`
`ZTE Corporation and ZTE (USA) Inc.
`
`
`
`RX-3034
`
`
`Inv. No. 337-TA-868
`
`Shared Channels for Packet Data Transmission in w-CDMA
`
`Amitava Ghosh, Mark Cudak“ and Ken Felix
`
`Motorola Network Solutions Sector, 'Motorola Labs
`
`1501 West Shore Drive, Arlington Heights, IL 60004
`
`ga0047@email.mot.com, am6005@;wil.mot.com, cellQ_6_@_emuil.mot.com
`
`In the simulation, the data traffic is modeled as distinct
`sessions with a Poisson arrival process. Each session
`marks a period of higher activity comprised of a number
`of packet calls. The number of packet calls per session is
`geometrically distributed while the time between packet
`calls
`is exponentially distributed.
`If
`the
`session
`modelling is for web browsing,
`then the packet call
`models a web page download, while the arrival
`time
`models the think time used to peruse the web page. Each
`packet cal] consists of one or more packets whose inter—
`arrival time and length are both exponentially distributed.
`The parameters associated with the data traffic model are
`summarized in Table 2.
`
`single service (Web
`
`Table 1 Simulation results for
`Browsing) implementation.
`Number
`of packet
`
`Average
`Transmis-
`sion lime
`(seconds)
`
`Total
`delay
`time
`(seconds)
`
`
`
`Abstract: One of the targets for a Wta'eband CDMA (W-
`CDMA) system is to provide on efliciem‘ transfer of low
`bit rate to very high bit rate packet data services. A
`shared channel concept has been proposed for both the
`downlink and uplink channels that is better suited for
`bursty packet data trafiic [I]{2][3][4].
`The shared
`channels allows UEs to transmit and receive data bursts
`at high rates by using short leases on the radio resource
`thereby lowering the overall delay by taking the greatest
`advantage of statistical multiplexing.
`The high rate
`bursts require the network tightly manage its resources
`to insure that the appropriate OVSF codes are asaigned
`on the downllnk and the aggregate interference does not
`exceed the noise rise on the uplink.
`In both cases, the
`network must convey the new assignments on a frame by
`flame basis.
`In this paper, packet data transmission
`using both Downlink Shared Channel
`(DSCH) and
`Uplink Shared Channel (USCH) is discussed for a W-
`CDMA system.
`
`1. Introduction
`
`The Shared Channel originated from the well known
`observation that packet call QoS is greatly enhanced by
`the use of fat pipe multiplexing where the scheduling of
`packet
`users
`is
`done
`collectively
`rather
`than
`stochastically. Section 2 presents the benefits of fat-pipe
`multiplexing for web browsing. The DSCH concept,
`introduced in Section 3, provides a method for fat—pipe
`scheduling of downlink packets by dynamically sharing
`the power
`and code
`resource
`among users
`thus
`overcoming the problem of downlink code shortage. The
`USCH concept,
`introduced in Section 4. provides a
`similar scheduling mechanism for the uplink. Although
`not code limited, the USCH requires that the notwork
`tightly manage the power resource while scheduling the
`uplink data packets. In Section 5, simulation results are
`presented composing possible control channel Options for
`communication assignments, either on dedicated or
`common channel, and the performance of continuous and
`discontinuous
`packet
`data
`transmission.
`Finally,
`conclusions are drawn in Section 6.
`
`2. Fat Pipe Multiplexing
`
`The benefits of fat pipe multiplexing are shown using
`simulation. Table
`] presents
`the
`results
`for Web
`browsing session over a 307 kbps channel scheduled
`together as one composite charms] (Le. a fat-pipe) or
`multiple dedicated channels. The results presented were
`for web browsing; however, FTP and email transfer have
`been considered as well [4].
`
`Table 2 Web Browsing Modcl.
`
`
`
`
`
`
`
`
`
`
`
`
`
`average web page
`reduction in the
`A significant
`download is achieved by using fat-pipe multiplexing.
`Table 1 shows the system performance in terms of the
`average
`queue
`delays.
`the
`average
`packet
`call
`transmission time, and the total time the packet call is in
`the system for a 75% system load. The latter is the sum
`of the queuing delay and transmission time. The results
`show that as long as the utilization is under 100%, the
`average time' in the system is lower when fewer channels
`
`‘ The average time in the systcm is defined as the time from when a
`transmission request
`is made till
`the time the.
