From smart cities to agriculture, LoRaWAN has demonstrated adaptability across various domains, continuously evolving and unveiling new capabilities. We now enter the era of 2.4 GHz LoRa, a development extending the reach of this already remarkable technology, offering exciting possibilities.
In this post, we explore the potential of amalgamating 868/915MHz LoRaWAN with 2.4GHz LoRa, journeying through wireless communication where the fusion of frequencies elevates innovations. The examples used in this blog post were presented by our CEO Luka Mustafa at the latest TTN Conference in Amsterdam.
Selecting wireless technology
Industrial and environmental challenges are constantly pushing the capabilities of wireless technologies. It has always been of key importance to select the best-fitted technology for your end application. Typically, we deal with the following considerations:
- Data throughput;
- Power requirements;
For successful IoT deployments, it is often crucial to utilize a combination of multiple communication technologies, with each one optimally covering a part of the application it is best suited for. Consider a system that:
- Uses a device with an e-paper display to show real-time sensor values.
- Implements cloud monitoring for a fleet of devices with a quick refresh time.
- Is installed in an industrial building with multiple rooms and floors.
- Is deployed independently of the existing IT infrastructure.
- Operates with no recurring costs and is user-operated on a daily basis.
For such a system:
- Actions on the cloud must feel responsive to the user.
- Devices need to be capable of sending updates every minute or so.
- Easy deployment is crucial, necessitating only a few gateways.
Could LoRaWAN meet the requirements?
Given the above conditions and requirements, could LoRaWAN fulfil these needs effectively?
Looking at the figure below, we see that no combination of technologies is really perfect for the use case. The red circle depicts a gap in which no technology is a perfect fit for what we need in terms of the mentioned boundary conditions.
Battle of the bands
Executing a ‘battle of the bands’ scenario reveals the capabilities and limitations of each band. The 868/915 MHz LoRa excels in indoor coverage and deep penetration, but it has a limited duty cycle (at least at 868 MHz in the EU). We are allowed:
- A payload 222B, which is 97 messages/hour at SF7;
- A payload 50B, which is 300 messages/hour at SF7;
- A payload 50B, which is 12 messages/hour at SF12.
Moving to the 2.4GHz band removes duty cycle limitations and increases throughput, allowing:
- One message every 2.5s;
- 1440 messages/hour.
A combination of both for maximal benefit
Much like we described the benefits of using a combination of different technologies, we can apply the same logic to using LoRa in two frequency bands, combining the benefits of both.
With 868/915 MHz we can successfully send one heartbeat message every 300s for maximal range at SF12. We can then utilize the 2.4GHz band to send one message every 2.5s with real-time data. This way, we get maximal coverage with basic service for critical data, and we get the best effort full service with real-time data.
Moreover, a system like this is able to notify us when we’re out of the 2.4GHz range, but we don’t lose any information.
The fusion of both bands should be a meticulously orchestrated, systemic endeavor. Properly configuring these bands unveils a plethora of applications previously deemed unfeasible with LoRa.
In AI, numerous Machine Learning models process small images in real-time. Historically, due to bandwidth constraints, only ML outputs were transmitted, not the actual image content. This can and will still persist; however, with 2.4GHz LoRa, we can also cover the scenarios when human validation is required (e.g. alert scenarios, or classification of new objects).
Consider, for instance, two JPEG images: one 6kB (160×160) and another 56kB (720×720).
- With 868MHz at SF7 in the EU: 21 kB/h (Duty-cycle limited)
- With 2.4GHz at SF12: 23kB/h (RX window limited in LoRaWAN)
The Faster Objects, More Objects (FOMO) approach segments images into smaller 8×8 chunks, performing ML on each segment. Only the segments of interest are reported as output. When necessary, we can send only a few 8×8 image chunks for validation. 2.4 GHz LoRaWAN can send about 1kB of data in 15s, sufficient for an image chunk.
Real-time ML models often analyze audio within a moving window, typically of 100ms duration. After we’ve trained the model, known phenomena can be successfully classified, however new and unknown anomalies can be collected to improve the future model enhancements.
Events of short duration, like impulses, bangs, or gunshots, can be conveyed over 2.4GHz LoRa in a few packets for manual classification and subsequent model refinement.
The incorporation of 2.4GHz LoRa introduces multifarious advantages, enabling innovative applications, especially in medical, security, and industrial domains focusing on swift and reliable communication of critical real-time information. With ML being present in an increasing number of end applications, 2.4GHz LoRa enables the enhancement of ML models post-deployment, at no additional costs.