Smart Sensors and Sensor Networks Lecture 4 Physical

Smart Sensors and Sensor Networks Lecture 4 Physical layer, MAC and link layer protocols 1

Smart Sensors and Sensor Networks Physical layer: the functions and components of a sensor node that mediate between the transmission and reception of wireless signals and the processing of data, including the higher – level protocol processing; n The physical layer is concerned with modulation and demodulation of digital data; this is carried by the transceivers; n In SNs, the challenge is to find modulation schemes and transceiver architectures that are simple, low cost, but still robust enough to provide the desired service; Wireless channel and communication fundamentals n Wireless channels are an unguided medium; n Frequency allocation: n q In RF based systems, the carrier frequency has to be carefully chosen; the carrier frequency determines the propagation characteristics, such as how well are obstacles penetrated and the available capacity; 2

Smart Sensors and Sensor Networks q The existing frequency bands: q Other solutions are infrared and optical communications: n n q q The infrared spectrum is between wavelengths of 1 mm to 2. 5 μm (300 GHz to 120 THz); The optical spectrum ends at 780 nm (≈ 385 THz); Most of today’s RF based systems work at frequencies below 6 GHz; the allocation of radio frequencies is governed by regulations to avoid unwanted interference; Some systems have special licenses for reserved bands: in Europe, GSM systems use the GSM 900 (880 – 915 MHz) and GSM 1800 (1700 – 1785 MHz) bands; 3

Smart Sensors and Sensor Networks q q There also license free bands, the ISM bands: In the public ISM bands, any system has to coexist with other systems, meaning they have to be robust against interferences: n n q q Coexistence needs to be approached both on the physical and the MAC layer; Requesting allocation of some exclusive spectrum is time consuming; An important parameter in a transmission system is the antenna efficiency, defined as the ratio of the radiated power to the total input of the antenna; the remaining power is dissipated as heat; Small dimensions are required in case of sensors but this is difficult even at high frequencies; 4

Smart Sensors and Sensor Networks n Modulation and demodulation q q q Groups of symbols (data) are mapped to one of a finite number of waveforms of the same finite length; with two different waveforms, a binary modulation results; m – ary modulation; The mapping from a received waveform to symbols is called demodulation; The most common form of modulation is the bandpass modulation: n n n The information signal is modulated onto a periodic carrier wave of comparably high frequency; The spectrum used is typically described by a center frequency fc and a bandwidth B and most of the signal energy can be found in the frequency range: [fc – B/2, fc + B/2]; The carrier is typically represented as a cosine wave, uniquely determined by amplitude, frequency and phase shift; As a consequence modulated signal s(t) can be represented as: s(t) = A(t) cos(ω(t) + Ф(t)) Accordingly, there are 3 fundamental modulation types: ASK, PSK and FSK; 5

Smart Sensors and Sensor Networks n Wave propagation effects and noise: q q Waveforms transmitted over wireless channels are subject to several physical phenomena that distort the waveform; this introduces bit errors at the receiver; The basic wave propagation phenomena are: n Reflection: when a waveform propagating in medium A hits the boundary to another medium B and the boundary layer is smooth, one part of the waveform is reflected back into medium A, another one is transmitted into medium B and the rest is absorbed; n Diffraction: all points on a wavefront can be considered as sources of a new wavefront; if a waveform hits a sharp edge, it can be propagated into another region; 6

Smart Sensors and Sensor Networks n n q Scattering: when a waveform hits a rough surface, it can be reflected multiple times and diffused into many directions; Doppler fading: when a transmitter and a receiver move relative to each other, the waveforms experience a shift in frequency, according to the Doppler effect; if the shift is important, this can cause the receiver to sample signals at wrong frequencies; The signal at the receiver is a superposition of multiple and delayed copies of the same signal; the copies have different relative delays, which translate for each frequency component of the signal into different relative phase shifts at the receiver; it results destructive or constructive interference, leading to fading of the signal; 7

Smart Sensors and Sensor Networks q Path loss and attenuation: n The received power at a distance of d ≥ d 0 m between transmitter and receiver is described by the Friis free – space equation: n For environments other than free space, the model is generalized: Prcvd(d) = Prcvd(d 0) x (d 0/d)γ, where γ is the path – loss exponent ε [2, 6] The path loss is defined as the ratio of the radiated power to the received power, Ptx/Prcvd(d) and is expressed in decibel as: PL(d)[d. B] = PL(d 0)[d. B] + 10γlog 10(d/d 0) Conclusions: n n q q q The received power depends on the frequency: the higher the frequency, the lower the received power; The received power depends on the distance according to a power of law; for example, assuming a path – loss exponent of 2, a node at a distance of 2 d to some receiver must spent 4 times the energy of a node at distance d to the same receiver, to reach the same level of received power Prvcd; Higher frequencies or larger distances must be compensated by an appropriate increase in transmitted power to maintain a specified Prcvd(d) value; 8

