Eeny, Meeny, Miney Moe: choosing a low power wireless network protocol - Part 1 - Embedded.com

Eeny, Meeny, Miney Moe: choosing a low power wireless network protocol – Part 1

One of the biggest trends in the electronics market is adding wirelessfunctionality and connectivity to products. Products as varied aselectric, gas or water meters, home security systems, TV remotecontrols or exercise equipment have added wireless functionality.

This is partly due to convenience for the user such as a wirelessremote control that can operate anywhere in a house as opposed to aline of sight (LOS) infrared remote.

Another driving factor is avoiding the expensive rewiring that wouldbe required to retrofit a home with a security system. Other trendssuch as automated meter reading (AMR)or advanced metering infrastructure(AMI) rely on low power protocols where battery lifetime can bemeasured in years.

In this three part series we will provide an overview of wirelessnetwork protocols, but will focus on ZigBee,802.15.4 and compare it withone example proprietary protocol, SimpliciTI,all designed for low power applications.

In Part 1, here, we will review network basics including commonnetworking terminology, the Open Systems Interconnect (OSI) networkmodel and considerations to keep in mind when deciding on a network fora particular application.

In Part 2, a set of network selection criteria will be presentedupon which the comparison of the three protocols will be framed.Finally, in Part 3 , the 802.15.4, ZigBeeand the proprietary SimpliciTIprotocols will be described in detail and some specific exampleswill be described,  using these selection criteria to determinethe optimum network forparticular applications.

The list of released wireless network protocols is a long one. Thefollowing is just a partial list of network protocols that areavailable:

* Standards-based protocols such as WiFi (802.11b), Bluetooth,ZigBee, ZigBee Pro, 802.15.4, RFID and Wireless USB

* Proprietary protocols from silicon vendors, third parties, or whatthis paper calls “do-it-yourself” networks, built and controlledcompletely by the company that uses the network in their own products.

Figure 1 below shows atable with various protocols, the applications these are focused on,typical resource requirements and so forth. When comparing the lowpower networks such as ZigBee/802.15.4 against other protocols, it isimportant to focus on the key features of a low power wireless networkthat distinguishes it from its peers: low data rates, reduced operatingrange, a low frame overhead, power management considerations builtdirectly into the protocol, and low complexity.

Each design consideration serves as a basis for the end goal ofreducing the power consumption per individual node on the network. Asthe figure shows, this is a perfect fit to the applications withinenvironmental monitoring and control markets in which the devices wouldbe optimally battery powered and be offered at price pointsconsiderably lower than hardware capable of supporting larger, morecomplex protocols.

Figure1–Wireless protocols parameters and focus applications (courtesy ofZigBee Alliance)

Low power networks were designed chiefly to provide wirelessconnectivity between products with battery lifetime in months or years.In most low power wireless systems, the element that uses the mostpower is the radio transmitting and receiving data. It is, therefore,extremely important to minimize the power used by the radio to maximizebattery life.

Since the distance a signal can be transmitted and received is afunction of the power input to the antenna, the wireless network rangebetween individual nodes is typically limited. Wireless protocols alsotend to be less complex to reduce computational overhead and the needfor larger memories, resulting in lower cost.

Network basics
Figure 2 below shows a blockdiagram of a typical low power network node. In general there is asensor (or sensors) collecting data or status and/or providing a userinterface to the system. A microcontroller interfaces the sensor andcontrols the radio (the CC1100 or CC2500 in this particular case) whilethe radio transmits and receives status.

Figure2–a typical low power network block diagram, shown for an automaticmeter reading application

All networks on which these nodes communicate, wired or wireless,can be conceptually described by what is referred to as the NetworkOpen Systems Interconnect (OSI) Basic Reference Model, depicted in Figure 3 below.

Developed in the late 1970's by the International Organization forStandardization (the ISO), this model separates the components of anetwork protocol implementation into software layers. For twoapplications on separate devices to communicate, a message must traveldown from the application layer, across the physical layer, and up theother side. Each layer communicates only with the layers above andbelow it.

Figure3–Elements of a network

One way to conceptualize a layered software architecture cold be bydescribing the mailing of a letter by mail. The letter itself can beconsidered application data. The individual drops it off to be pickedup by the mailman in the post-office box so that it may be transportedto the post office.

The post office separates all mail according to its destinationbefore shipping it, by sea, air and land to its final destination. Forthe letter to arrive to the recipient, it must work its way back up theother side, being dropped off at the post office, sorted bydestination, transported to the recipient's mailbox, to be interpretedby the reader.

