`
`ZTE Corporation and ZTE (USA) Inc.
`
`
`
`Inv. No. 337-TA-868
`RX-3087
`
`R1-02-1371
`
`3GPP TSG-RAN1 #29
`Shanghai, China
`November 5-8, 2002
`
`Agenda item:
`
`HSDPA Enhancements
`
`Source:
`
`Title:
`
`Lucent Technologies
`
`Code limitation and code reuse in HSDPA
`
`Document for:
`
`Discussion
`
`1
`
`Introduction
`
`The orthogonal code space is an important system resource for WCDMA DL transmission. Orthogonal transmission is
`achieved by allocating OVSF codes to different control and data cl~annels. For HS-DSCH, multiple SF=16 codes can
`be used for DL transmission within a subframe. The HS-DSCH provides a highly efficient radio link and can
`potentially support a large number of users. However, the capacity of HS-DSCH can be limited considerably due to a
`shortage of available orthogonal codes.
`
`1.1
`
`Causes of Code Limitation in HSDPA
`
`The shortage of codes can result due to various reasons including inefficient code space usage by associated DPCH for
`HSDPA users (basically carrying pilot and power control information) and other dedicated chamlels [1]. For R~99 data
`services, an inactivity timer is typically employed to ensure that code resources are released for other users. However,
`there are certain data applications (e.g. chatty applications, TCP acknoMedgements) for which long inactivity timers
`may be needed in order to ensure low- delay. As a result, these applications tend to use up significant fractions of the
`code space but have very low power requirements. Voice users could also make ineftScient use of code space since
`codes remain assigned during periods of inactivity. This leads to power code imbalance, an cftcct further compounded
`by soft handoff on the do~flink. Enhancements that provide power benefit (e.g. beamforming) also need
`corresponding improvements in the code dimension so tl~at the system capacity benefits can be realized. These effects
`can result in a disproportionately large amount of power available t’or HS-DSCH as compared to codes.
`
`1.2 Effect of Code Limitation on HSDPA
`
`In this section, we show how code limitation can substantially constrain the throughputs achieved in HSDPA.
`tS-ecisel~, we present the OTA and service throughput attained using single antenna (i.e., no transmit diversity) and
`CLTD Mode-1. In particular, we consider two cases: one where only 25% of the code space is available tbr HSDPA
`(i.e., the s~:stem is code space constrained) and another where 62.5% of the code space is available (i.e., the system is
`not code space constrained). In both cases, 63% of the overall power fraction is assumed to be available for HSDPA.
`Figure 1 and Figure 2 below plot the OTA and service throughput attained in each of the above cases in a single path
`Rayleigh fading channel model. Observe that in the code-constrained case, the service throughput and OTA remain
`virtually unchanged as the load in the system - measured in terms of the number of UEs - increase. This clearly
`illustrates the et~ct on system capacity due to the code-power imbalance in the code-constrained case.
`
`With thc availability of 2 transmit antennas, transmit diversity schemes such as closed loop (TxAA) can be used for
`HS-DSCH [2]. These transmit diversity techniques provide link level performance improvements leading to overall
`system capacity improvement. However, the gains achieved by the transmit d~versity schemes can be small when the
`system is code limited. This is due to the l~act that trax]smit diversity provides power benefit by achieving a given
`performance with smaller power (Eo/Io~) required compared to a single antenna transmission but does not help solve
`the code limitation problem.
`
`The code limitation problem can be partly resolved by using higher coding rates and high order modulations.
`However, high coding rates and higher order modulations come w~th associated penalties thus impacting the overall
`system capacity. Furthermore, in some cases, higher order modulations may not be available - the highest order
`
`NK8681TC013089781
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00001
`
`
`
`modulation available for HS-DSCH is 16QAM. Moreover, the introduction of HSDPA UE classes that only support
`QPSK ensures that higher rates can be achieved only by using a larger number of codes. MIMO techniques can also
`solve the code limitation problem but require multiple transmit and multiple receive antennas.
`
`In what follows, we address the code limitation problem through OVSF code reuse approaches that allow higher
`system capacity to be achieved without requiring multiple receive antennas.
