Designing Ultra-Low-Power Wireless Automation with ZigBee Green Power



New protocols and benchmarks will boost the development of ultra-low-power wireless systems that harvest energy from the environment. The Green Power specification from the ZigBee Alliance allows wireless systems to be built that use low duty cycles and can harvest mechanical, thermal, or solar energy from the surroundings.

A new ZigBee PRO specification called Green Power is now supporting low cost energy harvesting and ultra-long battery life for ZigBee wireless applications for the Smart Home, while an industry group is looking to benchmark ultra-low power devices.

The Green Power feature in ZigBee PRO 2012 gives battery-free, energy-harvesting devices the ability to join any Green Power-enabled ZigBee network. Without requiring batteries, the self-supporting devices typically harvest just enough power themselves to communicate a brief command from mechanical, thermal, or RF energy. For example, a light switch without a battery harvests energy from flipping the switch on or off. This energy is captured to send the “on” or “off” command to the network to switch on or off the appropriate light, and so the self-powered light switch can be located anywhere in the room without a need for running power wires.

However, the discussion on the performance of devices now moves from the current and power consumption to the overall energy consumed by the system. At these low levels of microjoules, there is no standardized way to evaluate different systems, so new benchmarks are also being developed separately that will give engineers key information about the performance of the microcontrollers and wireless systems that are used in these home automation installations.

Adapting the ZigBee protocol is a lower power approach than using other approaches. A typical Wi-Fi card in a laptop PC uses roughly 2 kJ of energy in an hour (assuming a 200 mA, 3 V transceiver runs continuously). In comparison, a 30 W lamp on for an hour consumes 100 kJ, 50 times as much. The ZigBee set-top box remote control using the R4CE protocol uses just 20 mJ per hour, a factor of 100,000 times less than the Wi-Fi card.

Using the Green Power protocol, the light switch that is flipped on or off an average 10 times a day uses around 200 µJ per hour, 100 times less than the ZigBee remote control and about 10 million times less than the Wi-Fi card. This energy was generated itself without having to have an external power source.

One reason for this is that the ZigBee network uses a very robust protocol to ensure high reliability, even in high-interference environments. Implementing that protocol is a challenge for devices that are using harvested energy, since those devices typically only have enough power to communicate a brief command and typically have no other source of power or stored energy.

A light switch without a battery harvests energy from flipping the switch on or off, capturing the energy with a piezoelectric crystal. This energy is used to send an on/off command to the network and control the appropriate light. As a result, the light switch is active on the network for an extremely short time to issue its command and then disappears from the network. For many protocols this can be a major issue. The Green Power protocol makes it possible for the device to send its ID and command code and allow the network to continue to operate flawlessly once the device has disappeared.

One of the first devices to support the new standard, and one of the ‘golden units’ used for testing is a microcontroller from Dutch low-power wireless chip designers GreenPeak Technologies. This is a fully-integrated system-on-chip solution for power harvesting end nodes for light switches, smart home devices, or for applications designed to run on a single battery for many years.

To get the ultra-low power, the controller integrates a radio transmitter, real-time Medium Access Control (MAC) processor, a security engine, and a harvester interface. The ultra-low power consumption ensures that multiple (redundant) packets can be transmitted with a small harvester or tiny battery. The device is designed to support both bursting and trickling energy harvesters, and is therefore well suited to energy harvesting from flipping a light switch as well as solar powered sensors.

The integrated features simplify design complexity. Developers can design low cost light switch products by using simple PCB antennas that require no shielding. No additional components, such as a voltage converter, are required to design a self-powered end node.

“GreenPeak is one of the first ZigBee Alliance members to offer Green Power products, and as Vice Chair of the Green Power working group, GreenPeak has been one of the driving forces and an influential contributor in establishing the Green Power feature,” said Cees Links, founder and CEO of GreenPeak Technologies. “The GP410 enables our customers to quickly develop and easily deploy low-cost, battery-free, self-powered solutions, the ultimate maintenance-free Smart Home end-devices. Green Power nodes can also be powered by a small battery, and due to the ultra-low power consumption, the battery will easily outlast the application’s expected life span. The high level of integration of the GP410 makes it easy to design and develop low-cost light switches and other self-powered Smart Home applications.”

