Resource Allocation Considerations for Multi-
`Carrier LTE-Advanced Systems Operating in
`Backward Compatible Mode
`
`Yuanye Wang*, Klaus I. Pedersen†, Preben E. Mogensen*†, and Troels B. Sørensen*
`*Aalborg University, †Nokia Siemens Networks – DK-9220 Aalborg East – Denmark
`Email: ywa@es.aau.dk
`
`Abstract— in this paper we focus on LTE-Advanced with
`backward compatibility, which serves a mixture of LTE-
`Advanced and LTE-Rel’8 users. Aggregation of multiple
`component carriers (CCs) to form a wide spectrum is assumed in
`order to fulfill the bandwidth requirement for the next generation
`systems, thereby leading to a multi-carrier system. Although a
`LTE-Advanced user can simultaneously access all the CCs, a
`LTE-Rel’8 user is restricted to operate on a single CC at a time.
`Different methods for balancing the load across these CCs will
`affect the system performance. This is investigated through both
`analytical methods and system level simulations. Bearing in mind
`that the LTE-Advanced users are scheduled on more CCs than
`the LTE-Rel’8 users, we propose a simple cross CC packet
`scheduling algorithm that improves the resource allocation
`fairness among the two categories of users. This simple scheduling
`algorithm is shown to be effective in providing better coverage
`performance with no loss of the overall cell throughput, as
`compared to independent scheduling per CC.
`
`INTRODUCTION
`I.
`The wireless communication systems have experienced
`dramatic growth since the introduction of the cellular concept
`in the 1960s [1]. They are now capable of providing users with
`high data rate and a variety of service types [2]. In continuation
`of this, the 3rd Generation Partnership Project (3GPP) has
`started a study item on Long Term Evolution (LTE)-Advanced
`[3], which aims at fulfilling the requirements for International
`Mobile Telecommunications - Advanced (IMT-Advanced) [4].
`For this purpose, a wider bandwidth than the 20MHz of the
`current LTE Rel’8 systems [5] will be used. This wide
`bandwidth can go up to 100MHz [6]. The current spectrum
`utilization pattern excludes the possibility of assigning a
`contiguous wideband; thereby it needs to be obtained via
`carrier aggregation (CA) of individual component carriers
`(CCs), where each CC follows the LTE Rel’8 numerology.
`This leads to multi-carrier LTE-Advanced. In order to allow
`backward compatibility so LTE Rel’8 and LTE-Advanced
`users can co-exist, it has been decided to use independent
`layer-1 transmission, which contains Link Adaptation (LA)
`and Hybrid Automatic Repeat request (HARQ) etc, per CC in
`coherence with the LTE Rel’8 assumptions [7].
`The migration from single to multi-carrier systems has
`previously been studied for High Speed Downlink Packet
`Access (HSDPA) [8], [9]. Two categories of users are
`specified in [8]: one is capable of simultaneously accessing all
`the CCs, whereas the other is restricted to operate on only one
`CC. Most of the studies in literature assume only one type of
`the users. However, in a backward compatible LTE-Advanced
`system, there will be a mixture of the two categories of users.
`
`Assuming that the objective is to have similar coverage and
`fairness for different user categories, it is necessary to update
`the resource allocation in upper layers, so that a LTE Rel’8
`user can get comparable resources with a LTE-Advanced user.
`
`Fig. 1. Structure of a multi-component carrier LTE-Advanced system.
`
`Fig. 1 illustrates the structure for a multi-component carrier
`LTE-Advanced system. The enhanced NodeB (eNB) first
`performs admission control to decide which users to serve, and
`then employs layer-3 CC Selection to allocate the users on
`different CCs. Different methods for balancing the load across
`these CCs will affect the system performance. Once the users
`are assigned onto certain CC(s), the layer-2 Packet Scheduling
`(PS) is performed. If independent PS per CC is used as
`pictured in Fig. 1, it cannot capture the difference between the
`number of CCs scheduled to the LTE-Advanced or Rel’8
`users. Therefore the Rel’8 users will suffer from operating over
`much fewer resources than the LTE-Advance users. In order to
`ensure the LTE-Rel’8 users get comparable resources as the
`LTE-Advanced users, cross CC PS is required. Finally, the
`layer-1 transmission is performed within each CC.
