LTE Capacity Enablers

15 downloads 346 Views 130KB Size Report
In order to achieve this, several new radio transmission technologies are ..... India Electronics limited and Qualcomm Inc. on 2G and 3G wireless technologies.
LTE Capacity Enablers Ashvin Chheda, Sagar Dhakal, Qi Hao, Gaurav Hemrajani, JoonBeom Kim, Sairamesh Nammi, Shankar Venkatraman Acknowledgements: Xiao-Dong Li, Patrick Lie

Abstract Long term evolution (LTE) of the 3GPP radio-access technology has been engineered to satisfy the increasing demand for broadband wireless communications. LTE systems are projected to have higher peak data rates, reduced latency, improved spectrum efficiency, greater system capacity, larger coverage, with reasonable system and terminal complexity. In order to achieve this, several new radio transmission technologies are included. Given the success of orthogonal frequency division multiplexing (OFDM), initially for broadcasting applications such as DAB, DVB-T followed by W-LAN (IEEE 802.11a/g), OFDM is a solid physical layer backbone for LTE systems. Multiple antennas at the transmitter and receiver using multiple input multiple output (MIMO) techniques have been designed. By using multiple parallel data streams transmission to a single terminal, data rate can be increased significantly. In this paper, we discuss various capacity enablers, and quantify the benefits provided over existing 3G technologies, such as 1xEV-DO.

1. Introduction 3GPP LTE represents the project within the third generation partnership project, with an aim to improve the UMTS standard. The goals are to support future requirements, and include; improved system capacity and coverage, reduced latency, higher peak data rates, and lowering costs. The LTE project is not actually a standard, but results in an evolved release of the UMTS standard. Mobility across cellular networks is supported. Bandwidth is scalable in view of spectrum allocations, higher data rate requirements and deployment flexibility. The LTE physical layer is designed to achieve higher data rates, and is facilitated by turbo coding/decoding, and higher order modulations (up to 64-QAM). The modulation and coding is adaptive, and depends on channel conditions. LTE supports both FDD and TDD modes of operation. This paper focuses on FDD. Orthogonal frequency division multiple access (OFDMA) is used for the downlink, while Single carrier frequency division multiple access (SC-FDMA) is used for the uplink. The main advantage of such schemes is that the channel response is flat over a sub-carrier even though the multi-path environment could be frequency selective over the entire bandwidth. This reduces the complexity involved in equalization, as simple single tap frequency domain equalizers can be used at the receiver. OFDMA allows LTE to achieve its goal of higher data rates, reduced latency and improved capacity/coverage, with reduced costs to the operator. The LTE physical layer supports HARQ, power weighting of physical resources, uplink power control, and MIMO. By using multiple parallel data streams transmission to a single terminal, data rate can be increased significantly. In a multi-path environment, such a multiple access scheme also provides opportunities for performance enhancing scheduling strategies. Frequency Selective Scheduling (FSS) can now be used to schedule a user over Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

1

sub-carriers (or part of the bandwidth) that provides maximum channel gains to that user (and avoid regions of low channel gain). The channel response is measured and the scheduler utilizes this information to intelligently assign resources to users over parts of the bandwidth that maximize their signal-to-noise ratios (and spectral efficiency). In other words, the end to end performance of a multi-carrier system like LTE relies significantly on sub-carrier allocation techniques and transmission modes. LTE allows for different opportunistic scheduling techniques; a source of significant product differentiation between competing companies. In this paper we compute the expected throughput of an LTE system under the different capacity enabling feature sets and compare and contrast to current 3G systems such as 1x EV-DO for both uplink and downlink. We show the full buffer system throughput for LTE under different scheduling techniques and MIMO techniques. We compute the expected system capacity of the LTE system for delay-sensitive applications such as VoIP. In terms of a reference point these are compared to the capacity and throughput of a 1xEV-DO system. We also extend our discussion to future work currently planned for.