Sink and proxy

An ingenious and simple design allows Green Power devices to complete communication with just 100 to 500 µJ of energy. The design sets up proxies so that each Green Power device is represented as always present on the ZigBee network, even when it is not on or present, avoiding the problem of whether the device is actually on or not. These proxies also handle forwarding and resending packets to ensure reliable communication, especially to destination devices that are not in the direct radio range of the Green Power device.

This proxy functionality is part of the ZigBee 2012 specification, implemented in the software on an existing device on the network. A ZigBee network with Green Power can be as simple as a light switch and a lamp, but with home automation becoming more and more popular, the networks are growing. The mesh approach of ZigBee allows new nodes to be easily added, with the proxies forwarding the data reliably from node to node to support networks of thousands of devices.

Figure 1: The ZigBee green sink.

Figure 2: Combining sinks and proxies in a ZigBee Green Power network.

Reducing the power consumption of the wireless transceiver is a key objective. Atmel has developed a low-power ZigBee transceiver with a static consumption of just 20 nA, a reduction of five over previous devices, coupled with a listening receive mode of just 6 mA and 14.5 mA transmit power at +3 dBm. This comes from using a 1.8 V supply and a 16 MHz 8-bit AVR controller.

Similarly, Silicon Labs is planning a family of highly-integrated, ultra-low-power wireless transceivers that will use even smaller amounts of energy, and will be key for the implementation of Green Power when they are launched later in 2013.

A ZigBee PRO2012 network is ready to support the Green Power protocol; some of its devices implement the Green Power Proxy function and others implement the Green Power Sink functions. The proxy handles the interfacing with the Green Power device and is typically part of an always-on device like a ZigBee PROrouter. The sink function handles the interpretation of the Green Power device commands and is typically implemented in the target device.

Both proxy and sink functionalities need to be present, and they can be implemented in the same device. A good example is a lamp and a light switch. The lamp is the device controlled by the Green Power switch and needs to implement the sink feature, and therefore becomes a Green Power Sink.

The lamp is usually connected to the mains power and always present on the ZigBee network. It has a defined network function and a dedicated power supply, also making it an ideal host for a proxy. This is not a requirement though, as any other always-on device can host the proxy functionality.

The Green Power light switch, however, can be offline with no power, but the ZigBee network still recognizes it because of the proxy present in the network.

Continuing the switch and lamp example, when the Green Power switch becomes active and transmits a command, the command is received by the device implementing the proxy, which could be the lamp, or a proxy and routed to the destination which can be the lamp itself (Figure 1), or another lamp (Figure 2) not in the direct radio range of the switch. This lamp will then interpret the Green Power command and act upon it. The proxy has the steady power source and can process all the required networking interactions required by the ZigBee network while the Green Power device uses all of its energy sending the command.

The mesh capability means the switch does not need to be near the lamp, any ZigBee device implementing the proxy can simply forward the command to the targeted lamp. The lamp can be in another room or another distant location in the network, and the command is handled just like any other ZigBee command.

With the increasing popularity of home automation networks comes an increased need for security. Adding a Green Power device and its sink to an existing network is a simple process based on the existing security protocols in the ZigBee standard.

Continuing to use the light switch and lamp example, ZigBee defines a pairing process that starts when a user triggers the target to begin the pairing process with the switch. Triggering the pairing process can be as simple as pushing a button on the lamp that is considered a sink by the ZigBee network. This triggers all the proxies in the network, allowing them to accept new Green Power devices and creates a pairing window for the switch to connect with the lamp.

The next step in the pairing process occurs when the user triggers the controller, for instance by flipping the switch and initiating a quick information exchange that pairs the switch to the lamp during the pairing window. Once this is completed, the pairing process ends and the lamp can now be controlled by the switch.