`
`3500
`
`3000
`
`2500
`
`2000
`
`1500
`
`1000
`
`500
`
`0
`
`3500
`
`3000
`
`2500
`
`2000
`
`1500
`
`1000
`
`500
`
`0
`
`[] CLTD (25% code space)
`
`[] SA (62.5% code space)
`
`I N S A (25% code space)
`
`[] CLTD (62.5% code space)
`
`37 UEs
`
`56 UEs
`
`75 UEs
`
`100 UEs
`
`Figure 1. OTA with single antenna and CLTD Mode-1.
`
`[] CLTD (25% code space)
`
`[] SA (62.5% code space)
`
`I N S A (25% code space)
`
`[] CLTD (62.5% code space)
`
`37 UEs
`
`56 UEs
`
`75 UEs
`
`100 UEs
`
`Figure 2. Service throughput with single antenna and CLTD Mode-1.
`
`NK8681TC013089782
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00002
`
`
`
`2 Code Reuse
`
`The reuse of OVSF codes results in additional in-sector interference that needs to be managed through suitable
`techniques (interference averaging, interference cancellation and/or interference rejection) in order to achieve the
`desired benefits without a significant power penalty.
`
`R’99 specifications allow the OVSF code space to be expanded through the use of one or more Secondary Scrambling
`Codes (SSCs) [3] (also see related references on code space expansion through QoFs and related topics [4][5]). Here,
`we explore both the use of SSCs and/or multi-antenna scheduling techniques to achieve code reuse in HSDPA with
`effective management of resultant interference.
`
`2.1
`
`Code Reuse Options
`
`Two alternatives for code reuse in HSDPA are now described:
`
`Option-l: Partial Code Reuse - Here, only the codes set aside for HSDPA are candidates for reuse (see Figure
`1). Reusing the codes set aside for dedicated channel users is not an option as they do not get any processing
`gain benefit in rejecting the cross interference. This approach does not expand the code space to the maximal
`extent but has the advantage of maintaining orthogonal transmissions to the dedicated channel users, thereby
`leaving the interference to them unchanged. The basic transmission approach suggested here is to
`simultaneously schedule data intended for different IdEs by utilizing two transmit antennas and reusing the
`available OVSF space for HSDPA UEs. The fundamental principle exploited in the simultaneous scheduling
`of multiple users with code reuse is to ensure that users selected on each antenna have good "cross-antenna
`rejection" [6], i.e., a user scheduled on antenna 1 has a strong charmel from that antenna and a comparatively
`weak channel from antenna 2. Users on antenna 2 are selected in the same manner. This will keep the cross
`interference experienced by each user low. With sufficient load, pairs of users that satisfy this condition can
`be found in the cell with high probability. The power distribution across the two antennas can be performed
`in a number of ways. A useful method is to split the total cell power equally across the two antennas (in a
`manner similar to STTD and CLTD Mode-l).
`
`Ant 1
`
`Ant2
`
`Codes used by control and
`dedicated channels.
`
`Codes not used (No
`transmission).
`
`Codes available for HS-
`PDSCH.
`
`HS-PDSCH codes reused.
`
`V
`
`Primaa’y scrambling code
`
`Figure 1. Partial code reuse scheme
`
`Option-2: Full Code Reuse - If a secondary scrambling code is activated, then the entire OVSF code space on
`the SSC is available. So HSDPA users can be scheduled on a part of the Primary Scrambling Code (PSC)
`space originally assigned and the entire SSC code space. The introduction of a SSC allows full reuse of the
`OVSF code space and can be achieved in the following ways:
`
`NK8681TC013089783
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00003
`
`
`
`Introduction of SSC on Same Antenna(s) as PSC: The use of a SSC on the same antenna(s) as a
`Primary Scrambling Code (PSC) may lead to excessive interference to UEs on both the PSC and
`the SSC since the instantaneous channel gains on the desired signal and interference
`components are perfectly correlated. Multi-user detection techniques can be applied to cancel out
`the resulting interference, albeit at the cost of additional UE complexity [5].
`
`2.
`
`Introduction of Antenna Specific SSC: This allows cross antenna interference rejection through
`scheduling, in a manner similar to that in Option 1 above, in addition to the interference
`rejection benefits of the SSC itself..
