System Level Modeling and Optimization of the LTE Downlink

4. Extensions to the L2S ModelTable 4.2.: Average deviation of the modeled 10% BLER points [dB].mAWGNPed-B0 1 2 3 0 1 2 34-QAM 0.02 0.11 0.06 0.03 0.11 0.22 0.14 0.2016-QAM 0.04 0.40 0.16 0.22 0.02 0.23 0.58 0.9864-QAM 0.07 0.30 0.29 0.79 0.11 0.39 0.85 2.59well as the coherence time assumptions in Section 3.1.1, the impact of the modelinaccuracies at high MCS and retransmission count is significantly reduced due tothe improbability of such retransmissions. This assumption is also backed by theresults in [109], in which it is shown that, for a MIMO cell setup such as that shownin Figure 3.7, 64-QAM retransmissions account for less than 0.05% of the totalnumber of TBs 1 .Table 4.3.: Minimum and maximum SNR gain due to the m-th HARQ retransmission withrespect to the previous retransmission for each of the employed modulations.ITU Pedestrian-B channel (5 km/h).4-QAM 16-QAM 64-QAM1st re-tx 3.56 dB - 5.3 dB 4.79 dB - 6.13 dB 6.38 dB - 14.28 dB2nd re-tx 1.98 dB - 2.77 dB 2.19 dB - 2.85 dB 2.53 dB - 4.19 dB3rd re-tx 1.11 dB - 1.6 dB 1.27 dB - 1.78 dB 1.68 dB - 2.89 dB4.2. Channel Estimation ErrorThis section extends the ZF-receiver-based post-equalization SINR to the case ofimperfect channel knowledge, adding to the model detailed in Section 3.1.1.1 [117].Analogously to Equation (3.8), the post-equalization SINR for the i-th layer, denotedas γ i , is expressed asγ i =P TX[ MSE ] ii, (4.9)where P TX denotes the signal sum power sent over the transmit antennas, denotedas σx 2 0in the remaining expressions in this chapter for mathematical consistencywith the transmitted symbol vector x, MSE the R ν×ν Mean Square Error (MSE)matrix, and [·] iithe i-th element of the matrix diagonal.1 In CLSM/OLSM, the level of spatial multiplexing can be adjusted, in addition to the MCS. As inSISO, varying the MCS is the only available rate-adjusting mechanism, the ratio would be higherfor SISO transmissions, but nevertheless of minor impact.58