduling policy was proposed
`that apportions noise rise to UEs allowing the system to schedule traffic on the broadest possible pipe and
`allowing scheduling to be adjusted for varying levels of Quality of Service. As a result, the system is better able
`to manage varying traffic types during periods of congestion. Finally, the size of a common downlink control
`channel supporting all UCD traffic inechanisins is examined showing that all
`the assigrniierits may be
`accommodated on a channel operating at a spreading factor of 64 or 128.
`
`7.0 Proposed modifications to standards
`
`Add a section after section 5.3.2.2 ofXX. 03
`7.1
`5.3.2.x Tertiary Common Control Physical Channel
`
`The tertiary CCPCH is used to carry downlink access control channel (ACCH) information. The ACCH
`aggregates the functions of uplink power control for packet data users, OVSF code assignment, dynamic
`persistence and noise rise apportionment. The ACCH is a low fixed rate channel (32 ksps or 64 ksps) which is
`coded and interleaved. The rate of ACCH may be different from cell to cell. The rate and spreading factor of the
`ACCH is broadcast on the BCCH. The frame structure of the ACCH is shown in Figure x {refer Figure 1 of
`above}. The ACCH is not power controlled and is transmitted over the entire cell when there are data packets to
`transmit on downlink or uplink or on both directions.
`
`NK868|TCO17317128
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1006-00005
`
`
`
`7.2
`
`Modijfjz thefigure 11 in section 6 ofXX. 03
`
`Transport Channels
`
`Physical Channels
`
`BCCH Primary Common Control Physical Channel (Primary CCPCH)
`
`FACH
`PCH _:
`ACCH Tertiary Common Control Physical Channel (Tertiary CCPCH)
`
`Secondary Common Control Physical Channel (Secondary CCPCH)
`
`RACH Physical Random Access Channel (PRACH)
`
`DCH Dedicated Physical Data Charmel (DPDCH)
`
`Dedicated Physical Control Channel (DPCCH)
`
`Synchronisation Channel (SCH)
`
`Figure11: Transport-channel to physical-channel mapping.
`
`
`
`7.3
`
`Add abbreviation in Section 3.3 ofXX.07
`
`ACCH
`
`Access Control Channel
`
`7.4
`
`Addfollowing steps to the random access procedure in Section 6 ofXX. 07
`
`3. The mobile station reads the ACCH to get information about:
`
`3 .1 The current dynamic persistence Value.
`
`8.0 References
`
`[1] Ericsson, “Modifications of the current RACH scheme for increased throughput”, SMG2 UMTS-L1 455/98.
`] Philips, “Modification to the RACH Scheme”, SMG2 UMTS-L1 533/98.
`]
`\/lotorola, “Comparisons of power ramping schemes for RACH”, SMG2 UMTS-L1 564/98.
`] Sampath, Mandayam, and Holtzman, “Analysis of an Access Control Mechanism for Data Traffic in an
`Integrated Voice/Data Wireless CDMA System”, VTC 1996
`\/lotorola, “Comparison of techniques for fast power control in FDD”, SMG2 UMTS L1 565/98
`]
`\/lotorola, “Channel Bandwidth Allocation Strategy”, SMG2 UMTS L23 534/98
`]
`\Tol<ia, “Efficient Radio Resource Usage with Shared Channels”, SMG2 UMTS L1 423/98
`]
`] Lucent, “Downlink Shared Channel (DSCH) Transmission”, SMG2 UMTS L1 436/98
`] Sony lntemational, “Multiplexing packet Data in FDD mode”, SMG2 UMTS L23 266/98
`0]\Iortel, “Analysis of the available proposals for shared channels and proposed merged solution”, SMG 2
`UMTS L1 584/98
`
`[11]\/lotorola, “Common power control charmel for UMTS FDD”, SMG 2 UMTS L1 440/98
`[12] \/Iotorola, “Shared Channel Options for Downlink Packet Data Transmission,” SMG 2 UMTS L23 533/98
`
`NK868|TCO1731 7129
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1006-00006