`
`Although dedicated channel users on the PSC get the processing gain benefit in rejecting SSC interference,
`unlike in Option 1, some additional interference due to the SSC cannot be avoided. This increase in
`interference will therefore have to be offset by increasing the power allocation to dedicated channel users.
`Although this does result in a reduction in the overall power fraction left for the data users, the increase in the
`code space available due to the introduction of the SSC still results in system capacity improvement in code-
`limited situations.
`
`Depending on the severity of code limitation encountered, full or partial code reuse may be employed in order to
`expand the OVSF code space. These techniques can be combined with other interference avoidance techniques (e.g.
`beamforming) in order to further improve performance. The signalling support needed for full or partial code reuse is
`for further study. Additional signalling - both in the uplink and downlink - will improve the performance of code
`reuse and improve the robustness of the schemes.
`
`3 Conclusions
`
`Code reuse is proposed as a method of increasing the OVSF code space for HS-DSCH. The schemes can alleviate the
`code limitation problem and enhance HSDPA throughput in these situations. The quantitative benefits of these
`schemes and the signalling support necessary to accommodate them are for further study.
`
`References
`
`ill
`
`[2]
`
`[3]
`
`[4]
`
`[5]
`
`[6]
`
`"DL structure in support for HS-PDSCH", R1-01-0478, Qualcomm.
`
`"Transmit diversity for HSDPA", R1-02-0530, Lucent.
`
`3GPP TS 25.213 V5.2.0, "Spreading and Modulation (FDD)," Release 5.
`
`K. Yang, Y-K. Kim, P.V. Kumar, "Quasi-Orthogonal Sequences for Code-Division Multiple-Access
`Systems," IEEE Transactions on Information Theory, Vol. 46, No. 3, May 2000.
`
`L. Jalloul and A. Shanbhag, "Enhancing Data Throughput Using Quasi-Orthogonal Functions Aggregation
`for 3G CDMA Systems," Proceedings, Spring VTC, 2002.
`
`Achilles Kogiantis and Lawrence Ozarow, "Downlink Best Eftbrt Packet Data with Multiple Antennas,"
`submitted to International Conference on Communications, ICC 2003, Anchorage, Alaska, May 11-15,
`2003.
`
`4 Appendix: System Performance with Code Reuse
`
`In this section, we present some preliminary results indicating the benefits of code reuse in scenarios where HSDPA is
`code limited. In general, the improvement in system level performance possible with each of the code reuse options
`presented in Section 2.1 will depend on the severity of the code limitation, as also the degree of code-power imbalance.
`For example, if a large fraction of the total OVSF space is reserved for dedicated channels with a low activity factor,
`i.e., the system is severely code-constrained with a code-power imbalance, then option-2 will result in better
`performance because it allows for full code reuse; otherwise, option-1 can be deployed. In the following we present
`system level performance improvement possible with the code reuse options 1 and 2 described above in a severely code
`constrained scenario. For comparison, we will also present the performance of two baseline schemes: namely, single
`antenna without code reuse and CLTD Mode-1 without code reuse. The simulation assumptions are as follows:
`
`NK8681TC013089784
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00004
`
`
`
`¯ We assume that 4 of 16 Walsh-16 codes are available for HSDPA.
`
`¯
`
`¯
`
`The channel model is a single-path Rayleigh fading channel.
`
`The power fraction available for HSDPA in the baseline schemes and code reuse option-1 is 63%. This power
`fraction is equally distributed across the two antennas in the CLTD and code reuse option-1 cases.
`
`As noted in Section 2.1, introduction of the SSC in code reuse option-2 results in increased interference to
`dedicated channel users on the PSC. In order to compensate for this increased interference, it will be
`necessary to increase the power fraction for dedicated channel users, thereby reducing the total available
`power for HSDPA users1. In keeping with this observation, we consider two cases here: one where only 40%
`and the other where only 50% of the power fraction is available for HSDPA. For convenience, we assume that
`power is distributed equally across the two scrambling codes (and, antennas) for the HSDPA users. Thus, the
`power fraction used for each antenna and scrambling code is 20% and 25%, respectively.