This is a four-tiered communication protocol where the individualwriting the letter represents a single layer, the transport to and fromthe local post office represents the second layer, the post officesorting the mail represents the third layer, and the shippingmethodologies represent the fourth.

Each layer concerns itself only with its own task, and communicatesonly with its adjacent layers. Only by working its way down one sideand up the other will the contents of the letter (or application data )be successfully communicated between individuals.

The OSI model services seven separate layers. An application layerservices the direct interface to the user. The presentation layerformats the message to or from a network format, oftentimes implementedas message encryption and/or encoding.

The session layer creates and manages the logical link between anytwo devices on the network. The transport layer is responsible forproviding reliable end-to-end communication. If the transport layerfails too often, perhaps the channel is too noisy or the link is bad,and it should inform the session layer to create a new link between thefailing nodes.

The network layer is responsible for network routing mechanisms,leaving the responsibility of message transmission between individualdevices to the data link layer. The data link layer assurespeer-to-peer delivery of a message, but leaves the actual transmissionof messages across the physical medium to be managed by the physicallayer. And so, a message is transmit down one side of the OSI model andup the other.

A designer has the option to select protocols that have differentnumbers of layers implemented, and can choose to customize theremaining layers to his or her application as needed.

Most networking implementations today do not actually implementthese layers wholesale, and may mix the functionality of certain layersaccording to the requirements of the protocols.

In fact, the OSI model is best to consider as a framework forconceptualizing the complexity and functionality of a protocol'sarchitecture The designer should always be aware of what functionalityhis or her solution implements and which functionality it isconsciously leaving out.

Giving structure to a wirelessnetwork
To give structure to selecting a wireless protocol when decidingbetween ZigBee, 802.15.4, or a proprietary network, the followingcriteria are helpful to keep in mind.

1. Applicationconsiderations
2. Robustness and reliability
3. Ease of use
4. Hardware and RFconsiderations

Application considerations
The initial steps in network design, much like any other system design,are to define the high-level requirements for the application. Thefollowing list captures some of the most important network parametersthat should be defined before considering any wireless protocol as thesolution of choice.

The possible implementations addressing these applicationconsiderations can then be traced throughout the paper, as the detailedexplanation of the selection criteria and the protocols themselvesrequire further elaboration.

1. Network topology
    a. How manynodes and in what general organization within the network does theapplication require?
2. Reliability of communications
    a. Howcritical is the reception of each data packet within the network?
3. Network security
    a. Does thedata need to be secure? If so, how critical is the absolute security ofthe network data transmissions?
4. Customization and designfreedom
    a. Howcustomized does the networking protocol solution need to be to fit theapplication, and does the protocol offer the design freedom to do so?
5. Development time andprotocol complexity
    a. Alongthe same lines of #4, what complexity exists within the protocol inquestion?
    b. How muchtime will be spent learning and integrating the protocol in question tothe existing application versus the time that would be spent developinga more customized solution?
6. Interoperability
    a. Would theend product benefit from interoperability with other vendors, or willit be a completely proprietary solution?

Figure 4 below presents fourof the most common wireless network topologies available forimplementation.

Figure4–low power network topologies

A peer-to-peer network topology is one that supports either uni- orbi-directional links between individual nodes on the network. Nodeswill only communicate if they are within range of each other, as theymust maintain direct physical links for communication; the onlyexception being a broadcast message, which could be re-broadcast andpropagated through the network.

A tree network topology is one each node in the network associatesitself to a parent node, and the network addressing is reflected assuch, much like an IP internet address. This allows for more efficientrouting algorithms to be implemented, as the more significant digits ofa node's network address can identify the node's location in relationto its peers.

A star network identifies a single node as the network coordinator,responsible for a variety of possible network management controls, suchas node associations, network join and linking permissions, messageforwarding, and security exchanges. A star network depends upon thecoordinator to keep the network communicating and is susceptible todisruption if the coordinator node goes down.

A mesh network, in its most general sense, is defined as a networkin which there are at least two pathways to each node and a fullymeshed network means that every node has a direct connection to everyother node in the network.

The latter case somewhat unreasonable in many cases, as this wouldquickly limit the size of a network to the minimum range of the weakestdevice, and the former is perhaps too strict a requirement.

Instead, one tends to see interpretations that exist somewhere inbetween the two cases, where a central node is responsible for startingthe network, and a tree addressing technique is used to locate nodesand manage the associations between them.

