Inside Energy Harvesting Sensors

Sensors Insights by Neil Cannon

The Internet of Things (IoT) combines input data, processing resources, and actuators to enable new applications. Delivering relevant input data is therefore a central requirement for IoT applications. For developers, energy-harvesting wireless sensors are of special interest as they offer flexible radio communication while eliminating the need for maintenance. These sensors have seen a dramatic rise in interest, as they are a reliable and easy to install technology that delivers the essential input data on which the whole IoT model depends.

Creating a sensor node that fulfils the essential data capture, processing, and transmission within the minimal energy that can be harvested from the environment is a challenge. There are three key tasks in energy harvesting wireless sensors: generating (harvesting) the required energy, sensing and processing environmental parameters (e.g. temperature, humidity, position), and wirelessly transmitting the collected information. All three tasks need to be optimized together to provide viable solutions.


Energy-efficient system design

The most common forms of energy harvested by IoT sensors are kinetic, solar and thermal (see “Technologies For Harvesting Energy” below). All three harvesting technologies provide comparatively small amounts of energy (typically in the microampere range). Power-optimized system design is therefore essential to enable wireless sensors based on these energy sources.

Three main tasks define the power budget of a wireless sensor node – sensor measurement, wireless transmission and idle (inactive) state. System design must balance the harvested energy with the power requirements of these tasks. This balance can be established in two directions – either the system functionality (and hence the required power) is fixed and the harvester is scaled or the energy delivery of the harvester is fixed and the system functionality needs to be optimized accordingly. The second case is the more common one.

To illustrate the requirements, we will consider the case of a solar-based room control unit. Its main tasks are to measure temperature and humidity in a room and compare them with user-defined set points (usually target temperature only, sometimes also target humidity). The available energy budget is limited here by the available size for the solar cell (e.g. 5 cm2) and the expected minimum illumination level (200 Lux for 6 hours).

Sensors Insights 2017-04-14 Fig1
For an example solar energy harvesting, the Thermokon SR04 room sensor relies on an EnOcean ECS 300 solar cell which will deliver 3 V and 4.5 uA give an illumination level of 200 Lux for 6 hours.

Considering the typical performance of standard indoor solar cells, this means that we need to design such sensor system to consume less than 1uA average current. For simplicity reasons we will subsequently calculate based on average current under the assumption that the supply voltage is fixed to 3 V.

To assess the functionality boundaries, we will initially allocate the available energy equally over the three main tasks giving approximately 300 nA of average current each for sensing, wireless transmission and sleep/power losses.

A highly optimized temperature/humidity sensing implementation would require the equivalent of approximately 1 mA current for a period of 10 ms for sensor operation, data exchange between sensor and processor (via I2C or a similar bus) and initial data processing. We can then calculate the maximum number of measurements per day by comparing the available energy per day (300 nA x 86.400 s) with the required energy per measurement (1 mA x 10 ms) and find that the initial energy budget would allow for 2592 measurements per day. Considering that temperature and humidity change only slowly and that we need to conserve energy, we set a rate of one measurement per minute (1440 per day).

Moving to the radio transmission, we assume an average current of 25 mA for formatting and transmission of data at a rate of 125 kBit/s. Based on the available power budget and the required transmission current we calculate the total possible transmission time per day, which equals just over 1 second (or 125,000 Bits/15,625 Bytes) per day. Putting this in relation to the possible number of measurements, we can identify a key constraint of energy harvesting wireless sensors – the radio protocol must be optimized for minimum size. We would need to limit the total telegram length to 10 Byte in order to transmit each measurement result in one radio telegram. From this, it is obvious that both the radio protocol and the amount of transmission need to be highly optimized.

Sensors Insights 2017-04-14 Fig2
The functionality described in this article, for example, has been implemented in the EnOcean STM 330 energy harvesting temperature sensor, which can be extended with the HSM 100 humidity sensor. It measures temperature and humidity every 100 seconds and transmits significant changes immediately. In addition, a regular heartbeat is transmitted every 1000 seconds. All required energy is provided by the EnOcean ECS 300 solar cell.

Energy-optimized protocols

The payload associated with sensors is often small (a few bytes), therefore an optimized protocol must limit the transmission overhead (frame control, preamble, synchronization, error checking) as much as possible while maintaining highly reliable communication.