2. Simulation Setup LTE [and 1xEV-DO] performance is studied via simulations. The LTE simulation set up is based on the 3GPP simulation evaluation methodology guidelines. Two layers of simulation are used. The first is a link level or physical layer simulation that models the frame transmission and reception over the wireless fading channel. The goal of the physical layer simulator is to provide physical layer data, such as packet loss probability per H-ARQ attempt for the different MCS schemes. This is used by the second layer of simulations, the system level simulator. The goal of this simulation is to model the cellular network, both uplink and downlink. A homogenous network of tri-sectored cells is modeled; antenna patterns, shadowing models, path loss models are included per 3GPP guidelines. The output of the simulator is sector throughput, or user capacity, based on a uniform distribution of terminals. The entire protocol stack, from the application layer, down to the physical layer is implemented. The physical layer implementation in the second simulator is based on data from the first simulator. The system level simulations used to evaluate performance of the LTE network use the following path loss model and fast fading channel mix. The traffic model used to evaluate the performance gains of various features assume the resource usage on each base station to be used 100%. Parameter Number of cells (3 sectored) BS-to-BS distance Minimum eNB and UE distance BS Transmission Power Shadowing Base Station Shadow Correlation Path-loss Antenna Pattern Thermal noise density eNB noise figure eNB maximum SINR eNB antenna gain UE antenna gain Max UE Tx Power UE other loss (Penetration) Channel Estimation

Value 19 500m 35m 46 dBm 8 dB 0.5 128.1+37.6*log10(d), d is in unit of km 70 degree (-3dB) with 20 dB front-to-back ratio -174 dBm/Hz 5 dB 30 dB 14 dB 0 dB 23 dBm (200 mW) 10 dB Ideal

Table 1. Simulation parameter set

Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

2

Mixed channels with the following probabilities were used; 40% Channel A (3 Km/h), 30% Channel B (10 Km/h), 20% Channel C (30 Km/h) and 10% Channel D (120 Km/h). The multipath energy and delay spread are shown in Table 2 below. All results presented (unless otherwise stated) assume a frequency reuse of 1. Path Rel. Power (dB)

1

2

3

4

5

6

-3

0

-2

-6

-8

-10

rms Delay (ns)

0

200

500

1600

2300

5000

Table 2: Multi-path Delay Profile 3. Downlink Capacity Enablers System capacity is significantly increased using various multiple antenna techniques such as receive/transmit beam-forming, closed and open loop single user MIMO, and multi-user MIMO. Since the sub-carriers remain orthogonal during an OFDM symbol time, MIMO receiver structure becomes simple. In Fig 1(left), the instantaneous SINR distribution is shown for SIMO, SFBC (space frequency block code) transmit diversity, SCW (single code word) based SM (spatial-multiplexing) and adaptive MIMO schemes. Clearly, due to the inter-layer interference, the SM has lower per layer SINR values. However, there will be a 3 dB multiplexing gain due to simultaneous transmission in two different layers. It can also be observed that due to transmit diversity, SFBC has a better SINR distribution compared to the SIMO scheme. The SINR CDF directly correlates to user throughput distribution shown in Fig. 1(right). The dynamic range of SM spans over larger as well as smaller user throughput values compared to SIMO and SFBC schemes. From an information theoretic point of view, we know that after a certain SINR threshold, capacity starts to saturate (logarithmically) and the throughput gain ceases to be proportional to SINR. However, by transmitting an independent stream of data on the other antenna SM can potentially exploits the 3 dB multiplexing gain, whereby it can achieve higher throughput than other schemes. However, if the resulting per layer SINR is below the saturation threshold, SM scheme becomes very sensitive to error events, resulting in throughput degradation. Clearly, a better strategy is to use the SFBC diversity technique if the per layer SINR of SM is below the saturation threshold, while use the SM technique if it crosses the threshold value. In Fig. 1(left), we have shown the performance of one such adaptive scheme, namely SM-SFBC, where we can see that the CDF tracks the SFBC CDF at low SINR values and then becomes parallel to SM CDF at about 0 dB SINR. As evident from the user throughput CDF in Fig. 1(right), the SM-SFBC scheme, even with a smaller dynamic range, achieves as high a spectral efficiency as the SM scheme. The average sector throughputs (in Mbps) for SIMO, SFBC, SM and SM-SFBC were found to be 11.57, 12.23, 11.11 and 12.76, respectively.

Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

3

1

1

0.9

0.9 0.8

SFBC SIMO SM SM-SFBC

0.7 0.6

Cumulative Probability

Cumulative Probability

0.8

0.5 0.4 0.3

0.7 0.6 0.5 0.4 0.3

0.2

0.2

0.1

0.1

0 -30

-20

-10

10

0

20

30

SFBC SIMO SM-SFBC SM

0

0

500

1500

1000

Instantaneous SINR (dB)

2000

2500

3000

3500

4000

UE Throughput (Kbps)

Figure1. SINR and user throughput CDF

AVG. SECTOR THROUGHPUT (Mbps)

The end to end performance of a multi-carrier system like LTE relies significantly on sub-carrier allocation techniques and transmission modes. In a frequency-selective fading environment, a good allocation technique is to assign sub-carriers selectively from stronger portion of the spectrum. Clearly, the effectiveness of a selective scheduling algorithm is based on the accuracy of channel frequency response information. In particular, when channel coherence time is larger than the channel update interval, the frequency selective scheduling (FSS) scheme provides good performance gain. But in the case of high-speed users, frequency selective scheduling is expected to become ineffective as the channel varies quite significantly within the update period. In Fig.2, we have shown average sector throughputs based on two types of transmission schemes, namely FSS and frequency distributed scheduling (FDS). In the FSS scheme, UEs are allocated resource blocks (RBs: localized set of tones) from sub-bands with stronger channel power. In particular, RBs from a sub-band can be assigned to a UE only if the power of the sub-band is above a certain configurable threshold. On the other hand, in FDS scheme, UEs are assigned a set of RBs in order to span the entire spectrum, irrespectively of the actual sub-band power. As expected, FSS scheme is better for 3 and 10 kmph users while the reverse is true for 30 and 120 Kmph users. Motivated by this result, we compared the performance of three transmission schemes: FSS, FDS and FSS-FDS, where FSS-FDS means that 3 and 10 Kmph UEs use FSS scheme while 30 and 120 Kmph UEs use FDS scheme. The sector throughputs (in Mbps) were found to be 11.26, 10.8 and 10.2 for FSS-FDS, FSS and FDS, respectively 13 FSS FDS