`In this paper we will first look into the problem of
`maximizing the system performance via layer-3 CC load
`balancing methods, then study the performance of independent
`and cross CC PS.
`The rest of this paper is organized as follows: Section II
`explains different techniques under investigation; Section III
`provides the theoretical analysis for the performance with
`different load balancing methods; In Section IV, the simulation
`methodology and assumptions are described; Section V shows
`both analytical and simulation results
`in a backward
`compatible system using different load balancing methods and
`PS algorithms; Section VI concludes the paper.
`
`370
`978-1-4244-5213-4/09/ $26.00 ©2009 IEEE
`Authorized licensed use limited to: Fish & Richardson PC. Downloaded on March 06,2021 at 10:12:22 UTC from IEEE Xplore. Restrictions apply.
`
`1
`
`APPLE 1009
`
`

`

`
`
`II. INVESTIGATED TECHNIQUES FOR THE BACKWARD
`COMPATIBLE LTE-ADVANCED SYSTEM
`In a multi-carrier LTE-Advanced system, the LTE-Advanced
`and Rel’8 users are treated differently, which is shown in Fig.
`2. A LTE-Advanced user will be assigned on all CCs in order
`to increase the user experienced performance and maximize
`the overall cell frequency domain packet scheduling (FDPS)
`gain [10]. A LTE-Rel’8 user supports transmission on only one
`CC, thereby the eNB needs to select a proper CC, and assign it
`to the user. Two methods for layer-3 CC load balancing are
`studied:
`
`
`Fig. 2. Different ways to treat the LTE-Advanced users and the LTE-Rel’8
`users for a LTE-Advanced system in backward compatible mode.
`
`Round Robin (RR) Balancing [11]
`The basic principle for RR balancing is to assign the newly
`arrived LTE Rel’8 user to the carrier that has the least number
`of users. Thus, it tries to distribute evenly the load to all
`carriers. However, there might be small variation for the cell
`load in different CCs, because the number of users does not
`necessarily make an exact even over the CCs or because one or
`more users may leave the system at random.
`Mobile Hashing (MH) Balancing [11]
`The MH balancing relies on the output from the user’s
`hashing algorithm. The output hash values are uniformly
`distributed among a finite set, which maps directly on the CC
`indices. Thereby, it provides balanced load across CCs in the
`long term. However, at each instant, the load across CCs is not
`balanced and the system will suffer from reduced trunking
`efficiency.
`
`After the CC assignment of the new user, PS at layer-2 is
`performed. This will decide on the resources one user can get
`within each CC. In our study, we select a commonly used
`frequency domain packet scheduler, namely Proportional Fair
`(PF) [12]. PF is aware of the channel condition for each user,
`and thereby can exploit the channel diversity to offer the FDPS
`gain compared to a channel blind scheduler. Two kinds of PS
`algorithms are investigated in this paper:
`Independent PS per CC
` This is similar to the PS in a traditional single carrier
`system, which does not
`consider
`the
`transmission
`characteristics from the other CCs. With the PF scheduler, the
`resource is assigned to a user that maximizes the metric on
`
` denoting the estimated throughput for user k on
`with
`jikR ,,
`~ the average delivered
`the ith CC at the jth PRB group and
`ikR ,
`throughput for that user on the same CC. In the long term, the
`PF scheduler achieves an equal share of resources within each
`CC among users with same fading statistics [14]. However,
`considering the difference in CC assignment between the LTE
`Rel’8 user and the LTE-Advanced users, a LTE-Advanced
`user will get N times the resources as a LTE Rel’8 user,
`where N is the total number of aggregated CCs in the system.