Standard IP protocol (UDP over IPv6) requires more than 50 Byte of overhead; therefore native IPv6 communication is usually not possible in energy harvesting sensor applications. The power optimized ISO/IEC 14543-3-1X protocol in contrast requires only 12 bytes in total for the transmission of 1 byte of sensor data. Using such protocol in conjunction with an intelligent transmission strategy (e.g. transmission of significant changes only) enables even use of redundant sub-telegrams to increase transmission reliability.


Minimizing sleep losses

Energy harvesting wireless sensors must be in an ultra-low power sleep state for more than 99.99% of the time. Minimizing power consumption in this state is therefore absolutely essential. Considering our design example, we have a total budget of 300 nA, which needs to cover processor consumption in sleep mode (with the ability for timer-based periodic wake-up) as well as losses due to leakage in the energy store. Such low level of power consumption is very difficult to achieve even with the latest processors and is probably the biggest design challenge. Custom mixed signal designs coupled with optimized system architecture is required to address these challenges.


From wired proximity to a wireless world

Today, input data for the IoT is often provided by wired sensors that are locally connected to controllers and actuators. Here, all network components are in close proximity and directly connected with each other. This approach is well suited to local applications with limited flexibility needs where data reuse is not required.

An Internet of Things, in comparison, no longer requires such proximity. It allows centralized or even cloud-based data processing. Thus, the same data can be used for several applications, driving down infrastructure cost and allowing dynamic network structures.

All of these characteristics require a second cloud, consisting of sensor and actuator nodes that can be deployed and expanded flexibly. Nodes that use minimal energy, which they harvest from their surroundings, provide a ‘fit and forget’ solution – they can be installed in the most inaccessible locations and relied upon to execute their task with minimal maintenance or attention.


Technologies For Harvesting Energy

Energy can be harvested from different sources, three of which are commonly used.

1. Kinetic Energy

Kinetic energy in different forms (lateral movement, rotation or vibration) has long been used to generate electrical energy using electromagnetic or piezoelectric harvesters. For most applications, the electromagnetic energy harvester is the better choice as it provides a more stable energy output at a longer lifecycle without ageing effects. These harvesters generally work by changing the magnetic flux through a coil either by moving a magnet relative to the coil or by changing the flux polarity. The latter approach is used in the EnOcean ECO 200 harvester, which can quickly switch magnetic polarity based on lateral movement of its spring. This type of kinetic energy harvesting is the technology of choice for mechanical switches and similar applications.

2. Solar Energy

Many sensor applications are powered by miniaturized solar cells. They are well suited for applications with sufficient illumination (indoor or outdoor) and often used for sensor applications such as temperature, humidity, illumination or CO2 sensors. Energy delivery can be scaled by adjusting the size of the solar cell based on the available space set by the application.

3. Thermal Energy

Temperature differences can be used to generate energy based on Peltier elements. The standard application for these elements is to cool an area (e.g. cooling box) when electrical energy is applied. The reverse effect – generating energy based on temperature differences – is used for thermal harvesting. The output voltage of Peltier elements depends on the temperature difference and is typically very small (e.g. 20mV for 2°C temperature difference). Specialized electronics is therefore required to utilize this energy.

In the SAUTER ecoUnit wireless room operating unit, a temperature sensor powered by a solar cell has been successfully deployed and has enabled increased comfort with significantly reduced energy consumption.
In the SAUTER ecoUnit wireless room operating unit, a temperature sensor powered by a solar cell has been successfully deployed and has enabled increased comfort with significantly reduced energy consumption.

About the author

Neil Cannon is President of EnOcean Inc. Before joining EnOcean, he was the Chief Marketing and Innovation Officer at Terralux Corp. where he started the LEDSENSE program of building and lighting controls. Before Terralux he was a Board member and the EVP of Business Development for Albeo Technologies, which was acquired by GE Lighting in 2012. Prior to Albeo and GE, he was VP of Advanced development at Picolight Corp.

Under Neil´s direction Picolight innovated the first 10G SFP+ fiber optic modules. This innovation led to design-ins at Google and other networking companies. In 2007 Picolight was purchased by JDSU. Earlier in his career Neil held senior management positions at Infineon Fiber Optics, Iolon Corp. and OR technologies.