Range extenders, or routing nodes, exist to route messagesthroughout the network and if one node or the coordinator goes down,the network can still continue to function, sacrificing some degree ofoperability. Additional features such as self-healing route discoveryand route expiration can make such routing algorithms increasinglyreliable and efficient.

Another important factor to consider may be the financial cost ofusing a certain protocol. It is not uncommon that a membership orroyalty fee must be paid to the organization representing a proprietarysolution.

ZigBee, for example, does not have royalties but does requiremembership in the ZigBee Alliance for a nominal, yearly fee. There canalso be a cost, both in time and money, resulting from a certificationprocess. Silicon vendor proprietary protocols typically require thattheir devices be used in lieu of a royalty.

Robustness and reliability
The extent of the implemented robustness and reliability of a low powerwireless network protocol can be summarized into three categories:message delivery, physical layer considerations, and the messagingprotocol.

Message delivery concerns routing methodologies that assure asuccessful packet transmission and the security of networktransactions. Physical layer considerations address the interference ofnoise or other transmissions within the channel of operation.

The messaging protocol then defines the division of the channel sothat all devices can use the physical medium without packets collidingmid-transmission. All three contribute to the improved quality ofservice (QOS) for a network, defined as a set of network metrics usedto gauge the efficiency, transmission rate and error rate for packagecommunication.

Channel scanning, or the ability to sense the amount of activity ornoise on a channel, is a physical layer consideration that networkprotocol can use to find the channel within the specified frequencyband of operation that is least likely to impede communication betweennodes.

Frequency agility is the ability of a network to change the channelof operation for all nodes on the network, so that if a channel isbombarded with interference operation of the network may continuewithout concern.

Improved message delivery can be achieved through acknowledgementschemes, where the receiving node will transmit an ACK to the originalsender after the successful reception of the packet. Peer to peeracknowledgements, partnered with a defined number of message retrieswill highly decrease the possibility of a packet transaction beinglost.

End to end acknowledgements will provide a second layer of securitythat a packet transaction will not be lost, and are especiallyimportant in large, multi-hop networks supported by complex routingalgorithms.

The messaging protocol defines how the network bandwidth iscontended for or partitioned. Different wireless protocols will definedifferent partitioning of the bandwidth, and the possibilities includedivision in frequency, space, time, or code.

A division in frequency parallels to a room of people talking at adifferent pitch of voice; a division in space parallels to a room ofpeople talking in different directions; a division in time to a roomwith people contending for the right to speak but backing off ifsomeone beats them to speak first; and a division in code to a room ofpeople speaking different languages in all different pitches of voice.

The protocols presented in this paper discuss only the division intime, or Time Division Multiple Access protocol, for which there aretwo possible implementations: synchronous and asynchronouscommunication.

Synchronous communication is enabled by the coordinating nodebroadcasting a periodic network beacon and partitioning the resultingtime in between the beacons into equal-size time slots.

A single network beacon and the time slots that occur before thenext beacon are referred to as the superframe. The time slots of thesuperframe can be further partitioned into an active and inactiveperiod of communications, so that the coordinator can sleep in lowpower modes during the inactive period. Time slots can be guaranteed orcontended for using a Channel Sense Multiple Access (CSMA), orlisten-before-speak algorithm.

A CSMA algorithm defines the protocol that will arbitrate the use ofthe RF channel when multiple nodes are attempting to communicate at thesame time. The most common implementation is a CSMA/CA algorithm, whereCA stands for Collision Avoidance because a transmitting node willavoid the transmission of its message if it senses the channel iscurrently busy.

There are other implementations of the CSMA algorithm, such asCSMA/CD (collision detection), and CSMA/CR (collision resolution), butthey are not commonly found in RF protocol implementations and areoutside the scope of this essay.

Security is also a key concern affecting the robustness of wirelesscommunication and may be the main function of the network. For example,a home security network may include a garage door opener or lock andunlock doors.

These systems need security to prevent eavesdropping, securitybreaches or to maintain privacy. Security can be addressed by multiplelevels of security keying and encryption, message integrity andauthentication, and the use of a trust center, meaning that allsecurity is handled by a single node on the network (usually thenetwork coordinator) rather than a distributed scheme where individuallinks exchange symmetric keys upon link creation and may allow anattacking node entrance to the network without direct authenticationfrom the managing node.

To read Part 2, go to : Additionalcriteria to consider
To read Part 3, go to: ComparingZigbee to TI's SimpliciTI

Miguel Morales is MSP430Applications Engineer and Kevin Belnap is MSP430 Product MarketingManager at Texas Instruments, Inc.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.