`Cross CC PS
`By taking into consideration the statistics from all CCs, the
`packet scheduler can achieve an overall better resource
`allocation than independent scheduling. In order to reduce the
`complexity for upgrading the existing LTE systems, we
`propose a scheduling algorithm that still operates within each
`CC. The only difference from independent scheduling is that it
`takes the past user throughput over all aggregated CCs into
`account, i.e.
`
`Metric
`
`jik
`,,
`
`=
`
`R
`N
`
`i
`
`1
`
`(cid:166)=
`
` (3)
`
`jik
`,,
`~
`R
`ik
`,
`
`the LTE-Advanced users have a reduced
`With (3),
`scheduling metric because their overall throughput is higher
`than the throughput per CC. On the other hand, the LTE Rel’8
`users maintain
`their scheduling metric, because
`their
`transmission and reception are restricted on only one CC. They
`are thereby prioritized as compared with the LTE-Advanced
`users in resource allocation, which meets the objective of
`similar fairness for all user categories. The possibility of
`further improving the single-CC user (Rel’8) throughput is
`investigated in details in [15], by putting different weights on
`the scheduling metric for the users.
`
`III. THEORETICAL ANALYSIS OF LAYER-3 CC LOAD
`BALANCING METHODS
`In this section, the performance for different load balancing
`methods is analyzed theoretically. In order to de-couple the
`transmission over multiple CCs, we use the simple approach of
`independent PS per CC. The performance with cross CC PS
`will later be evaluated based on system level simulations.
`Within our analysis, we assume a frequency domain PF
`packet scheduler, which offers FDPS gain in terms of increase
`average cell throughput that follows a logarithmic function of
`
`371
`Authorized licensed use limited to: Fish & Richardson PC. Downloaded on March 06,2021 at 10:12:22 UTC from IEEE Xplore. Restrictions apply.
`
`each CC [12]:
`
`k
`
`i
`
`,
`
`j
`
`=
`
`arg
`
`{max Metric
`
`
`k
`
`}
`
`
`
` ,,ik
`
`j
`
` (1)
`
` is the selected user on the ith CC at the jth Physical
`where
`jik ,
`Resource Blocks (PRB) group. According to [13], each PRB is
`constituted with 12 consecutive sub-carriers, and one PRB
`group contains 3 neighboring PRBs. The traditional way to
`calculate the PF metric is [12]:
`
`j
`
` (2)
`
`
`
`ik ,
`
`RR
`
`
`
`ik ,,
`
`~=
`
`Metric
`
`
`
`ik ,,
`
`j
`
`2
`
`

`

`The cell throughput in this case is ( N times the throughput
`per CC):
`
`(cid:166)− =
`
`TP
`cell
`
`(
`
`
`
` KMH,
`
`)
`
`a
`
`=
`
`C
`∞
`G
`(
`)
`k
`0
`From (5) and (11), the overall cell throughput can be
`estimated by:
`
`aKK
`
`
`KGkP )( (
`
`
`
`+
`
`k
`
`)
`
`a
`
` (11)
`
`(cid:176)(cid:175)(cid:176)(cid:174)(cid:173)
`
`
`
`
`
` )(kG
`
`=
`
`
`
`the number of users k on each CC [10]. For our modeling
`purposes, we represent the FDPS gain for a LTE system in
`[10] with the sample approximation as:
`1
`ln(*11.0
`38.1
`
`+
`10.1)
`
`k
`
`=
`k
`1
`≤<
`k
`1
`>
`k
`13
`
`13
`
` (4)
`
`Now let us consider a general case with N aggregated CCs
`and K active users per cell. If we assume that a user has
`probability α of being a LTE-Advanced user, then the
`aK LTE-Advanced users out of the K
`probability of having
`active users equals:
`
`KP
`(
`
`)
`
`a
`
`=
`
`(cid:168)(cid:168)
`
` −αα
`
`K
`)
`1(
`(cid:184)(cid:184)
`
`a
`
`−
`KK
`
`a
`
` (5)
`
`TP
`cell
`
`(
`
`MH
`
`)
`
`=
`
`C
`∞
`(
`
`)
`
`G
`
`K
`
`(cid:166)
`
`=
`
`a
`
`KP(
`
`)
`
`a
`
`−
`KK
`
`=
`
`(cid:166)
`
`a
`
`
`
`KGkP )( (
`
`
`
`+
`
`k
`
`)
`
`a
`
` (12)
`
`
`K
`k
`0
`0
`Although the average user throughput on each CC can be
`estimated using a similar form as with RR, the overall user
`throughput is dependent on the load condition on all CCs. Due
`to this correlation, it is non-trivial to formulate the average
`user throughput and we will evaluate its performance based on
`simulation.