Smart Sensors and Sensor Networks 9

Smart Sensors and Sensor Networks n q Attenuation is determined by the media (for example fog or rain); it is frequency – dependent; usually is accounted for in the path – loss exponent; Noise and interference: n Interference refers to the presence of any unwanted signals from external sources, which obscure or mask a signal; q q q n n Multiple – access interference: when the signals come from transmitters sending in the same band at the same time; Co – channel interference: the interference sources radiates in the same or in an overlapping frequency band as the transmitter and receiver under consideration; Adjacent – channel interference: the interferer works in a neighboring band; The noise (or thermal noise) is caused by thermal motions of electrons in any conducting media; it is typically measured by the single – sided noise Power Spectral Density, N 0, given by: N 0 = K x T [watts/hertz], where K = Boltzmanns constant (≈ 1. 38 x 10 -23 J/K), T = temperature in Kelvin degrees; The thermal noise is additive, that is, the received signal r(t) is a sum of the transmitted signal s(t), arrived at the receiver after path loss, attenuation, scattering and so forth, and the noise signal, n(t): r(t) = s(t) + n(t) 10

Smart Sensors and Sensor Networks q Symbols and bit errors: n n The symbol/ error probability depends on the modulation scheme and on the ratio of the power of the received signal, Prcvd, to the noise and interference power; when only thermal noise is considered, the ratio is (in decibels): N 0 is the noise power and Prcvd is the average received signal power; When other sources of interference are considered too: Ii is the power received from the i – th interferer; The SINR describes the power that arrives at the receiver and is related to the symbols sent over the channel; in the end, the data bits are relevant; to correctly demodulate and decode an arriving bit, the energy per such a bit Eb in relation to the noise energy N 0 is relevant: The bandwidth efficiency, ηBW = R/W, is a measure of a modulation scheme’s efficiency: 11

Smart Sensors and Sensor Networks n Spread – spectrum communications: q q In spread – spectrum systems, the bandwidth occupied by the transmitted waveforms is much larger than what would be really needed to transmit the given user data; The user signal is spread at the transmitter and dispread at the receiver; By using a wideband signal, the effects of narrowband noise/ interference are reduced; Spread – spectrum systems offer an increased robustness against multipath effects but the receiver operation is more complex; n There are 2 types of spread – spectrum communications: q q Direct Sequence Spread Spectrum: the transmission of a data bit is replaced by transmission of a finite chip sequence c = c 1 c 2…cn or /c 1/c 2…/cn with ci ε {0, 1}; Frequency Hopping Spread Spectrum: the available spectrum is divided into a number of equal – sized sub bands or channels (for example Bluetooth divide the spectrum in the 2. 4 GHz range in 78 sub bands 1 MHz wide; the user data is always transmitted within one channel at a time; all nodes in a network hop synchronously through the channels according to a prespecified schedule; different networks can share the same geographic area by using no overlapping hopping schemes; 12

Smart Sensors and Sensor Networks n Packet transmission and synchronization: q q q The MAC layer uses packets or frames as the basic unit of transmission; The receiver must know certain properties of an incoming waveform to understand it and to detect a frame: synchronization becomes necessary due to the drift of the oscillators; To compensate this drift, the receiver has to learn about the frequency or time base of the transmitter; the receiver has to extract synchronization information from the incoming waveform; Frames must be equipped with a training sequence that allows the receiver to learn about the parameters of the transmitter; the training sequence can be placed at the beginning of frames or in the middle; After the receiver has successfully acquired initial synchronization from the training sequence, it enters a tracking mode, continuously readjusting its local oscillator; Synchronization is needed at: carrier (frequency correction bursts are used), bit/symbol (the coding technique is important) and frame (frame bounds must be detected) levels; 13

Smart Sensors and Sensor Networks Physical layer and transceiver design considerations in WSNs: n Some of the most crucial points influencing PHY design in WSNs are: q Low power consumption; n n q q q n Consequence 1: small transmit power and thus a small transmission range; Consequence 2: low duty cycle; most hardware should be switched off or operated in a low power standby mode most of the time; Low data rates (tens to hundreds kb/s); Low implementation complexity and costs; Low degree of mobility; A small form factor for the overall node; Low cost; Energy usage profile: q q The radiated energy is small but the overall transceiver consumes much more energy than is actually radiated; for ex. for the Mica motes, 21 m. W are consumed in transmit mode and 15 m. W in received mode for a radiated power of 1 m. W; For small transmit powers the transmit and receive modes consume more or less the same power; therefore it is important to put the transceiver into sleep state instead of idle state; 14

Smart Sensors and Sensor Networks q q n This rises the problem of startup energy/ startup time which a transceiver has to spend upon waking up from sleep mode, for example, to ramp up phase – locked loops or voltage – controlled oscillators; during this startup time, no transfer of data is possible; for example, the μAMPS-1 transcveiver needs 466 μs and a power dissipation of 58 m. W; therefore, going into sleep mode is unfavorable when the next wakeup comes fast; Computation is cheaper than communication: the ratio is hundreds to thousands of instructions/ 1 transmitted bit; Choice of modulation scheme: q q The choice of modulation scheme depends on several aspects, including technological factors, packet size, target error rate and channel error model; The power consumption of a modulation scheme depends much more on the symbol rate than on the data rate; it leads to desire of high data rates at low symbol rates which ends to m – ary modulation schemes; trade – offs: q q q M – ary modulation schemes require more hardware than 2 – ary schemes; M – ary modulation schemes require for increasing m an increased Eb/N 0 ratio; Generally, in WSN applications most packets are short; for them, the startup time dominates overall energy consumption making the other efforts irrelevant; 15

