Measuring LoRaWAN capacity in dense urban environments
01 June 2020
Much has been written about wide area networks (WANs) relying on wireless technology to transfer data across long distances. However, just how reliable are they in dense urban environments? What packet success rate can be achieved and is interference a problem? Or is gateway density an issue – and what happens when multiple sensors interact with multiple gateways?
This case study was originally featured in the June 2020 issue of EPDT magazine [read the digital issue]. Sign up to receive your own copy each month.
Here, Patrick van Eijk, Director of IoT Solutions at high performance analogue & mixed signal semiconductors specialist, Semtech tells us about a recent study that evaluated the ability of enterprise-grade networks to handle sensor-generated traffic...
The recent study in Philadelphia set about measuring exactly how effective a long-range wide area network, LoRaWAN, is at handling large amounts of sensor-generated traffic in a busy environment. It looked at an urban area of around a ¼ square mile, which contained roughly 100 residences per city block and 1,000 residences in the deployment area. Each site would have an average of 10 LoRa® enabled Internet of Things (IoT) devices, which would send upstream packets every hour, meaning that in a 24-hour period the network would handle just short of 250,000 packets.
For comparison purposes, stressful traffic conditions were simulated handling 500,000 (Phase 2) and 1m packets a day (Phase 3), while transmitting at rates that achieved the same traffic as 10,000 sensors. The packets were sent at randomised intervals through the 10 MultiTech Conduit 8-channel indoor gateways to mimic a real-world setting and ensure packet overlap took place.
One of the objects of the exercise was to assess if a packet success rate of at least 90% could be achieved when sending each packet once and without acknowledgment. At the same time, the study also sought to understand LoRa devices’ high-capacity features and check the technology’s ability to deliver multiple overlapping packets successfully.
The aim was to understand gateway diversity and how each device would communicate with multiple gateways when transmitting data, while checking the suitability of a cloud infrastructure for high-volume commercial applications.
Other objectives included testing network server scalability and monitoring what effect environments have on signal propagation, packet success rate and gateway diversity. It was also useful to see what impact factors such as multipath fading, radio interference and backhaul throughput would have on capacity.
Placement and use
All gateways were self-installed by the study’s participants, who made sure they were placed near Wi-Fi access points or TV cable boxes. Multiple sensors were installed in such a way as to represent real-life scenarios with different levels of radio frequency (RF) penetration – such as in basements, ovens, between appliances and inside kitchen cabinets. Up to 15% of the sensors were located in parked cars around the trial area to mimic light indoor use.
To represent a generic LoRaWAN-based sensor with a 5cm rubber antenna, MachineQ used the standard Semtech evaluation kit (Nucleo Pack), which includes an SX1276MB1LAS radio board paired with a STM32L073RZT Nucleo board. Kits were powered by generic USB batteries and all sensors were configured for 18 dBm TX power and an 11-byte payload. All gateways and end-devices used Channels 24-31 as defined in the LoRaWAN North American regional specification.
The time between packets being sent by the devices varied from four seconds to 60 seconds, while maintaining an overall average packet rate of once every 32 seconds. The Phase 2 rates were used to evaluate network behaviour where there is a high probability of on-air packet collisions, while Phase 3 was intended to identify a maximum sustainable network load, while highlighting any variation in inter-packet delay.
One way the success rate of the study was judged was by monitoring the packet error rate (logging missing sequence numbers). If a packet wasn’t received, it was missing in the incremental sequence. The results showed that whatever the packet volume, the delivery success rate was over 95%, with an average packet loss of 3.77%.
In detail, with a 247,779-daily packet count there was a 97.73% success rate, at 500,000 the rate was 97.25% and when the count doubled again to 1m the rate was 96.23%. Achieving this level of success under such extreme load conditions demonstrated the enterprise-grade potential of a professionally installed and managed dense LoRaWAN network.
Also, out of the 915,000 packets generated by the study, several thousands of them were received nearly 1.5 miles away, by multiple gateways at the MachineQ laboratory, which is obstructed by a substantial urban infrastructure. Furthermore, packet overlap was monitored where two or more packets survived the overlap and were delivered error-free to the cloud.
Throughout the length of the study, the network server and cloud infrastructure were found to support all the traffic successfully, despite a real-world cluttered urban environment. It was clear that propagation through a residential urban environment wasn’t a problem for an enterprise-grade LoRaWAN network. The results also indicated that multipath fading, radio interference and backhaul throughput would not have a significant impact on real-world deployment.
The study also recorded that a significant percentage of packets were overlapping with traffic from other sensors. This happens when two or more packets are received at the same time in high-density traffic conditions.
LoRa enables overlapping packets to survive potential collision and to arrive at the network server error-free. In any other viable IoT technology, two or more packets concurrently appearing on the same channel would result in packet loss. LoRa ‘orthogonality’ increases network capacity and, when combined with high gateway density – which represents the number of gateways that receive each packet, and is an important measure of a full-scale network – results in enterprise-grade high capacity.
It’s clear from the study that a LoRa-based system can perform reliably at various capacity levels, and a full-scale LoRa deployment would be able to handle more than the expected number of packets per day in a dense network environment.
The study showed that with a success rate almost up to 98%, LoRa can be relied on to transmit data at scale, and a maximum network load was close to two million packets per day. At full-scale deployment, the network used gateway diversity successfully, with each device communicating with different gateways when transmitting data. On average, each sensor communicated with 2.5 gateways, confirming the success of orthogonality in full-scale deployment, which is a major advantage of a multi-tenant network.
The Cloud infrastructure and network server also performed successfully in the full-scale deployment scenario, but more work is needed to analyse how specific environmental factors can affect propagation, packet success rate and gateway diversity.
Looking ahead, network designers can be confident that capacity is not an issue for LoRaWAN coverage, and because the technology allows end-devices to transmit at any time, complexity is shifted from the devices to the network. This significantly reduces the amount of energy that’s needed at the end-device to manage and maintain network co-ordination.
In short, the study confirmed the availability of a reliable, enterprise-grade LoRaWAN network for a variety of IoT applications.
Contact Details and Archive...