` equals 1 for all k
`For both equation (9) and (12), if
`)(kG
`(other than 0), we obtain the performance with frequency
`domain channel blind RR packet scheduler.
`
`IV. SIMULATION METHODOLOGY AND ASSUMPTIONS
`The performance of the algorithms is evaluated in a quasi static
`downlink multi-cell system level simulator that follows the
`LTE specifications defined
`in [16],
`including detailed
`implementations of layer-3 CC selection, layer-2 PS, HARQ
`and LA functionalities. The simulation scenario is Macro-cell
`case #1 as defined in [7]. The simulation parameters are
`summarized in Table I. The link to system mapping is based
`on the exponential effective metric model [17].
`Note that, we aggregate 4 CCs, each of 10MHz to form a
`wide bandwidth of 40MHz. In case an even wider bandwidth is
`needed, more CCs can be aggregated together, or the
`bandwidth per CC can be extended. Simulation campaigns are
`conducted with 40 simulation runs (5.0 seconds in each run)
`with constant number of 10 users per cell. Multiple simulation
`runs are required for this traffic model in order to get sufficient
`statistics, since the traffic model is static in the sense that the
`10 users per cell are active all the time.
`Three kinds of throughput measures are used in our study as
`performance indicators:
`• Cell throughput: Average throughput per cell, i.e. equals
`the summation of the user throughput in each cell.
`• LTE-Advanced (or Rel’8) user throughput: Average
`throughput over all the simulated LTE-Advanced (or
`Rel’8) users.
`• Coverage: The 5th percentile worst user throughput, over
`the simulated users.
`
`
`a
`K
`0
`the
`is
`throughput one each CC
`The average user
`corresponding per CC cell throughput divided by the average
`number of users. Because the LTE-Advanced users are
`scheduled on N CCs, their throughput is expected to be N
`times that of the LTE-Rel’8 users. As a result, we have the
`following equation for the user throughput:
`TP
`RR
`(
`)
`cell
`(~
`K
`KPK
`0
`TP
`RR
`(
`)
`cell
`
`α −+
`α
`N
`1(
`/)
` if the user is LTE-Advanced, and
`
`K
`
`(cid:166)=
`
`a
`
`)
`
`a
`
` (9)
`
`TP
`user
`
`(
`
`RR
`
`)
`
`=
`
`I
`
`N
`
`
`
`(cid:166)=
`
`a
`
`K
`I
`
`KN
`
`=
`
` for
`
`1−=I
`
`0=I
`In (9),
`Rel’8 users.
`B. Throughput analysis with MH Balancing
`If MH is used for carrier load balancing, each LTE Rel’8
`user has equal probability of selecting any of the CCs, which is
`. It offers balanced load in the long term and the
`N/1
`throughput is N times the performance with only one CC. With
`aK LTE-Advanced users, the probability for one CC to have
`k LTE Rel’8 users is:
`
`−
`−
`kKK
`a
`
` (10)
`
`(cid:184)(cid:185)(cid:183)
`
`1
`
`−
`N
`N
`
`(cid:168)(cid:169)(cid:167)
`
`k
`
`(cid:184)(cid:185)(cid:183)
`
`1
`N
`
`(cid:168)(cid:169)(cid:167)
`(cid:185)(cid:183)
`
`(cid:184)(cid:184)
`
`−
`KK
`k
`
`a
`
`(cid:169)(cid:167)
`
`(cid:168)(cid:168)
`
`kP
`)(
`
`=
`
`372
`Authorized licensed use limited to: Fish & Richardson PC. Downloaded on March 06,2021 at 10:12:22 UTC from IEEE Xplore. Restrictions apply.