12 11 10 9 8 3Kmph

30Kmph 10Kmph CHANNEL MODEL

120Kmph

Figure 2. FSS/FDS sector throughputs for different channel models

Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

4

Low data rate users who are typically assigned one RB will observe drastically reduced frequency diversity. In such cases, in order to improve frequency diversity, a DVRB is constructed such that the first seven OFDM symbols and the last seven OFDM symbols are from uncorrelated sub-bands. For high speed users, we observed that the distributed virtual resource block (DVRB) based scheduling provides significant performance gain. In particular, for120 Km/h users the sector throughput improved from 8.4 Mbps (FSS) to 9.7 Mbps (DVRB). Though SM enhances the data rate by transmitting independent data streams on different branches simultaneously it is sensitive to ill-conditioning of the channel matrix. One way to guard against the rank deficiencies in the channel is to pre multiplying the transmitted data stream by a precoding matrix. The precoding matrix is chosen based on the channel information. The full gain from precoding are achieved with full channel state information (CSI) at the transmitter, since this allows the transmitted signal to be customized based on the eigen structure of the channel matrix. In a frequency division duplex (FDD) system, full CSI must be conveyed through a feedback channel and it is highly impractical. One approach to achieve performance gains close to that of precoding is to use limited feedback. In a limited feedback a set of matrices which are known both at the transmitter and receiver. The whole set of matrices are called code book. For a downlink LTE system the UE will choose the precoding matrix based on some criterion and inform the ENB the precoding matrix index (PMI) . In our simulations, we choose the precoding matrix index which maximizes the largest AWGN capacity. The average sector throughput for closed loop MIMO was found to be 11.82 Mbps. Space division multiple access (SDMA) allows us to reuse limited bandwidth by multiplexing multiple users in the same time slot, thus multiplying the throughput of a wireless network resources. If the number of transmit antenna is , then spatial layers can be serviced. These spatial layers can be used either for single-user multipleinput multiple-output (MIMO) transmission or for SDMA with multiple-users. Specifically, this wideband SDMA can be implemented by beamforming or beam pattern that does not rely on beamforming. This paper shows the system performance of fixed beam SDMA, which is cost-effective deployment of advanced antenna technology. To determine a serving beam for an UE with , the UE calculates received power at each receive antenna. Then, if the difference is larger than threshold , the UE chooses the beam with strong receive power, otherwise this UE is served by both beams, called overlapping region. Since CL-MIMO is not adopted in SDMA configurations, the serving beam for an UE is determined by long-term SINR instead of using short-term SINR. SFBC and SIMO are used for an UE for both beam and single-beam region, respectively. Figure 1 shows the antenna beam patterns for 2-beam SDMA configuration. 0 Beam 1

-5

Beam 2 -10 -15 -20 -25 -30 -35 -40 -180 -150 -120 -90 -60

-30

0

30

60

90

120 150 180

Figure 3: Antenna Beam Pattern of 2-Beam SDMA

Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

5

LTE Spectral Efficiency (b/s/Hz) SDMA*

DO Spectral Efficiency (SIMO) (b/s/Hz)

SIMO

SFBC

SM

SMSFBC

CLMIMO

( Pth = 9 dB )

0.96

1.16

1.22

1.11

1.28

1.18

2.47

Table 3: Downlink Spectral Efficiency Comparison between LTE and DO-RevA

Pth (dB)

6 9 12

No SDMA (SIMO) Spectral Efficiency (b/s/Hz) 1.14

2-Beam SDMA (Mixed channels) Spectral Efficiency (b/s/Hz) 2.42 2.28 2.23

Success Rate (%)

Gain (%)

Avg. # of UEs

% scheduling for OL

99.94 99.94 99.90

113.19 100.09 95.69

5.19 6.13 8.33

26.97 29.23 38.18

Table 4: Throughput Comparison for different Pth with 30 UEs

As illustrated in 2, the average sector throughput of 2-beam SDMA configuration is 1.96 – 2.13 times larger than that of configuration without SDMA configuration depending on the threshold, Pth . For Pth = 6 dB, we can achieve up to 113% throughput gain under mixed channel environments. The percentage of scheduling UEs at an overlapping region can be changed depending on the number of UEs at an overlapping region due to the fairness of the professional fair scheduler. Thus, the average sector throughput is the function of this percentage. As the percentage increases, the sector throughput tends to be decreased. Furthermore, as mentioned before, if the CLMIMO is implemented based on short-term SINR, then the gain from SIMO will be further increased.

4. Uplink Capacity Enablers Uplink (virtual) V-MIMO is based on the same principle as MIMO using Spatial Multiplexing (SM) wherein multiple streams of information are transmitted over the same resource (frequency/time) to increase spectral efficiency. Multiple mobile stations, each using a single transmit antenna can be assigned the same resource to create a virtual MIMO transmitter. The main advantage of V-MIMO over conventional MIMO is that each mobile only needs to support a single transmit antenna. Further, V-MIMO as a feature is completely transparent to the mobile and requires no additional mobile processing. Extracting performance gains using V-MIMO in a multi-cell environment is more challenging when compared to a single user MIMO scenario. The scheduler has to select users capable of sustaining a robust link in the presence of additional interlayer interference & determine appropriate MCS downgrades. Scheduling of multiple mobiles on the same sub-carrier could result in additional inter-cell interference that will negatively affect low SINR users & coverage. The scheduling of users and user pairing are areas that need optimization to ensure that performance gains are realized without impacting coverage. SNR based Candidate List Filtering coupled with Orthogonal Factor based Pairing can be used to extract gains without affecting coverage. Extracting performance gains using VMIMO requires additional filtering schemes to be implemented to select Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