`transmission is
`completed.
`
`0480364354199” 1 0.00 © 1999 [BEE
`
`943
`
`VTC ‘99
`
`NK868IT0011070394
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1030-00001
`
`
`
`of greater bandwidth are implemented. As one would
`expect, as the number of channels increases the average
`queue delay decreases even though the bandwidth of
`each channel decreases. However, the increased queuing
`delay is more than compensated by the decrease in
`transmission time on larger bandwidth channels. These
`characteristics hold true in general provided the channel
`holding time distribution is
`suflicienfly regular {5].
`Considering that the end users perception of Quality of
`Service for data applications is limited to the time
`interval from when the request for service is submitted to
`when the service has completed,
`the system performs
`better when the allocated bandwidth is configured into
`fewer communications channels of larger bandwidth. As
`a result, both downlinlt and uplink shared channel would
`provide these scheduling benefits for W-CDMA.
`3. Downlink Shared Channel (DSCH)
`
`The DSCH provides a method for sharing code and
`power resources to overcome a potential code shortage
`when bursty data traffic is typical. When assigned in a
`conventional manner on a Dedicated Channel (DCH), a
`User Equipment
`(UE)
`is
`reserved a portion of the
`Orthogonal Variable Spreading Factor (OVSF) code tree
`for the duration of the web browsing session based on the
`peak data throughput. As an example, consider the code
`tree in Figure l where seven 384 kbps UEs operating at
`activity rate of 10% are allocated resources. Notice that
`87% of the code tree is consumed. However, if codes are
`re-assigned dynamically on a frame-by-framc basis only
`14% of the code tree is consumed since only one OVSF
`code with SFwB may be shared between seven users.
`This latter case is depicted in Figure 2.
`am
`
`5.
`“Mg;
`
`.,
`
`Figure 1 OVSF Code Tree for DCH.
`33-1
`
`IFII we:
`
`
`“II
`
`.:
`
`
`mu .
`“35.4900;
`
`Figure 2 OVSF Code Tree for DSCH.
`
`The OVSF codes for the DSCH can be assigned using i)
`multiple DCH's or ii) a DSCH control channel which is
`also termed as Physical Shared Channel Common
`Control Channel (PSCCCH). The slot structure of the
`DSCH when associated with a DCH is shown in Figure
`3. This is a special case of multicode transmission where
`the low rate DCH is spread by a fixed rate OVSF code
`
`known to the UE and the spreading factor of the DSCH
`varies from frame to frame. The DSCH only carries the
`data field whereas the DCH comprises of Pilot, TPC,
`'I‘FCI and Data field. The slot structure of the DSCH
`when associated with a PSCCCH is shown in Figure 4.
` Slol 0.635 In:
`
`DFCCH DPDGH
`
`
`
`Figure 3 Slot structure of DSCH associated with DCH.
`Slc|0.625 ms
`
` +—-—————*fi“—-—'—-—-————D
`
`PSCCCH
`
`Figure 4 Slot
`PSCCCH.
`
`structure of DSCH associated with
`
`4. Uplink Shared Channel (USCH)
`In the case of the USCH there is no code limitation as
`
`each [IE will have its own scrambling code. However.
`there is still a limited power resource, hence the USCH
`represents a shared power
`resource.
`The USCH
`coordinates the fast scheduling of uplink data packets so
`as
`to insure a uniform interference power profile
`protecting. voice users.
`In the USCH concept. each
`active HE is assigned a fraction of total noise rise, which
`translates into 3. Spread Factor (SF) assignment. Similar
`to that of a DSCH, the data rate of the USCH can be
`reassigned on a frame-by—frame basis. The signaling for
`the USCH is conveyed on a PSCCCH. The structure of
`PSCCCH which aggregates
`the functions of power
`control, downlink OVSF code assignment and uplink
`spread factor assignment
`is shown in Figure 5. The
`PSCCCH carries Common Transmit Power Control
`(CTPC) information, a Dynamic Persistence Indication
`(DPI), Common OVSF Code Assignments and Uplink
`spreading factor assignments (SPA). For flexibility, it is
`assumed that the PSCCCH operates at multiple spreading
`factors and is sized appropriately depending on the
`traffic load in the cell. The rate and SF of the PSCCCH
`
`would be broadcast on the BCH. Each slot contains 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 slot are encoded with a
`rate 1/3 convolution code and interleaved over the 10 ms
`
`frame to enhance the reliability of the DPI as well as the
`DSCH and USCH assignment
`information. The
`PSCCCH is not power controlled and is transmitted over
`the entire cell.