`
`(cid:185)(cid:183)
`
`KK
`
`(cid:169)(cid:167)
`
`a
`A. Throughput analysis with RR Balancing
`If the RR carrier balancing method is used, the number of
`users on each CC equals:
`~
`=
`KK
`
`(
`
`+
`
`−
`KK
`
`a
`The average cell throughput with
`is:
`
`a
`
` (6)
`
`
`) N/
`aK LTE-Advanced users
`
`=
`
`∞
`
` (7)
`
` /)~( GKCG
`
`TP
`, KRR
`
`(
`)
`(
`)
`cell
`a
`where C is the maximum achievable throughput over all
`CCs.
`Combining (5), (6) and (7), we get the performance with α
`probability LTE-Advanced users:
`C
`∞
`G
`(
`
`TP
`cell
`
`(
`
`RR
`
`)
`
`=
`
`)
`
`
`
` KGKP( ) )~(
`
`
`
` (8)
`
`3
`
`

`

`
`
`Carrier frequency
`Aggregation configuration
`Number of PRBs per CC
`Sub-frame duration
`
`Modulation and coding schemes
`
`TABLE I
`SYSTEM SIMULATION SETTINGS
`Parameter
`Setting / description
`Test scenario
`3GPP Macro-cell case #1 (19 sites,
`3 cells per site)
`2 GHz
`4 CCs, with 10MHz per CC
`50 (12 subcarriers per PRB)
`1 ms (11 OFDM data symbols plus
`3 control symbols )
`QPSK (1/5 to 3/4)
`16-QAM (2/5 to 5/6)
`64-QAM (3/5 to 9/10)
`2-Rx Interference Rejection
`Combining
`Ideal chase combining
`4
`6 ms
`1 CQI per 3 PRBs
`Log normal with 1dB std.
`2 dB
`None (the eNB is able to schedule
`up to 20 users simultaneously)
`Proportional fair
`10%
`10
`Full buffer
`
`User receiver
`
`HARQ modeling
`Max. number of retransmissions
`Ack/nack & CQI feedback delay
`CQI frequency domain resolution
`CQI reporting error
`CQI reporting resolution
`Time domain PS
`
`Frequency domain PS
`1st transmission BLER target
`Number of UEs per cell
`Traffic type
`
`
`2. With the co-existence of both LTE-Advanced and LTE
`Rel’8 users, the Rel’8 users achieve much lower
`throughput than the LTE-Advanced user. Because the
`outage performance is collected from the 5% worst user
`throughput, a low LTE-Rel’8 user throughput therefore
`leads to poor coverage performance. MH offers lower
`throughput for the LTE-Rel’8 users than RR, which will
`leads to even poorer coverage as compared with RR.
`
`
`Fig. 3. Gain in average cell throughput by using RR balancing as compared to
`MH. Performance is evaluated with 10 users per cell and different ratios of
`LTE-Advanced users. Both estimated and simulated results are shown.
`
`
`
`V. PERFORMANCE WITH DIFFERENT LOAD BALANCING
`METHODS AND INDEPENDENT / CROSS CC SCHEDULING
`In this section, the performance for different carrier load
`balancing methods is evaluated using the theoretical analysis
`and
`the simulation methodology described previously.
`Different ways to perform the PS at layer-2 are also
`investigated.
`A. Load balancing with independent PS per CC
`As discussed before, the RR load balancing method achieves
`a balanced load across the multiple CCs in a short duration as
`compared with MH. In term of throughput, RR is also
`expected to be higher than MH. The relative gain in average
`cell throughput by using RR over MH is shown in Fig. 3, with
`different ratios of LTE-Advanced users. Both theoretical
`estimation and simulation result are shown. The absolute
`values of average user (both LTE-Advanced and LTE-Rel’8)
`throughput are summarized in Fig. 4.
