Performance Analysis of Scenario-2

6.4.2.1 Analysis within Effective Coverage

Analysis: Since the distance impact to communication performance is pretty complex, merely judging from the average delay of each node will not be enough. Instead, we observe three indicators for our performance study, namely average delay, BER and packet receiving rate. Firstly, we can see that, within a 1Km distance, all nodes have almost the same average delay and are slightly increased according to distance from data source. This can be explained by an increased BER, which is acceptably given in Figure 6-9, due to a relatively greater attenuation by distance. Since the delay is a compositive performance, including hardware efficiency, propagation, processing, etc., in this case, we can consider the slight increase in node 05 delay being down to more processing time taken by correcting received

bits as well as propagation delay. Furthermore, we know that the relationship between SNR and BER under QPSK from [60] is as follows:

BER=ଵ_ଶ‡”ˆ…ට୉ౘ ୒బ (Equation 6-1) SNR=୉ౘ ୒బ ୖ౩ ୆ (Equation 6-2) Where ୉ౘ

୒బ is carrier-to-noise ratio, ୱ and B represent data rate and bandwidth, respectively.

The two equations above actually indicate that the BER will be dynamically affected by SNR according to distance change (i.e. lower SNR results in higher BER). In real application, with the purpose of extending coverage, vehicles should switch to a lower data rate (i.e. a simpler modulation mechanism to obtain a lower sensitivity requirement or to intelligently increase its transmission power, together with more advanced hardware components, such as an equalizer, in order to cope with multipath propagation and a fading phenomenon).

Meanwhile, there is an abrupt packet receiving rate drop at Node 05, from 98.92 packets/s to 92.45 packets/s, due to an edge effect related to a sensitivity issue, which will be explained in detail later on.

6.4.2.2 Analysis Outside of Effective Coverage

Next, we observe the performance of testing nodes 06 and 07, in contrast with 01, based on their SNR, delay and bit error rates from Figure 6-(11̱13). It can be observed from Figure 6-12 that the longer distance has the larger swing range of delay and thus more scarce sampling nodes on the diagram. Through comparing the BER between nodes 06 /07 and 00, they increased almost by 213% and 100% on average. What we need to focus on is the peak AC 3 delay for each testing node, rather than their average values, as we found that the peak delay for node 00 is around 0.323 ms, compared with 0.766 ms for node 07, which can be explained by the same reason for a processing delay (mentioned above). Although the delay has increased by 137% outside of coverage, under VANET with only three vehicles, it might seem acceptable. However, on the other side, the scarcity observed actually impacts on the

coverage. Firstly, we know that CCH always works on the broadcast model and that there is no retransmission sent differently from normal WLAN applications, in which a user will experience extremely slow throughput and frequent disconnection, but can still feel “useful”.

A typical case might be that your wireless adapter detects weak Wi-Fi signals without

encryption leaking from your neighbour’s house; you can take advantage of slowly opening some web pages rather than watching YouTube smoothly. However, in traffic safety application, almost all messages on CCH are expected to be received without exemption and there is no selective or occasional receiving that is tolerated. Secondly, the cause of lower receiving rates is mainly the sensitivities of receivers; the turning point in communication performance is varying the different modulation schemes. According to Equitation 7-2, the SNR will increase under a higher data rate. Most of the time, the receiving status will abruptly deteriorate when the signal strength is below the sensitivity, thus causing a great amount of data which cannot ultimately be decoded and dropped. Here, we need to distinguish the difference between receiver sensitivity and the packet reception-power threshold. The hard figure of the former is provided by a few manufacturers but is a measure of the minimum threshold at which a device can properly detect and interpret a signal, which implies two facts: firstly, under this threshold, a device is still allowed and might be able to detect signals. On the other hand, even though it can detect the signal, the interpretation will perform improperly without any guarantee. The latter (reception-power threshold), however, is a very precise value under which a signal will not be sensed nor decoded. Therefore, by

setting this parameter with a smaller value, we can see that the signal is being ‘cut off”

immediately without any exemption, even if it can still be properly decoded by a receiver. Since the sensitivity is merely hardware-related and cannot be changed, the power threshold actually has a great meaning to vehicle communication under the two scenarios stated below: 1. If the data source would like to confine the transmission range within certain

coverage or eliminate receivers under a poor communication environment, it can define the fixed value of the power threshold in its packet header rather than just merely increasing its transmission power. This is based on several reasons: firstly, not all vehicles have to receive data, especially in the SCH case. Secondly, since VANET

operates in a highly dynamic environment, merely changing the transmission power would not always help to better the performance but rather increase interference. Furthermore, by informing receivers of the threshold, we actually handle the rights of decisions of acceptance rather than forcing the communication.

2. As we know, under the broadcast mode, CSMA/CA will still be involved in media access contention, which contributes to most of the delay. Generally speaking, how long a certain node needs to wait in order to gain access to the media is dependent on the number of nodes within the effective communication coverage. At the edges of broadcast distances or even further, even though the signal strength is much lower than sensitivity, nodes there might be content with redundant counterparts, which are not supposed to be taken into consideration. For instance, based on common sense, we know that the pre-crash messages are generally only meaningful to the vehicles behind the sending source; this is in order to let drivers brake as soon as possible. However, since mainly isotropic antenna will be adopted in VANET, these messages will inevitably send to the vehicle not only in front but also outside of pre-defined coverage and will thus make receivers try to sense them. Therefore, if vehicles are able to set a reception threshold according to distance and dynamic environment, their performance will not be affected by vehicles outside of their coverage and will look better due to the receivers ignoring signals below their threshold. In fact, in a long queue of vehicles in a traffic congestion period, most vehicles can be considered at the edge of certain effective communication coverage. Thus, this parameter will improve the overall network performance rather than several members of it.