`
`As the results below indicate both the code reuse options outperform the baseline schemes in service flaroughput and
`OTA (see Figure 4 and Figure 5). The packet call throughputs of the baseline schemes can, however, be better than
`option-1 in some cases. Packet call throughputs with option-2, however, are better than that of the baseline schemes in
`almost all the cases considered.
`
`[] SA (0.63)
`[] CLTD (0.63)
`[] Option-1 (0.63)
`
`[] Option-2 (0.4)
`[] Option-2 (0.5)
`
`37 UEs
`
`56 UEs
`
`75 UEs
`
`100 UEs
`
`Figure 3. Packet call throughputs in kbps with code reuse options 1 and 2 compared with the baseline single
`antenna system and multiple antenna CLTD Mode-1 system. In this case, only 4 of 16 Walsh-16 codes are available
`for HSDPA on the PSC. For each scheme, the power fraction available for HSDPA users is indicated within
`brackets in the legend.
`
`1 A simple analysis of the increase in power fraction necessary to maintain QoS for the voice users on the PSC due to the
`
`introduction of SSC is presented in Section 4.1.
`
`NK8681TC013089785
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00005
`
`
`
`] SA (0.63)
`[] CLTD (0.63)
`[] Option-1 (0.63)
`
`[] Option-2 (0.4)
`[] Option-2 (0.5)
`
`37 UEs
`
`56 UEs
`
`75 UEs
`
`100 UEs
`
`Figure 4. OTA in kbps with code reuse options 1 and 2 compared with the baseline single antenna system and
`multiple antenna CLTD Mode-1 system. In this case, 4 of 16 Walsh-16 codes are available for HSDPA. For each
`scheme, the power fraction available for HSDPA is indicated within brackets in the legend.
`
`] SA (0.63)
`[] CLTD (0.63)
`
`[] Option-1 (0.63)
`
`[] Option-2 (0.4)
`
`[] Option-2 (0.5)
`
`37 UEs
`
`56 UEs
`
`75 UEs
`
`100 UEs
`
`Figure 5. Service throughputs in kbps with code reuse options 1 and 2 compared with the baseline single antenna
`system and multiple antenna CLTD Mode-1 system. In this case, 4 of 16 Walsh-16 codes are available for HSDPA
`on the PSC. For each scheme, the power fraction available for HSDPA is indicated within brackets in the legend.
`
`4.1
`
`Impact of the use of SSC
`
`As mentioned in Sections 2.1 and 4, introduction of the SSC in code reuse option-2 generates interference to dedicated
`being carried on the PSC. Consequently, it is necessary to increase the power fraction allocated for these users in order
`to maintain quality of service. This, in turn, reduces the power fraction available for HSDPA. In this section, we
`present a sflnple analysis that quantifies the increase in power fraction allocated for voice due to the introduction of the
`SSC.
`
`NK8681TC013089786
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00006
`
`
`
`Let y be the power fraction allocated to the SSC. If we assume that the total available HSDPA pouter is equally divided
`between the primary scrambling code and the SSC, then it follows that the total HSDPA power fraction is 2y. The
`question we ask is as follows: If the total power fraction allocated to voice is 0.37 when no SSC is present, then what is
`the largest value of power y that can be assigned to the SSC such that voice quality is maintained despite the
`introduction of SSC?
`
`For our analysis we assume that 10 voice users are present in the system. Additionally, we assume that STTD is being
`used for voice, the receive Ec/Nt target for which is -21dB, and the power control is perfect. We measure voice quality
`in terms of the probability that the power fraction allocated to voice is sufficient to achieve this receive target for all
`the voice users as a function of the power fraction available allocated. We shall denote this probability by x. Figure 7
`plots x as a function of voice power fraction for different values ofy. For comparison, we have also plotted the case
`when no SSC is present. Note that in this case, when voice power fraction is 37%, x = 0.9875. Next, consider the case
`when y = 0.315. If, as mentioned earlier, the total available HSDPA power is equally divided between the primary
`scrambling code and the SSC, then it follows that the available voice power fraction is unchanged and equals 0.37.
`Observe in the figure that voice quality does suffer in this case, and x reduces to less than 0.985. Ify is reduced to 0.25
`(and, consequently, voice power fraction is increased to 0.5), then x is greater than 0.9875, thereby implying that voice
`quality is restored. Therefore, it follows that 25% power fraction can be allocated to the SSC for HSDPA without
`affecting voice quality. Note that this is in line with the assumption on the power fraction allocated for HSDPA in the
`results presented for code reuse option-2 in the section above.