6

users that are capable of VMIMO. Firstly, mobiles at low SNR cannot be considered for VMIMO because the additional interlayer interference could lead to the user not being able to support even the most robust MCS at the required error rate. Further, if the difference in received signal between users is significant, that could also degrade the performance of the lower SNR user. Filters for minimum SNR and SNR difference along with appropriate MCS downgrade will ensure that the VMIMO paired users can meet required error rates. In addition, interference increase due to pairing needs to be mitigated. One approach is to only consider users that contribute least to the overall inter-cell interference. This leads to creating a VMIMO Candidate List based on forward link geometry. The approach allows the system to extract maximum spectral efficiency gains without impacting cell-edge performance. As an enhancement to randomly selecting users from the Candidate List for pairing, Orthogonal Factor (OF) based pairing makes use of channel information available at the base station for pairing users. If the channel between users is highly orthogonal, the amount of inter-layer interference generated by the paired users to each other will be minimal. OF based scheduling makes use of this principle in pairing users to reduce inter-layer interference and increase MIMO gains. Since OF based scheduling relies on channel information, it works best for slow fading channel models. OF based pairing used with Candidate List Filtering has shown 10% improvement in aggregate sector throughput without impacting coverage. Spectral Efficiency LTE η LTE (bps/Hz) 0.62 0.83 0.87

WBS FSS FSS & V-MIMO

Spectral Efficiency 1xEVDO-RevA η DO-RevA (bps/Hz)

%Gain (over WBS) 33% 40%

0.32

η LTE / η DO-RevA 1.93 2.59 2.71

Table 5. Uplink Spectral Efficiency Comparison between LTE and DO-RevA

Assumptions: LTE simulations assume 32 users per sector, DO simulations assume 15 users per sector.

User Throughput CDF 100%

Probability

80%

60% No FSS SIMO

40%

FSS SIMO FSS VIMO

20% 0% 0

100

200

300

400

500

600

700

Throughput (kbps)

Figure 4. FSS/FDS sector throughputs for different channel models

Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

7

5. VoIP Capacity The 1xEV-DO Rev-A & LTE VoIP capacity numbers for the three codecs, AMR12.2, AMR7.95 & EVRC are given in the table below. The capacity in 1xEV-DO was derived with 12-slot HARQ termination target & 120 ms delay bound, while the LTE capacity numbers in the table above are with a delay bound of 50 ms. Capacity is defined such that no more than 2% of packets exceed the delay bound. LTE VoIP capacity is uplink limited because of the inter-cell interference from the UEs using the same tones in other cells. Codec

Capacity

Average Delay per Packet

98% Delay per Packet

LTE-AMR-12.2

190 (5 MHz)

2.8 ms

18 ms

LTE-AMR-7.95

230 (5 MHz)

2.5 ms

17 ms

LTE-EVRC (1/8 On)

324 (5 MHz)

2.2 ms

17 ms

EVDO-EVRC (1/8 On)

42 (1.25 MHz)

