Jan 4, 2021
Qualcomm products mentioned within this post are offered by Qualcomm Technologies, Inc. and/or its subsidiaries.
We’ve seen several questions raised about the performance of C-V2X sidelink performance versus Dedicated Short Range Communication (DSRC) under highly congested conditions with active congestion control (for example, see  and  under References below). The primary contention is that with congestion control, active collisions between the semi-persistent scheduling (SPS) transmissions of C-V2X nodes may lead to successive losses, resulting in a long tail for the inter packet gap (IPG) distribution. To understand the significance of IPG as a figure of merit for Basic Safety Message (BSM) transmission, we must first understand what it means in a safety context. IPG is the time between two consecutive BSMsso when there is a wide distribution with a long tail means a number of vehicles at those tails do not receive timely safety messages. That is not good.
In this blog post, we show that the purported longer tail of the distribution from C-V2X versus DSRC is due to:
- choosing corner cases where transmission periodicities are almost equal to a round multiple of 100ms
- an incomplete implementation of the C-V2X protocol standard
- the use of highly forgiving path loss models that mask DSRC’s lower link budget.
First and foremost, C-V2X has a similar or shorter IPG tail distribution at all distances when compared to DSRC – showing again the superior performance of this more modern radio.
How are those studies able to claim poor C-V2X IPG performance? In real-world scenarios with real devices, the transmit periodicities do not align as the corner cases that are simulated. Based on our experience from field testing, devices will have a wider distribution of the Inter-Transmit Time (ITT), and this is also seen in our simulations of the same corner cases. This distribution serves to decorrelate overlap times between devices that share resources.
Consequently, C-V2X does not exhibit the tail behavior claimed by some. Furthermore, the muting mechanisms introduced in SAE J3161/1 (On-Board System Requirements for LTE V2X V2V Safety Communications)  for C-V2X eliminate the tail behavior even when periodicities are rounded to the closest multiple of 100ms. As a result, even when the transmit periodicities do align in heavy congestion where all devices back off to 600ms periodicities, the tail behavior is not seen for C-V2X implementations compliant to SAE J3161/1. Another contentious aspect is the use of the Utra-High Frequency (IHF) path loss model from ITU-R P.141 , which does not take into account the shadowing caused by intervening cars in the simulated congested highway scenarios; a more stringent path loss model such as the freeway path loss model used in 3GPP TR 36.885  is more appropriate for these scenarios, and is used in the simulation results shown in this post.
We illustrate our point with a simulated six-lane highway scenario with 2000 C-V2X equipped cars distributed over a 5000m stretch of highway. We used the line-of-sight (LOS) freeway path loss model described in 3GPP TS 36.885. This model is appropriate for crowded highways considered in these simulations, whereby cars act as blockers that weaken signal propagation.
The plot below shows the IPG performance of C-V2X with and without the J3161/1 muting mechanism at 10 and 20 MHz with 200m between vehicles. In these plots, the IPG represents the duration of time between the successful reception time of a packet received from a car within the 190m to 210m range, and the reception time of the preceding successfully received packet from the same car. Thus, IPG is a base indicator of how well a car can be tracked by the receiving device. As seen in the Figure 1, the tail behavior is not present in any of the cases. Muting does improve IPG tail performance. Also, C-V2X’s ability to make use of flexible bandwidth means that when 20 MHz are available (as in China and the U.S.), C-V2X will perform even better. Finally, the DSRC performance is relatively poor due to its lower link budget.
Figure 2 shows the performance of C-V2X with and without the SAE J3161/1 muting mechanism at 10 and 20 MHz with 300m between devices. As expected, performance is worse at this longer distance. Still, messages transmitted from devices 300m away have a 99% probability of being received within 3.5s in the 20 MHz case, and a 95% probability in the 10 MHz case. Again, the tail behavior is not present in any of the cases. Muting and more bandwidth (if available) improve IPG tail performance. DSRC performance is again relatively worse at this longer distance due to its lower link budget.
To emphasize, in realistic scenarios C-V2X devices do not have periodicities that are almost equal to a multiple of 100ms, so resource usage will occasionally de-overlap, leading to reduced IPG. In the two aforementioned examples, the calculated ITT, according to SAE J2945/1 congestion control algorithms, oscillated in a 30ms range close to 300ms. However, let’s consider the corner case where the BSM periodicity is fixed to 300ms for each V2X device in the 2000-car scenario. Figure 3 is generated with this fixed ITT, using the 3GPP freeway path loss model. The tail of the distribution appears only when muting is not implemented. Even in this artificial ITT scenario, when the SAE J3161/1 muting enhancement is used, the tail of the distribution disappears.
This observation leads us to conclude that the tail behavior for IPG observed by some is an artifact of the assumed vehicles locations, ITT setting, or an inaccuracy in modelling of muting mechanisms.
Hence, C-V2X performs very well in dense highway scenarios. Moreover, the cumulative distribution of the IPG shows good tail behavior in each scenario with the implementation of the muting mechanism specified in SAE J3161 and even in practical congestion scenarios without muting.