Connecting more than one switch to the lamp is also possible during this pairing window. All that is required to connect multiple controllers to a single target device is to repeat the pairing procedure above for each device. There is no pre-set maximum to the number of switches that can control the lamp; however, the amount of memory available in the lamp may limit the maximum number for that device.

A key point is that adding a Green Power device to an existing network does not come at the price of lower security. The Green Power protocol provides the same three levels of security found in the ZigBee PRO2012 specification.

The basic level filters any undesired device trying to communicate on the network using a 32-bit unique device identifier and a rolling frame counter. The intermediate level adds an authentication mechanism based on AES 128 technology and a shared key. This provides the assurance that the device that is communicating is not falsified and is authorized on the network. The highest level of protection adds encryption also based on AES 128 for protection against eavesdropping so that the data exchanged between the device and the network remains confidential.

All of this security is available, as Green Power devices can get the security key using an out-of-band mechanism, or over-the-air from the ZigBee network.

Measuring power consumption

The EEMBC benchmarking organization is also looking at the same ultra-low-power market to find a standard way to quantify the real world power consumption of such devices in applications such as portable medical devices, security systems, building automation, smart metering, and also applications using energy-harvesting devices.

EEMBC already has a non-intrusive energy benchmark that provides data on energy consumed by a processor while running EEMBC’s performance benchmarks, and this applies to all benchmarks provided by EEMBC, which ties performance with energy consumption for specific benchmarks.

Every processor vendor typically has its own power measurement methods, making it nearly impossible to make accurate comparisons among competing vendors. Many processor vendors offer "typical" power specifications on product datasheets that are difficult to compare with one another. The problem of interpreting these values is exacerbated when designers attempt to compare processor cores for system-on-chip implementations. EnergyBench fills this knowledge gap by providing data on how much energy a processor consumes when running a real application workload and not just arbitrarily chosen test vectors.

Besides providing design engineers with comparable information regarding energy consumption, EnergyBench gives designers insights into the power budget "cost" of a device’s performance by allowing a performance/energy number to be derived using the consolidated performance score in each benchmark suite. EnergyBench makes it possible to calculate a "Netmarks per Joule" score for devices tested against EEMBC's networking benchmarks, or a "Telemarks per Joule" score for devices tested against the consortium's telecomm benchmarks.

However, the energy consumption is still too high for energy-harvesting applications, and a new Ultra-Low-Power (ULP) benchmark is being developed.

This will initially look at the power used when a controller moves from running to its real-time clock sleep mode, said Horst Diewald, chief architect of MSP430 Microcontrollers at Texas Instruments (TI) in Germany who is chair of the EEMBC ULP working group.

“We have seen a significant need for a well-constructed, industry-accepted benchmark to equitably evaluate the energy efficiency of microcontrollers,” said Diewald. “Unfortunately, the application developer cannot rely on datasheet parameters alone to compare total microcontroller power consumption and select an appropriate microcontroller.”

Members of the group include Analog Devices, ARM, Atmel, Cypress, Freescale, Fujitsu, Microchip, Renesas, Silicon Labs, STMicroelectronics, and Texas Instruments, although some notable energy-harvesting device developers such as GreenPeak and EnOcean are missing.

The benchmarking methodology will allow the microcontrollers to enter into their idle or sleep modes during the majority of time when they are not executing code, thereby simulating a real-world environment where products must support battery life measured in months, years, and even decades.

Conclusion

The new generation of protocols focusing on ultra-low power is opening up the use of energy-harvesting technologies in smart networks. Through a combination of the developing ULP benchmarking of devices and the ULP-aware ZigBee Green Power protocols, designers can implement innovative wireless networks that bring in new equipment, increasing the controllability and performance of homes and buildings without increasing the power footprint. The new protocol delivers an effective way to implement a secure, reliable, and ultra-low-power wireless network, while reducing power consumption of the coming generation of devices will allow them to be powered by something as simple as the click of a switch.
Supplier