System Level Modeling and Optimization of the LTE Downlink

More documents

Recommendations

Info

5. Performance Evaluation of Fractional Frequency Reuse in LTEServing as an exemplifying point, the FFR configuration with frequency allocationβ FR = 0.31 and SINR threshold Γ thr = 18 dB yields seemingly all-improving UEthroughput results. For this case, the following UE throughputs are observed: meanUE throughput of 3.66 Mbit/s (+11.15 % compared to reuse-1), edge throughput of1.28 Mbit/s (+73.04 %), and peak throughput of 12.91 Mbit/s (+75.28 %). Do notethat this point does not correspond to the optimum point shown on Figure 5.3,but is rather an exemplifying FFR point. A more exhaustive examination of thesimulation results offers a complete view of the distribution of the UE throughput.In Figure 5.6, the UE throughput distribution, shown both as an ecdf (left) anda scatterplot over the FR wideband SINR are shown (reuse-1 is considered for thecalculation of Γ, hence the term “FR SINR”).In the right plot, the boundaryseparating the PR (left, low SINR range) and FR (high SINR range) UEs at Γ =18 dB is marked.Empirical CDF1UE average throughputWideband SINR−to−throughput mapping700.90.8600.7500.6400.50.430PR zone FR zone0.3200.20.11000 10 20 300−5 0 5 10 15 20 25 30average UE throughput [Mbit/s]UE wideband SINR [dB]average UE throughput [Mbit/s]Figure 5.6: UE throughput distribution for the β FR = 0.31 Γ thr = 18 dB FFR point. Left:UE throughput ecdf. Right: UE throughput over wideband SINR (Γ).The ecdf shows that 80 % of the UEs experience low throughput, which albeit low, isstill higher than the reuse-1 edge throughput. This translates into an edge throughputincrease, which combined with a small group of UEs close to the cell centerthat obtain most of the throughput (peak throughput gain) that pushes the averagethroughput up, results in a gain in mean, edge, and peak throughput. However, anequally valid assertion is that the majority of UEs experience a performance degradation,despite what the throughput metrics may indicate, which obviously does notsound as desirable as the first assertion of overall gain.As shown, the typical metrics to evaluate FFR performance can lead to results which,albeit seemingly good, are undesirable. By combining the previously-mentionedthroughput metrics with a fairness metric, a better-suited performance evaluationof FFR performance is proposed [135].Fairness, as first introduced in [57], rates how equally a resource (in this case through-72
5. Performance Evaluation of Fractional Frequency Reuse in LTEput) is distributed over N users. It is defined as( ∑Ni=1 x i) 2J (x) =N ∑ , (5.1)Ni=1 x2 iwhere x is a vector of length N containing the resources obtained by each of theN users. Applied to the results shown in Figure 5.6, fairness can be interpretedas either the steepness of the throughput ecdf or alternatively the flatness of theSINR-to-throughput mapping.Figure 5.7 depicts, employing the same visualization as in Figure 5.5, the obtainedfairness for all of the simulated β FR and Γ thr FFR value pairs, as well as the UEthroughput ecdf associated the following three points of interest in the plot: (i)reuse-1, (ii) reuse-3 (its closest FFR approximation, where β FR = 0.01), and (iii)β FR = 0.34, Γ thr = 10 dB, where a high fairness gain relative to reuse-1 is achieved.Empirical CDF0.91 reuse-10.80.70.60.50.40.30.20.100 1 2 3 4 5 6 7 8 9UE throughput [Mbit/s]SINR threshold [dB]20151050(i)0.90.7Fairness0.5(iii)0.30.1(ii)0.90.80.70.60.50.40.30.20.1Empirical CDFEmpirical CDF0.91 reuse-30.80.70.60.50.40.30.20.100.5 1 1.5 2 2.5 3 3.5 4UE throughput [Mbit/s]0.910.80.70.60.50.40.30.20.100 2 4 6 8 10UE throughput [Mbit/s]Figure 5.7: FFR fairness results employing round robin scheduling. The associated UEthroughput ecdf is shown for the (i) reuse-1 case, (ii) reuse3 case, and (iii)β FR = 0.34, Γ thr = 10 dB case.Ideally, an operator of an LTE network would find it desirable to obtain a fairnessgain (or at least not lose any fairness so as to avoid starvation of some UEs),while maintaining or ideally improving average throughput. Thus, ensuring that cellthroughput (i) is not reduced and (ii) is shared optimally among UEs.With this constraint, a fairness increase while maintaining average throughput impliesa throughput gain for the UEs with poor channel conditions, and vice versa.For the case depicted in Figure 5.6, a degradation in fairness from 0.69 (reuse-1fairness) to 0.32 was observed, which indicates that the observed throughput gain isdue to a more biased throughput distribution.If just the area where J FFR > J reuse1 is taken into account, the mean, edge, and73
Page 1:
DissertationSystem Level Modeling a
Page 5:
I hereby certify that the work repo
Page 9:
KurzfassungDie vorliegende Arbeit p
Page 13 and 14:
ContentsContents1 Motivation and Sc
Page 15 and 16:
Contentsix
Page 17 and 18:
1. Motivation and Scope of Work1. M
Page 19 and 20:
1. Motivation and Scope of Worklaye
Page 21 and 22:
1. Motivation and Scope of Workperf
Page 23 and 24:
BibliographyThe combined throughput
Page 25 and 26:
2. 3GPP Long Term Evolution2. 3GPP
Page 27 and 28:
2. 3GPP Long Term Evolutionup to 30
Page 29 and 30:
2. 3GPP Long Term EvolutionNAS proc
Page 31 and 32:
2. 3GPP Long Term Evolutionchannel
Page 33 and 34:
2. 3GPP Long Term EvolutionSince th
Page 35 and 36:
2. 3GPP Long Term EvolutionThe chan
Page 37 and 38: 2. 3GPP Long Term EvolutionTable 2.
Page 39 and 40: 3. Physical Layer Modeling and LTE
Page 67 and 68: 4. Extensions to the L2S Model4. Ex
Page 69 and 70: 4. Extensions to the L2S Model4.1.2
Page 71 and 72: 4. Extensions to the L2S ModelTo ac
Page 73 and 74: 4. Extensions to the L2S ModelBLER1
Page 75 and 76: 4. Extensions to the L2S ModelThe M
Page 77 and 78: 4. Extensions to the L2S Modelthe a
Page 79 and 80: 4. Extensions to the L2S ModelIn th
Page 81 and 82: 5. Performance Evaluation of Fracti
Page 87: 5. Performance Evaluation of Fracti
Page 97 and 98: 6. Summary and Outlook6. Summary an
Page 99 and 100: 6. Summary and Outlook6.2. OutlookW
Page 101 and 102: A. SNR-independence of the CLSM Pre
Page 107 and 108: B. Correlation Matrices for Shadow
Page 109 and 110: C. Taylor Expansion of the ZF MSEC.
Page 111 and 112: C. Taylor Expansion of the ZF MSEAp
Page 113 and 114: D. Evaluation of Multi-User GainD.
Page 115 and 116: D. Evaluation of Multi-User GainAvg
Page 117 and 118: Abbreviations and AcronymsAbbreviat
Page 119 and 120: Abbreviations and AcronymsKPILLRL2S
Page 121 and 122: Abbreviations and AcronymsU-PlaneUT
Page 123 and 124: Abbreviations and AcronymsBibliogra
Page 125 and 126: Abbreviations and Acronymsdescripti
Page 127 and 128: Abbreviations and Acronyms[69] Tech
Page 129 and 130: Abbreviations and Acronymsfemto cel
Page 131: Abbreviations and Acronyms[134] Z.
show all

System Level Modeling and Optimization of the LTE Downlink

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?