The Kanepi system was developed by Twisthink for industrial lighting control in a large warehouse or manufacturing-type configurations. Typically these types of facilities have large rows of overhead lighting with fixtures every 25-40 feet that were traditionally controlled in banks by switches, contactors or circuit breakers. Wireless control with step dimming at the light fixture enables optimization of energy use and decouples the lights from the occupancy sensors and the daylight harvesting sensors. Each light fixture includes a Zigbee device. All data reports are sent to a central concentrator. Control of the system is done using a central controller implemented on an EPIC single board computer running embedded Linux. Additional light in the facility can also be manually requested using touch screens or wireless light switches. The traffic from these switches is also routed through the gateway for processing.
A typical network is from 300 to 1200 lighting fixtures. All lighting fixtures are Zigbee devices connected to an EPIC single board computer running embedded Linux. The EPIC computer can also leverage additional Zigbee-based coordinators using the XTEND product from Digi Corporation.
- Typical data flow from the gateway out to devices in the network is intermittent based on schedule, sensors, and operator action. All routing is done using "many to one" routes into the gateway, and broadcasts from the gateway to devices in the network.
- The gateway updates the "many to one" routes throughout the day.
- Typical latency on the system when lights are triggered from an occupancy sensor is less than 1 second. Standard security is in use with a trust center link key.
- The system operates on Silicon Labs Zigbee software. In-field upgrades of host software have been conducted during the life of these systems using over-the-air bootloaders.
The picture below shows a typical system as seen from the control screen view.
The systems operate in industrial and commercial facilities that also include typical WiFi networks. The building structures are usually metal frames and metal roofing but there are few interior walls. The Zigbee radios are installed as part of the light fixture. The design life of a light fixture is 15-20 years.
In the test network of 800+ lighting control modules, lighting commands are sent approximately 30 times per day. Within the network, 35 unique groups are defined, and each group contains from 5-20 lights. This network is controlled using a combination of scheduled (time of day triggered), reactive (sensor triggered), and manual (operator triggered) events. Each event typically is controlling several groups. After each lighting command, the gateway verifies each light has properly changed state. Any failure in the process generates alerts that are logged and investigated. On average the network processes approximately 3000 lighting commands per day. The network has been running since December of 2007. At the time of this writing, we are not aware of a single communications failure attributed to the PHY/Zigbee layer of the system, with well over 1,000,000 messages logged.
Conclusions and Summary
This data shows the expected reliability of a distributed Zigbee network in a commercial environment. The particular environment is typical for many Zigbee commercial uses and provides a good indicator of expected operation. These networks experience little to no message loss over time indicating a robust and reliable system.
Zigbee wireless devices have been implemented in a variety of different networks for field deployment. These networks utilize different topologies, different numbers and densities of devices, and different traffic patterns. However, each can be analyzed and provide some valuable data on network reliability and robustness. Silicon Labs will continue to publish these case studies documenting such customer networks as an information base for other customers. This case study is intended for Silicon Labs' customers who are designing a Zigbee system and are interested in field experiences and reliability. The specific target audience is system architects and software engineers developing new applications for their Zigbee-enabled products.