How to Broadcast Multimedia Contents? II Lessons from The Channel

What Is The Next for Mobile System Design? I A Single-Cell Model Perspective on Downlinks
[How to Broadcast Multimedia Contents? I Introduction]
[How to Broadcast Multimedia Contents? IV Hierarchical Modulation]
[How to Broadcast Multimedia Contents? V Overloaded Transmission and IC]
[How to Broadcast Multimedia Contents? VI Open-Loop MIMO for BCMCS]
[How to Broadcast Multimedia Contents? VII Network Layer or Stream Layer Design]

COST 231 model, which was developed by European COST Action 231, and its variations are the most popular radio propagation model adopted in various standardization bodies, such as 3GPP, 3GPP2 and IEEE. Its modifications include COST 231-Hata Model and COST 231-Walfisch-Ikegami Model. The mathematical formulation of the COST 231-Hata model path loss in dB is

PL = 46.3 + 33.9 logf - 13.82log h_BS - a( h_MS ) + [ 44.9-6.55logh_BS] log d + C

which features a carrier frequency f between 800MHz and 2GHz, an above-neighborhood base station antenna with a height h_BS of 30~300m and a mobile station with an antenna height h_MS of 1~10m. One instance of this path loss model can be shown in Figure 1.

Figure 1. An example of COST 231-Hata urban propagation model.

With Figure 1, we can observe that there are at least two different regions along the channel. The small area close to the base station usually has a high achievable throughput, while the large area on the cell-edge has a low achievable throughput. There two kind of areas are separated in space domain. On the other hand, we alway expect to achieve higher achievable throughput with a good coverage for broadcasting multimedia contents.

Lesson I. Coverage and Throughput Dilemma

Figure 2. The tradeoffs inside the channel

The COST 231 channel model confirms us that there is a well-known trade-off between reception and coverage. In Figure 2, with a 300-meter transmitter antenna, it shows the path-loss changes 0.6560 dB at every 90% coverage change, 1.3894 dB at every 80% coverage change, 2.2209 dB at every 70% coverage change and 3.1807 dB at every 60% coverage change. In general, if you want more coverage, then you may lose some capacity on the cell-edge.  Otherwise, you have to shrink your coverage.

Figure 3. Spectral Efficiency and Coverage Tradeoff

Lesson 2. Gaussian Broadcast Channel and Superimposed Transmission

From Figure 2 and 3, it shows you can't get both coverage and cell-edge throughput at the same time. Though we can't break the trade-offs, is there any way to move the trade-off curves in Figure 3 upwards a little bit? Fortunately, information theory has told us there is another option which is worth thinking about. The idea is called superposition precoding or superimposed transmission. The existing of superposition precoding is mostly due to the nonlinearity of Shannon capacity curve. Superimposed transmission suggests splitting one data stream and simultaneously transmitting together instead of orthogonally transmitting the multiple streams, such as in a time-division multiplexing or frequency-division multiplexing fashion. With this way, a higher achievable capacity is possible. This means, if Signals can be sent from two layers, a base layer and a enhancement layer, in which the base layer has the best coverage and the enhancement layer provides additional throughput, the user capacity can increase about 60% with a 3.2 dB path-loss difference between the two layers. If the difference is 1.4dB, the user capacity can increase about 80%.

Figure 4. Capacity Improvement through SPC

Figure 5. User Capacity and Coverage Tradoff

Lesson 3. Fading Channel and Overloaded Transmission

Figure 6. Fading Channel Capacity

It is well-known that homogeneous fading has no effect on CDMA spectral efficiency. However, higher forward-link spectral efficiency is achievable with taking advantage of multiuser diversity. In fact, if an optimum receiver is used and the system loading β=K/N is sufficiently high, even randomly spread CDMA incurs negligible spectral efficiency loss relative to no-spreading. And when β=K/N increases to the infinity, the achievable capacity with and without fading grows to the same ultimate limit. This means, the fading effect vanishes even for the linear receivers, such as matched filter and the MMSE receiver.

Lesson 4. Open-Loop MIMO Channel Capacity

Consider a M-transmit-antenna and N-receive-antenna MIMO link. When both transmitter and receiver know the channel response, it is well-known that the achievable channel capacity C is proportional to min( M, N )log(SNR) + O( 1 ). If only receiver knows the channel response, the achievable channel capacity C is proportional to min( M, N )log(SNR) + O( 1 ). If neither receiver nor transmitter knows the channel response, the achievable channel capacity C is proportional to min( M, N )( 1 - min( M, N )/T ) log(SNR) + O( 1 ). Therefore, in higher SNR region, that transmitter knows the channel may not help much in terms of achievable channel capacity, even though with proper CSI feedback and MIMO precoding at Tx, the required Rx design can be simplified.

Lesson 5. Break The Dilemma with Relay

Figure 8. Extend coverag with relay