32.6 ms

60 ms

Table 6: VoIP capacity of LTE vs 1xEV-DO Rev A. The capacity of EVRC in LTE is larger than AMR 7.95 codec because the average packet size of EVRC is 7 bytes (without headers) which is less than the average packet size of 10 bytes for AMR 7.95. As can be seen, VoIP capacity using EVRC in LTE is greater than that of 1xEV-DO Rev-A, when adjusted for bandwidth. In 1xEV-DO, beyond 42 simultaneous VoIP UEs, the RoT outage exceeds 7 dB more than 1% of the time, which is an additional criterion for capacity in addition to the delay bound 2% factor. It should be noted that the LTE capacity numbers in the table above are with a delay bound of 50 ms, whereas the 1xEV-DO Rev-A capacity is using a delay bound of 120 ms. Further, the size of the headers used in the LTE simulations for EVRC is 10 bytes whereas 7 bytes was used for EVRC simulation in 1xEV-DO. In spite of a tighter delay bound & relatively larger EVRC VoIP packet size, the 98-percentile delay experienced per packet in LTE is far lower than 1xEV-DO Rev-A at their respective capacity points. This is due to the fact that LTE Uplink overall provides at least 2-times greater spectral efficiency than 1xEV-DO Rev-A Reverse link. Also LTE Uplink gets some gain from pooling the 600 tones to form a 10 MHz bandwidth compared to the 1.25 MHz bandwidth of 1xEV-DO Rev-A. Pooling multiple 1xEV-DO Rev-A carriers with MCTA-like algorithm will increase the capacity for 1xEV-DO Rev-A beyond 42 per 1.25 MHz band as well. With a random allocation of users to three carriers (5 MHz DO deployment), the capacity would be 126. A smart MCTA is expected to increase this by a factor of at least 10%. In the LTE simulations for VoIP above, the HARQ scheme used was synchronous, non-adaptive with Chase combining with maximum of 4 HARQ re-transmissions. With 6 HARQ transmissions, the capacity numbers will further improve. Frequency selective scheduling was done for 3 Km/h UE speeds only. For the higher speed UEs, wideband MCS was used for scheduling. The capacity numbers will further increase with the usage of an enhanced delay-based scheduler, which takes into account the channel quality of the VoIP UEs as well. Full rate 12.2 kbps & 7.95 kbps source rate AMR & EVRC codecs used with no bundling & 20 ms frame length. The AMR codecs have 50% voice activity factor with SID payload of 15 bytes every 160 ms during Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

8

silence. The EVRC codec generates one of the following bits every 20 ms: 171 (29%), 80 (4%), 40 (7%), 16 (60%).

6. Additional points of interest We are currently researching spectral efficient link adaptation algorithms for H-ARQ. To recover from link adaptation errors, ARQ can be used. When ARQ is used at the physical layer along with FEC it is called Hybrid – ARQ (H-ARQ). In H-ARQ, the receiver uses the old information (failure packets) for decoding the current transmitted packet. Depending on the way the packets are combined H-ARQ systems can be typically classified into two categories namely, Type- I or Chase combining or Type-II Incremental Redundancy (IR). In Chase combining, the packet transmitted is an exact replica of the original packet. In IR, instead of repeating the same packet, additional redundant information is transmitted incrementally. Typically, the code rate is effectively decreased over each retransmission. With the current link adaptation algorithms, we find that though we get link level gains using IR, we don’t get any large gains at the system or end-to-end throughput level. We propose two mapping schemes for the product that do not require standardization namely maximum effective spectral efficiency mapping -static (MESEMS), maximum effective spectral efficiency mapping –dynamic (MESEM-D) for link adaptation. Our simulation results indicate that, by using these mapping schemes for link adaptation (LA) large gains can be achieved compared to the conventional HARQ schemes. Conventional LA schemes are based on the targeting 10 percent of FER in each HARQ attempt. In our current research, we propose LA schemes which are based on maximization of effective spectral efficiencies (ESE). The main concept is to perform LA with the objective of maximization of ESE defined as the ratio of number of information bits to that of retransmission attempts. We proposed MESEM-S, which maps the MCS obtained for conventional LA to a higher MCS that maximizes the ESE based on a staticTable, while in MESEM-D, the table is generated dynamically. Table 5 shows the performance comparison between the traditional LA schemes using Chase, IR and the proposed schemes namely MESEM-S, MESEM-D. HARQ Scheme

