The tremendous value in connecting devices and harvesting data comes with a big responsibility of making the right choices to achieve the engineering excellence. The main challenge here is not the lack of choices, but that there are way too many choices to make, be it the protocol selection, power consideration, data sampling, wireless technology and so on. It gets more complicated as each one of these are not to be seen as independent choices to be made, but inter-dependent. However, understanding the relationship amongst them empowers you to make the right choice.
It is important to understand the key factors that influence the design choices when it comes to IoT Edge design. Those key factors are,
The IoT Edge devices could vary from coin-cell battery powered to solar-powered to industrial devices that are powered 24×7 with backup batteries. Power is one of the important considerations, as the choice of the data transmission frequency and amount of edge computation will have an impact on the battery life. For example, an IoT BLE device running on a coin-cell battery will have to consider how much and how frequently it will transmit data to optimize the battery life to sustain for a year or two. Another factor to consider is that the broadcasting power setting in the chipset is typically fine-tuned based on the required range and deployment environment. Two important elements that has a direct relationship to the power are data transmission frequency and amount of CPU cycles that are required on the edge device. One of the common optimization techniques employed for power and bandwidth includes sampling the data at a set interval and aggregate, but turning on the transmitter radio at a different interval. In cases of using LTE radio, turning on the LTE radio only during transmission saves considerable power. But, caution must be taken not to turn ON and OFF the radio too frequently as it can be counter-productive. Also, the tolerance for latency must be considered based on a case-by-case basis before making the choice of store-and-forward. One other way to optimize the data transmission is to perform some basic computation on the edge to avoid sending noise, but send only significant information to the cloud. It is important to understand that computation involves significant power, sometimes even more than using LTE radio. For example, sending data using a LTE radio at 1 Mb/s rate could consume approximately 1700 mW, whereas running ARM processor with 100% CPU cycles could consume 2000 mW.
Before making IoT Edge power design choices here are some common questions we should be asking ourselves –
- “Is the edge device connected to a power socket or battery powered?”
- “If battery powered, do we have a separate Field Gateway (that is powered to a socket) that this IoT device can connect to send data?”
- “What is the amount and frequency of the data to be sent?”
It is important to understand that the choice of the network, application and messaging protocol is a function of various factors – where data transmission rate, battery power and range are the most important.
The data transmission rate will usually depend on the required fidelity and amount of data that is transmitted. Often, optimizing on the data transmission rate will compromise the fidelity which might have a negative impact on the use-case for this data. In some cases, the missing data can be extrapolated or interpolated using techniques such as regression, curve-fitting and smoothing.
It is also important to strongly understand the deployment environment to make the right design choices. For example, a device in the mining field with little to no cellular signal might have to resort to satellite or data muling. In cases where devices are not connected to the internet, data muling can be used. This is when a mobile device (such as a user’s mobile phone) can connect to the IoT device when it comes in proximity to harvest the data to send it to cloud. Some IoT use cases are implemented using crowd-sourced data muling where the general public’s mobile phone acts as a gateway when they come in proximity to the IoT device. The disadvantage with data muling is the latency that is not deterministic often.
These factors are very closely related to each other and they all have a direct relationship to the cost of the edge device hardware and also the software solution. The importance of the IoT edge device design phase cannot be understated as one undertakes the journey of connectivity.