`From these two figures, we have the following observations:
`1. In terms of average cell throughput, there is a good match
`between the theoretical estimates and the extensive
`system-level simulations, with maximum 1% deviation
`between the two. When all users are Rel’8, RR
`balancing provides ~7% higher cell throughput than
`MH. However, the gain decreases quickly with the ratio
`of LTE-Advanced user and it vanishes for the ratio
`beyond 20%. The reason is that the assignment of CCs
`to the LTE-Advanced users is always balanced. A high
`ratio of LTE-Advanced users thereby means a better
`balancing than with low ratio, hence less room for RR
`to improve over MH.
`
`
`Fig. 4. Average user throughput for the two carrier load balancing methods.
`Results are obtained via simulation with 10 users per cell and different ratios
`of LTE-Advanced users.
`
`B. Performance with independent or cross CC PS
`We have seen that the load balancing method of RR offers
`better performance than MH. However, the Rel’8 users still
`suffer from much lower performance than LTE-Advanced
`users, which will cause poor coverage. By using cross CC PS
`as introduced in Section II, we can increase the priority for the
`Rel’8 users to be scheduled. Thereby cross CC PS is expected
`to offer better coverage performance than with independent PS
`per CC.
`The performance with independent and cross CC PS is
`shown in Fig. 5 and Fig. 6 for average cell throughput and
`coverage, respectively. We are mostly interested in the
`performance with the load balancing method of RR. However,
`the performance for MH with independent PS is also presented
`for reference.
`From Fig. 5 we can see that, there is no obvious gain, or
`loss, by using cross CC PS over independent PS. However, in
`
`373
`Authorized licensed use limited to: Fish & Richardson PC. Downloaded on March 06,2021 at 10:12:22 UTC from IEEE Xplore. Restrictions apply.
`
`4
`
`

`

`
`
`terms of coverage performance, cross CC PS has a significant
`improvement over independent PS. When LTE-Rel’8 users are
`mixed together with the LTE-Advanced users, the gain is
`50~90% over independent scheduling, as shown in Fig. 6.
`
`Fig. 5. Average cell throughput for independent or cross CC PS.
`
`48%
`
`93%
`
`80%
`
`
`
`
`
`Fig. 6. Coverage performance for independent or cross CC PS.
`
`VI. CONCLUSION
`In this paper, we have studied the problem of how to
`accommodate the Rel’8 users in a multi-carrier LTE-Advanced
`system. Due to the terminal capability constraints, Rel’8 users
`can access only one carrier. They are
`thereby
`in a
`disadvantageous situation as compared to LTE-Advanced
`users, who can simultaneously access all carriers.
`Both analytical and simulation results are obtained for two
`load balancing methods and different component carrier (CC)
`scheduling. The results show that with low number of users
`and
`low percentage of LTE-Advanced users,
`the
`load
`balancing method of RR achieves better performance than the
`MH balancing. However, the LTE Rel’8 users still suffer from
`lower throughput than the LTE-Advanced users, because of the
`independent scheduling per CC.
`Motivated by the poor coverage performance, we proposed a
`cross CC PS algorithm, which is a simple extension of the
`existing PF scheduler. The cross CC algorithm is aware of the
`user throughput over all the aggregated CCs, it improves the
`scheduling priority for the Rel’8 users on its serving CC. As a
`result, this algorithm is able to significantly improve the
`performance of the LTE Rel’8 users. In terms of coverage, it
`offers a gain of 50~90% when LTE Rel’8 users and LTE-
`Advanced users co-exist in the same system. Despite of the
`
`high gain in coverage, it gives no degradation in the average
`cell throughput, and thereby provides a good solution for the
`LTE-Advanced systems to work in backward compatible
`mode.
`As continuation of this study, we suggest to conduct further
`evaluations under more realistic traffic models, as well as
`including various updates coming from 3GPP as more
`decisions on LTE-Advanced are coming.