`
`A more thorough analysis of the impact on voice due to the introduction of the SSC in the presence of real power
`control, as also other channel models (such as Ped-B) is for further study.
`
`0.9
`
`0.5
`
`0.4
`
`0.2
`
`0
`
`--SSC PF = 0
`
`--SSC PF = 0.15
`
`SSC PF = 0.2
`
`SSC PF = 0.25
`
`--SSC PF = 0.315
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`Voice power fraction
`
`Figure 6. Impact on voice due to the introduction of the SSC with perfect power control and STTD.
`
`NK8681TC013089787
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00007
`
`
`
`v
`
`1
`
`0.995
`
`0.99
`
`0.985
`
`-- No SSC
`
`y=O.2
`
`y = 0.25
`
`--y = 0.315
`
`0.98
`0.35 0.36 0.37 0.38 0.39 0.4 0.41 0.42 0.43 0.44 0.45
`
`Voice power fraction
`
`Figure 7. Figure 6 zoomed in.
`
`NK8681TC013089788
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00008
`
`
`
`Annex: Simulation parameters
`
`The system level simulation parameters are listed in Table 1 below.
`
`Parameter
`
`Cellular layout
`
`Table 1 Basic system level simulation assumptions.
`
`Explanation/Assumption
`
`Comments
`
`Hexagonal grid, 3-sector sites
`
`Provide your cell layout picture
`
`Site to Site distance
`
`2800 m
`
`Antenna pattern
`
`As proposed in [2]
`
`Only horizontal pattern specified
`
`Propagation model
`
`L 128.1 + 37.6Log10(R)
`
`R in kilometers
`
`CPICH power
`
`Other common channels
`
`-10 dB
`
`- 10 dB
`
`to HSDPA Variable
`Power allocated
`transmission, including associated
`signaling
`
`Slow fading
`
`As modeled in UMTS 30.03, B 1.4.1.4
`
`Std. deviation ef slow fading
`
`Correlation between sectors
`
`Correlation between sites
`
`8 dB
`
`1.0
`
`0.5
`
`Correlation distance of slow fading
`
`50 m
`
`Carrier frequency
`
`2000 MHz
`
`BS antenna gain
`
`UE antenna gain
`
`UE noise figure
`
`14 dB
`
`0 dBi
`
`9 dB
`
`Max. # of retransmissions
`
`Specify the value used
`
`Retransmissions by fast HARQ
`
`Fast HARQ scheme
`
`A21R
`
`BS total Tx power
`
`Up to 44 dBm
`
`Active set size
`
`Frame duration
`
`Scheduling
`
`3
`
`2.0 ms
`
`normalized Max C/I
`
`Specify Fast Fading model
`
`Jakes spectrum
`
`Maximum size
`
`Generated e.g. by Jakes or Filter
`approach
`
`The fundamentals of the data-traffic model are captured in Table 2 below.
`
`Table 2 Data-traffic model
`
`)arameters
`
`Process
`Packet Calls Size
`
`I
`
`Random Variable
`Pareto with cutoff
`
`I
`
`Time Between Packet Calls
`Packet Size
`
`Packets per Packet Call
`Packet Inter-arrival Time
`(open- loop)
`
`Geometric
`Segmented based on MTU
`size
`Deterministic
`Geometric
`
`Parameters
`A=I. 1, k=4.5 Kbytes, m=2 Mbytes, # = 25
`Kbytes
`,Lt = 5 seconds
`(e.g. 1500 octets)
`
`Based on Packet Call Size and Packet MTU
`# = MTU size/peak link speed
`(e.g. [1500 octets * 8]/2 Mb/s = 6 ms)
`
`NK8681TC013089789
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00009
`
`
`
`Packet Inter-arrival Time
`(closed-loop)
`
`Deterministic
`
`TCP/IP Slow Start
`(Fixed Network Delay of 1 O0 ms)
`
`10
`
`NK8681TC013089790
`ZTE Corporation and ZTE (USA) Inc.
`Exhibit 1018-00010