In Figure 5.8, numerical result of the link data rate is shown via offline simulation for different feedback schemes that were discussed in Section 5.4. The distribution of link data rate from
the relay to different indoor spatial locations (1000 samples) is shown. The median data rate is measured at the 50 percentile point of the cumulative distribution function (cdf) shown in Figure 5.8.
Coverage improvements are represented by improvements in coverage data rate and minimum data rates, which are defined as follows. Coverage date is measured at the 5 percentile point of the cumulative distribution function of the data rates collected within an indoor cell. The minimum data rate is the tail of the cdf curve, i.e the data rate corresponds to the zeroth percentile. All data rates are calculated by applying (5.6).
From the relay-to-user link data rates, the end-to-end efficiency of relaying from (5.4) can be deduced by taking into account the time sharing between the relay’s feeder and access links, and the feeder link efficiency. In practice depending on an operator’s choice of feeder bandwidth and the feedback scheme, the effectiveness of relaying may vary, as is evident from (5.1) - (5.5). To observe the effect of feedback, the data rate distribution with and without a relay are compared for different feedback qualities in Figure 5.8 using a target PER of 10−6. For these simulations, the stored set of channel matrices from the measurements on floor 6 were used. For consistency, we ensure that the simulations with the fine feedback scheme closely fit the real-time bit rate measurements shown in Section 5.6.1. A slight deviation can occur because the real-time measurements underwent impairments such as phase noise error which is not modeled to generate Figure 5.8. Results show that in the median the coarse feedback scheme behaves almost as well as the fine feedback scheme for the macrolink coverage (without a relay). The averaging effect over the frequency selective channel in the coarse scheme compensates for the lack of granular feedback. A significant gain is obtained from the fine feedback scheme for minimum data rate. Even though the minimum rate is zero for both the schemes, the probability is just 7 % in the fine scheme as compared to 40 % in the coarse scheme.
The extended scheme improves both the minimum and median data rates from the outdoor base station. This shows the effectiveness of closed loop precoding for the macro link, especially when more channel coding options are available. But the downlink gain may come at the cost of uplink loss. The implication on the uplink can be better understood in the light of voice over IP simulation
0 20 40 60 80 100 120 140 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Link data rate [Mbps]
Empirical CDF Macro − Coarse Macro − Fine Macro − Extended Relay − Coarse Relay − Fine Relay − Extended
Figure 5.8: Performance comparison of various feedback schemes in distribution of data rate using 20 MHz bandwidth. The better performing curves are to the right.
0 20 40 60 80 100 120 140 0 0.05 0.1 0.15 0.2
Link data rate [Mbps]
Empirical CDF Macro − Coarse Macro − Fine Macro − Extended Relay − Coarse Relay − Fine Relay − Extended 5 percentile
Figure 5.9: Lower percentile data rate region zoomed from Figure 5.8. The better performing curves are to the right.
results shown in [107]. The uplink satisfaction chances are seen to drop sharply with path loss for a 42 byte payload and for 200 users in a cell. Thus it may be challenging to support several voice over internet protocol (VoIP) users who are indoors at the same time.
The 128 bits of extended feedback may be encoded to 32 bytes with 12 rate channel coding. If the feedback were to be updated once in 40 ms, then 32 byte of feedback comparable in size to the 42 byte VoIP payload. A macrocell may prioritise VoIP data in the uplink, and hence compromise the feedback quality. In the case of a relay, the 30 dB average SNR gain can be used for more feedback bits. Suppose there are only few resource blocks available for feedback. An indoor user
can communicate more feedback bits to the relay by using higher order modulation. For example, instead of QPSK, 16 QAM modulation can be used on the feedback bits.
Based on this viewpoint, relaying has better chances of employing the extended scheme (red curve in Figure 5.8). In Figure 5.9 we observe that closed-loop precoding on the indoor channel increases the coverage data rate (5 percentile of the cdf) from 47 Mbps to 62 Mbps, which is a 32% gain. The minimum achievable data rate increases from 30 Mbps to 45 Mbps which is a 50% gain. The main benefit from this access link improvement is that a higher feeder to access bandwidth ratio ∆ can be utilised to attain the same spectral efficiency of (5.4). The higher feeder bandwidth in turn will provide a better user efficiency Ur as can be seen in (5.5).
With an indoor relay, the coarse level feedback is able to achieve 60 Mbps in the median i.e, a spectral efficiency of 3 b/s/Hz. This result is significant for deployment, because the peak data rate is achievable in 35% of the locations even with the coarse feedback scheme. A two-pronged approach will therefore improvise: a) for median data rate, employ coarse feedback based on user selection and b) for minimum data, increase the feedback to an extended level.