`
`ACKNOWLEDGMENT
`The authors are grateful to Daniela Laselva, Jens Steiner and
`Mads Brix of Nokia-Siemens-Networks for their valuable
`suggestions and help in carrying out this study.
`
`REFERENCES
`[1] S. Z. Asif, “Wireless Communications Evolution to 3G and Beyond,”
`Artech House, Inc. 2007.
`[2] E. Dahlman, S. Parkvall, J. Sköld, and P. Beming, “3G Evolution: HSPA
`and LTE for Mobile Broadband,” Elsevier, 2007, ch. 2.
`[3] S. Parkvall, E. Dahlman, A. Furuskar, Y. Jading, M. Olsson, S.
`Wanstedt, and K. Zangi, “LTE-Advanced – Evolving LTE towards IMT-
`Ad¬vanced,” in Proc. IEEE VTC, pp. 1-5, Sept. 2008.
`[4] Recommendation ITU-R M.1645, “Framework and overall objectives of
`the future development of IMT 2000 and systems beyond IMT 2000,”
`June, 2003.
`[5] H. Holma and A. Toskala, “LTE for UMTS, OFDMA and SC-FDMA
`Based Radio Access,” John Wiley & Sons, 2009.
`[6] 3GPP TR 36.913 v8.0.0, “Requirements for further advancements for E-
`UTRA (LTE-Advanced),” June, 2008.
`[7] 3GPP TR 36.814 v1.0.0, “Evolved Universal Terrestrial Radio Access
`(E-UTRA); Further advancements for E-UTRA Physical layer aspects,”
`Feb. 2009.
`[8] 3GPP TR 25.825 v1.0.0, “Dual-cell High Speed Downlink Packet Access
`(HSDPA) operations,” June, 2008.
`[9] K. Johansson, J. Bergman, D. Gerstenberger, M. Blomgren, and A.
`Wallén, “Multi-Carrier HSPA Evolution,” in Proc. IEEE VTC, Apr.
`2009.
`[10] A. Pokhariyal, T. E. Kolding, and P. E. Mogensen, “Performance of
`downlink frequency domain packet scheduling for the UTRAN long term
`evolution,” in Proc. IEEE PIMRC, pp. 1-5, Sept. 2006.
`[11] T. W. Wong, and V. K. Prabhu, “Capacity Growth for CDMA System:
`Multiple Sectors and Multiple Carrier Deployment,” in Proc. IEEE VTC,
`vol. 2, pp. 1608-1611, May, 1998.
`[12] A. Jalali, R. Padovani, and R. Pankaj, “Data Throughput of CDMA-HDR
`a High Efficiency-High Data Rate Personal Communication Wireless
`System,” in Proc. IEEE VTC, pp. 1854-1858, 2000.
`[13] 3GPP TS 25.814 v7.1.0, “Physical layer aspects for evolved universal
`terrestrial radio access,” Sept. 2006.
`[14] J. M. Holtzman, “Asymptotic Analysis of Proportional Fair Algorithm,”
`in Proc. IEEE PIMRC, vol. 2, pp. F-33-F-37, Oct. 2001.
`[15] Y. Wang, K. I. Pedersen, M. Navarro, P. E. Mogensen, and T. B.
`Sørensen, “Uplink Overhead Analysis and Outage Protection for Multi-
`Carrier LTE-Advanced Systems,” in Proc. IEEE PIMRC, Sept. 2009.
`[16] 3GPP TS 36.300 v8.0.0, “Evolved Universal Terrestrial Radio Access
`(E-UTRA) and Evolved Universal Terrestrial Radio Access Network (E-
`UTRAN); Overall description,” Mar. 2007.
`[17] K. Brueninghaus, D. Astely, T. Salzer, et al. “Link performance models
`for system level simulations of broadband radio access systems,” in Proc.
`IEEE PIMRC, vol. 4, pp.2306 - 2311, Sept. 2005.
`
`374
`Authorized licensed use limited to: Fish & Richardson PC. Downloaded on March 06,2021 at 10:12:22 UTC from IEEE Xplore. Restrictions apply.
`
`5
`
`