Spectral Efficiency (b/s/Hz)

% FER Outage

% Gain

Chase IR

0.68 0.69

0.51 0.8

---0.34

MESEM-S

0.82

0.8

20.36

MESEM-D

0.83

0.7

22.56

Table 7. Performance gains using spectrally efficient LA algorithms for H-ARQ for an uplink 3Kmph channel

7. Conclusion LTE systems are projected to have higher peak data rates, reduced latency, improved spectrum efficiency, greater system capacity, larger coverage, with reasonable system and terminal complexity. In order to achieve this, several new radio transmission technologies are included. Multiple antennas at the transmitter and receiver using multiple input multiple output (MIMO) techniques have been designed. Our results shows that by using multiple parallel data streams transmission to a single terminal, data rate can be increased significantly. Performance with various capacity enablers is discussed.

Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

9

About the authors Ashvin Chheda has worked for Nortel for 12 years. He is an advisor in the Systems Design group; primarily working on PHY/MAC layer optimization for a variety of technologies such as CDMA, 1xEV-DO, and LTE. He holds an B.Eng in Electrical Engineering & Electronics from UMIST (U.K) and an MSc in Electrical Systems Engineering from the University of Michigan. Sagar Dhakal has worked as an engineer in the Systems Design group at Nortel since January 2007. His primary activity has been in the system level design and simulation of LTE systems. He received the B. Eng. in electrical and electronics engineering in May 2001 from Birla Institute of Technology, India. He received the MS and PhD degrees in electrical engineering, respectively, in 2003 and 2006, from the University of New Mexico, USA. Some of his research interests include statistical signal analysis, wireless communication systems and queuing theoretic model for stochastic networks. Qi Hao is currently a senior network designer in 4G Systems Design group. She has been working in Nortel for 8 years. She has been working on system design, analysis, and simulation for CDMA 1xRTT, 1xEVDO, and LTE systems. Qi Hao got her Ph.D on electrical and electronic engineering in Northeastern University, China, and Post Doctoral Fellows in Carleton University, Canada, and Nanyang Technological University, Singapore. Gaurav Hemrajani joined Nortel in 2004 in the Systems Design team. Initially his work was related to the performance aspects of CDMA & 1xEV-DO systems. In 2006, he started working on the network simulations for 1xEV-DO Rev B & LTE systems. Gaurav holds a Masters of Science in Electrical Engineering with a major in Telecommunications from the University of Texas at Arlington. He is a member of Tau Beta Pi. JoonBeom Kim joined Nortel in 2006. He received the BS and MS degree in Electronics Engineering from Yonsei University, Korea and the Ph. D degree in Electrical and Computer Engineering from Georgia Institute of Technology. He is currently working on the PHY/MAC layers simulation and performance evaluation for LTE systems. His research interests include channel estimation, coding theory, MIMO technologies, and statistical signal processing and communication. Sai Ramesh Nammi is a system designer for OFDMA performance simulation and analysis in Nortel. Prior to this he worked for Motorola India Electronics limited and Qualcomm Inc. on 2G and 3G wireless technologies. He received his masters from Indian Institute of Technology, Madras and received his Ph. D from New Mexico State University. His research interests include iterative detection and decoding techniques for wireless communication systems. Shankar Venkatraman has worked for Nortel for 4 years. He is a Member of Scientific Staff in the Systems Design group; primarily working on Systems Capacity, PHY/MAC layer optimization for a variety of technologies such as CDMA, 1xEV-DO, and LTE. He holds a B.E. in Electronics and Telecommunications Engineering from the University of Mumbai and a M.S. in Electrical Engineering from the University of Houston.

Nortel Technical Conference – June 13-16, 2008

Nortel Confidential Information – For internal use only

10