Frequency-domain scheduling, multiple antenna techniques, and rate adaptation enable next generation orthogonal frequency division multiple access (OFDMA) cellular systems such as Long Term Evolution (LTE) to achieve significantly higher downlink spectral efficiencies. However, this comes at the expense of increased feedback overhead on the uplink. LTE uses a pragmatic combination of several techniques to reduce the channel state feedback required by a frequency-domain scheduler. In subband-level feedback scheme specified in LTE, the user reduces feedback by only reporting the channel quality indicator (CQI) computed over groups of resource blocks called subbands. LTE also specifies an alternate user selected subband feedback scheme, in which the feedback overhead is reduced even further by making each user feed back the indices of the best M subbands and only one CQI value averaged over all the M subbands. The coarse frequency granularity of the feedback in the above schemes leads to an occasional incorrect determination of rate by the scheduler for some resource blocks. The overall throughput of LTE depends on the method used to generate the CQI and the statistics of the channel, which depends on the multiple antenna technique used. In this thesis, we develop closed-form expressions for the throughput achieved by the user selected and subband-level CQI feedback schemes of LTE. The comprehensive analysis quantifies the joint effects of four critical components on the overall system throughput, namely, scheduler, multiple antenna mode, CQI feedback scheme, and CQI generation method. The performance of a wide range of schedulers, namely, round robin, greedy, and proportional fair schedulers and several multiple antenna diversity modes such as receive antenna diversity and open-and closed-loop transmit diversity is analyzed. The analysis clearly brings out the dependence of the overall system throughput on important parameters such as number of resource blocks per subband and the rate adaptation thresholds. The effect of the coarse subband-level frequency granularity of feedback is explicitly captured. The analysis provides an independent theoretical reference and a quick system parameter optimization tool to an LTE system designer. It also helps us theoretically understand the behavior of OFDMA feedback reduction techniques when operated under practical system constraints. Another contribution of this thesis is a new statistical model for the effective exponential SNR mapping (EESM), which is a highly non-linear mapping that is widely used in the design, analysis, and simulation of OFDMA systems. The statistical model is shown to be both accurate and analytically tractable, and plays a crucial role in facilitating the analysis of the throughput of LTE when EESM is used to generate the CQI.

Wireless communication networks have evolved continuously over the last four decades in order to meet the traffic and security requirements due to the ever-increasing amount of traffic. However this increase is projected to be massive for the fifth generation of wireless networks (5G), with a targeted capacity enhancement of 1000× w.r.t. 4G networks. This enhanced capacity is possible by a combination of major approaches (a) overhaul of some parts and (b) elimination of overhead and redundancies of the current 4G. In this work we focus on OFDM reference signal or pilot tones, which are used for channel estimation, link adaptation and other crucial functions in Long-Term Evolution (LTE). We investigate two aspects of pilot signals pertaining to its evolution - (a) impact of targeted interference on pilots and its mitigation and (b) adaptation of pilot patterns to match the channel conditions of the user. We develop theoretical models that accurately quantify the performance degradation at the user’s receiver in the presence of a multi-tone pilot jammer. We develop and evaluate mitigation algorithms to mitigate power constrained multi-tone pilot jammers in SISO- and full rank spatial multiplexing MIMO-OFDM systems. Our results show that the channel estimation performance can be restored even in the presence of a strong pilot jammer. We also show that full rank spatial multiplexing in the presence of a synchronized pilot jammer (transmitting on pilot locations only) is possible when the channel is flat between two pilot locations in either time or frequency. We also present experimental results of multi-tone broadcast pilot jamming (Jamming of Cell Specific Reference Signal) in the LTE downlink. Our results show that full-band jamming of pilots needs 5 dB less power than jamming the entire downlink signal, in order to cause Denial of Service (DoS) to the users. In addition to this, we have identified and demonstrated a previously unreported issue with LTE termed ‘Channel Quality Indicator (CQI) Spoofing’. In this scenario, the attacker tricks the user terminal into thinking that the channel quality is good, by transmitting interference transmission only on the data locations, while deliberately avoiding the pilots. This jamming strategy leverages the dependence of the adaptive modulation and coding (AMC) schemes on the CQI estimate in LTE. Lastly, we investigate the idea of pilot pattern adaptation for SISO- and spatial multiplexing MIMO-OFDM systems. We present a generic heuristic algorithm to predict the optimal pilot spacing and power in a nonstationary doubly selective channel (channel fading in both time and frequency). The algorithm fits estimated channel statistics to stored codebook channel profiles and uses it to maximize the upper bound on the constrained capacity. We demonstrate up to a 30% improvement in ergodic capacity using our algorithm and describe ways to minimize feedback requirements while adapting pilot patterns in multi-band carrier aggregation systems. We conclude this work by identifying scenarios where pilot adaptation can be implemented in current wireless networks and provide some guidelines to adapt pilots for 5G. / Master of Science