`
`Since the assignment and reassignment of the power
`resource to various packet data users are made on a
`frame by frame basis, the convergence of the fast power
`control loop, availability of good channel estimates, and
`searcher performance are critical for the operation of the
`USCH. A method for power control and channel
`
`onus-5435—4/99/stom o 1999151313
`
`944
`
`VTC ‘99
`
`NK868ITC011070395
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1030-00002
`
`
`
`link maintenance channel between the
`bi-directional
`packet bursts. The link maintenance consists of power—
`conno] commands and pilot. symbols needed to preserve
`power control and synchronization of the dedicated
`physical channel. However, there is a cost associated
`with the use of the link maintenance channel since code
`and power resources are consumed even though no data
`is being transmitted. The cost increases linearly with the
`number of users engaged in a packet call.
`
`Table 3 Simulation Results for Various Shared Channel
`Utilizations.
`Shared
`Likelihood a
`Mean number of
`
`
`
`
`Simultaneous Users
`Channel
`Common Channel
`
`
`
`
`Utilization
`will be more
`
`
`Efficient
`
`
`
`
`
`
`
`
`Analytic -Siniulated-lO:l
`
`
`
`20:1
`
`
`
`
`
`
`
`mm
`
`
`"M-
`
`
`
`
`estimations for bursty data traffic is introduced in Section
`5 along with some simulation results.
`«mm Nm—wu—Iv <—~——N,,_,-———-—-—w—>
`commence“
`
`not: no. mn'hiu (ta-M)
`
`
`
`
`Figure 5 Structure of the PSCCCH.
`5. Simulation Results
`
`The use of PSCCCH in association with DSCH is
`efficient when the number of simultaneous packet data
`calls exceed a certain threshold.
`The packet data
`simulation that was used in Section 2 to present the
`benefits of a "fat—pipe” scheduling was modified to
`monitor the number of simultaneous packet calls. For
`this investigation, the simulation models the scheduling
`of multiple users on a 384 kbps shared channel.
`The
`model parameters are identical
`to those presented in
`Section 2. The simulations runs were conducted for 75%.
`90%, 92%, and 95% shared channel utilizations based on
`a session arrival rate of 0.60, 0.70, 0.72, and 0.76 per
`second, respectively Figure 6 presents the cumulative
`probability of N or more simultaneous packet calls. Table
`3 tabulates the statistics for the simulation runs. The
`simulation shows the dedicated control channels will be
`more efficient with respect to power budget for low
`utilizations, however, the common channel will be more
`efficient when resources are needed the most. Assuming
`a 10:1 ratio of PA resources [6] required to support one
`common channel versus one dedicated channel,
`the
`simulation shows that the common channel will be more
`
`efficient in terms of power budget when shared channel
`utilization is above 92%. The common control channel
`bounds the PA resources consumed by shared channel
`signaling bolstering the stability of the shared channel
`during periods of peak loading.
`
`
`
`Figure 6 Probability of N or more Simultaneous Packet
`Users.
`
`loop and the
`Convergence of the fast power control
`availability 0t" good channel estimates at the base station
`are critical in the case of USCI-I where packets arrive in a
`bursty manner [8]. A simple solution is to use a low rate
`
`for
`the modified approach
`illustrates
`10
`Figure
`discontinuous packet data transmission as in the case of
`the USCH. Three cases are considered, a) packets are
`transmitted only in the downlink ustng DSCH, h) packets
`are transmitted only in the uplink using USCH and c)
`packets are transmitted in both the uplink and downlink
`direction.
`In all
`the cases an upiink channel
`is
`maintained so as to convey the power control bits for
`forward link, piggybacking information and/or to carry
`the data using the USCH. To prime the fast reverse link
`power control loop, searcher and channel estimator, the
`transmission of preamble using DPCCH starts one frame
`(16 slots) prior to the scheduled uplink or downlink
`packet data transmission. The preamble is transmitted
`with an additional negative power offset
`from the
`computed open loop estimate. Further, the initial power
`control step size for transmitting the preamble is set at a
`higher value (eg. 2dB) so that power control
`loop
`converges faster if the UE is in a deep fade. On the
`receipt of the first down power control command at the
`UE during the preamble transmission phase, the step size
`reverts back to normal power control (PC) step size {e.g.
`ldB). It may be noted that the step size always resets to
`its normal setting in the beginning of actual packet data
`transmission.
`
`Next, simulation results are presented using a W»CDMA
`reverse link simulator for the above three cases.
`In this
`simulator, the packet data source model is represented by
`a function that produces a random draw of a new packet
`transmission time along with a subsequent inter-arrival
`time before the next packet. If the generated inter-arrival
`time occurs prior to the completion of transmission of the
`generated packet,
`a
`concatenation process occurs.
`Another packet is generated and appended to the original
`packet. This process continues until there is time for idle
`
`.
`
`0-7303-5435-4lQ9I$10.00© 1999 IEEE
`
`945
`
`VTC ‘99
`
`N K868|TC01 1070396
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1030-00003
`
`
`
`.In the simulation the mean packet size is
`slots to occur.
`assumed to be 480 bytes and a code rate of two blocks
`per frame is used, which translates into 144 kbps UDD
`service. Figure 7 Shows the GDP of consecutive idle
`frames within a packet call for system utilization values
`of 75% and 92%.
`It may he observed from the figure
`that as the systetn utilization increases the number of
`consecutive idle frames within a packet call
`also
`increases. The increase in idle frames is caused by a
`greater reliance on time multiplexing which occurs due to
`the higher likelihood of multiple active packet calls
`sharing the channel.
`
`
`
`for a chip rate of 4.096Mcps and carrier frequency of
`2.0GI-Iz for a flat fading channel. It may be noted that the
`simulations did not use the negative power offset or the
`varying step size of Figure 10. The three cases simulated
`were
`a)
`continuous packet data
`transmission. b)
`discominuous
`packet
`data
`transmission without
`preamble, and c) discontinuous packet data transmission
`with one frame preamble transmission. Simulations were
`run at three values of vehicle speeds (3, 30 and 120
`kmph) under flat fading channel conditic‘ms for various
`values of system utilization.
`Table 5 and Table 6
`summarize the received Eb/Nt for target Frame Erasure
`Rate (FER) of 10% and 1% at a system utilization of
`75% and 92% respectively. Figure 8 and Figure 9 give a
`pictorial representation of the summary.
`
`
`
`
`5
`
`6.1
`
`3.5
`
`3-5
`
`Figure 7 Consecutive idle frames within a packet call for
`various values of system utilization.
`
`Table 4 Simulation Parameters.
`
`
`
`
`
`
`
`
`
`
`Table 6 Received Eb/Nt at 1% and 10% at a system
`utilization of 92%.
`
`Discontinuous Discontinuous with
`‘
`
`
`mm
`mm
`Meagan
`
`
`
`10% FER 1%FER
`10% PER
`
`1.7
`
`
`
`-
`
`-—n-
`
`
`
`
`
`
`
`
`—_
`
`
`—_
`
`
`__ a. At high and medium values of vehicle Speed the
`
`
`to discontinuous packet data
`degradation due
`—1%_ that the power control does not track the fading at
`
`
`Power Control Delay
`1 slot
`transmission is within 0.5dB. This is due to the fact
`
`
`[TOT
`Eower
`Control
`Feedback
`
`
`
`
`-_ c. At operating FER of 10% the loss in performance
`due to discontinuous packet data transmission is less
`than at the 1% FER operating point.
`
`The following observations are made from Table 5 and
`Table 6.
`
`high values ofDoppler.
`b. The performance degradation at slow speed at 1%
`FER and high system utilization is approximately
`2.3.
`
`The W-CDMA chip level link simulator was used to
`evaluate
`the
`performance
`of
`continuous
`and
`discontinuous packet data transmission.
`The above
`source model was used to model
`the discontinuous
`packet data transmission. The parameters used in the
`simulator are shewn in Table 4. The simulation was run
`
`|
`_
`d. The performance at‘slow speeds is improved for
`discontinuous transrnlsston With one frame preamble
`Izransnussron.
`
`0-7803-5435-4199/slono o 1999 IEEE
`
`946
`
`VTC ‘99
`
`NK868ITC011070397
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1030-00004
`
`[HI-It]I
`
`I I I
`
`. E".-
`
`--
`I I
`
`Table 5 Received Eb/Nt at 1% and 10% at a system
`utilization of 75 %.
`Continuous
`(dB)
`
`Discontinuous
`(dB)
`
`Discontinums with
`Preamble
`(dB)
`
`
`
`
`
`[7] Motorola, “State Occupancy Estimations for Shared
`Channel Concept, TSGR]#2(99)066'
`[8] Motorola,
`“Discontinuous
`Transmission,” TSGR1#2(99)064.
`7 on i._ ,. W. i.
`
`Packet
`
`Data
`
`
`
`
`“b- Dimmdnwnwith Paw-hie "0%)
`
`
`/
`-
`
`sammm
`:gmflng-WW-“t‘
`momma»,
`
`0
`
`m
`
`‘2
`
`m
`
`m
`
`mo
`
`I20
`
`Figure 8 Received Eb/Nt vs Speed at 75% Utilization.
`7
`_e-._.__....._w_1..__.__~._.._...
`
`{'56 “m
`
`I
`
`6. Conclusions
`The Shared Channel originated from the well known
`observation that packet call QoS is greatly enhanced by
`the use of fat pipe multiplexing where the scheduling of
`packet
`USCl'S
`is
`done
`collectively
`rather
`than
`stochastically. For W-CDMA this equates to doing a
`frmIB-by-frame scheduling of resources allocating short
`leases on high data rate transmissions. The DSCH
`provides a method for fat-pipe scheduling of downlink
`packets by sharing a low SF (high data rate) OVSF code
`between multiple packet data users. Similarly, USCH
`also provides a method for fat pipe scheduling using a
`common
`control
`channel
`for
`transmitting
`the
`assignments.
`Finally,
`the use of one frame preamble
`helps to prime the searcher, power control and channel
`estimator
`for
`discontinuous
`packet
`data
`transfer
`associated with the USCH.
`
`Acknowledgements
`
`The authors would like to thank MyMy Nguyen and Nick
`Tolli for carrying out the simulations and to Bob Love,
`Steve Barrett and Lousy Jalloul for many stimulating
`discussions.
`
`References
`
`[1] Motorola, “Shared Channel Options for Downlink
`Packet Data Transmission,” Tdoc SMGZ UNITS-Ll
`681/98.
`
`[2] Nokia, “Utilisation of UTRA FDD Downlink Shared
`Channels", Tdoc SMGZ UMTS-L23 29698.
`
`for Managing Uplink
`“Mechanisms
`[3] Motorola,
`Interference and Bandwidth,” Tdoc SMGl UMTS-Ll
`683/98 -
`
`[4] Motorola, “Channel Bandwidth Allocation Strategy,”
`Tdoc SMGZ UNITS-Ll 682/98.
`[5] Ronald W. Wolff, Stochastic Modeling and the
`Theory of Queues, Englewood Cliffs, NJ:
`[1998]
`Pram“ Ha“
`[6] Motorola, “Power Requirement for Common vs.
`Dedicated Channel,” TSGR1#3(99)382.
`
`
`
`
`
`4
`
`i“
`
`2
`
`.
`
`+ lemmlfl 5)
`-I- olwmumm (to)
`
`:2:m:;;f“m“““’
`iscnnl uani
`m In!»
`:Elumuij‘wmuan hm
`
`w
`M
`
`0.,
`
`m
`
`m
`
`m
`(Woman)
`Figure 9 Received Eb/Nt vs Speed at 92% Utilization.
`
`m
`
`Chin
`mum-a meg-I
`MCCEH
`USCH
`HAP-H
`DFCCH
`um
`upunh maul
`
`aim-II
`
`Figure 10 Discontinuous packet data transmission using USCH.
`
`macs-5435.4I99/sioco o 1999 IEEE
`
`947
`
`VTC ‘99
`
`NK868ITC011070398
`
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1030-00005
`
`