Smart Sensors and Sensor Networks q n Dynamic modulation scaling is necessary; Antenna considerations: q q q The small form factor of the overall sensor restricts the size and the number of antennas; If the antenna is much smaller than the carrier’s wavelength, it is hard to achieve good antenna efficiency and transmitted energy must increase; In case of multiple antennas, they should be spaced apart at least 40 – 50% of the wavelength used to achieve good effects; for ex. for 2. 4 GHz, a spacing of 5 – 6 cm between the antennas is necessary, which is difficult to be accepted; Radio waves emitted from antennas close to the ground, typical in some applications, are faced with higher path – loss coefficients than the common value of α = 2; a typical value, considering the obstacles too, is α = 4; Nodes randomly scattered on the ground, deployed from an aircraft, will land in random orientations, with the antennas facing the ground or being otherwise obstructed; this can lead to nonisotropic propagation of the radio wave, with considerable differences in the strength of the emitted signal in different directions. 16

Smart Sensors and Sensor Networks MAC layer protocols n n The fundamental task of any MAC protocol is to regulate the access of a number of nodes to a shared medium in such a way that certain application – dependent performance requirements are satisfied; For WSNs, the main requirement is energy efficiency and there are MAC – specific sources of energy waste to consider: overhearing, collisions, overhead and idle listening; protocols address one or more of these issues; Within the OSI reference model, the MAC is considered as part of the Data Link Layer (DLL) but it is a clear division of work between the MAC and the remaining parts of the DLL; The remaining part of the DLL approaches error and flow control: q q Error control ensures correctness of transmission and to take appropriate actions in case of errors; Flow control regulates the rate of transmission to protect a slow receiver from being overwhelmed with data; 17

Smart Sensors and Sensor Networks Fundamentals of MAC protocols n Requirements and design constraints for wireless MAC protocols: q Traditionally, the most important performance requirements for MAC protocols are throughput efficiency, stability, fairness, low access delay, low transmission delay, low overhead; n q The operation and performance of MAC protocols is heavily influenced by the properties of the underlying physical layer; n n q The overhead can result from per – packet overhead, collisions or from exchange of extra control packets; The problems of wireless transmission influences the MAC protocols; one problem is the time – variable and sometimes high error rates, which is caused by physical phenomena like fading, path loss, attenuation and noise; Usually the bit error rates can vary between 10 -3 and 10 -2; The fact that the received power Prcvd decreases with the distance between transmitting and receiving node, combined with the fact that any transceiver needs a minimum signal strength to demodulate signals, leads to a maximum range that a sensor node can reach with a given transmit power; 18

Smart Sensors and Sensor Networks q If 2 nodes are out of reach, they cannot hear each other; this rises the hidden – terminal/ exposed – terminal problems: n n The problems occur for the class of Carrier Sense Multiple Access (CSMA) protocols, where a node senses the medium before starting to transmit a packet; if the medium is found to be busy, the node defers its packet to avoid a collision and a subsequent transmission; Simple CSMA in a hidden – terminal scenario leads to needless collisions; Simple CSMA in an exposed – terminal scenario leads to needless waiting; The solutions are: busy – tone solutions and the RTS/CTS handshake used in the IEEE 802. 11 WLAN standard; 19

Smart Sensors and Sensor Networks q q q On the wired media, it is often possible for the transmitter to detect a collision at the receiver and to abort packet transmission; this is called collision detection (CD); collision detection protocols are not applicable in wireless media; Another problem arises when there is no dedicated frequency band allocated to a WSN and it has to share the spectrum with other systems; for ex. the ISM bands are used by several systems, such as IEEE 802. 11/ IEEE 802. 11 b WLANs, Bluetooth and IEEE 802. 15. 4 WPANs; therefore, WSNs must coexist with other systems; The design of MAC protocols depends on the expected traffic load patterns: n n If a WSN is deployed to continuously observe a physical phenomenon, for example, the time – dependent temperature distribution in a forest, a continuous and low load with a significant fraction of periodic traffic can be expected; If the goal is to wait the occurrence of an important event and to report immediately as much data as possible, the network is close to idle for a long time and then is faced with a bulk of packets that are to be delivered quickly; a high MAC efficiency is desirable during these overload phases; 20

Smart Sensors and Sensor Networks n Important classes of MAC protocols: they can be divided in: q Fixed assignment protocols: n n q The available resources are divided between the nodes for long term and each node can use its resources exclusively without the risk of collisions; To account for changes in the topology, signaling mechanisms are needed to renegotiate the assignment of resources; Demand assignment protocols: n The exclusive allocation of resources to nodes is made on a short – term basis, typically the duration of a data burst; they are divided in centralized and distributed protocols; q q q In centralized protocols a central node receives requests and allocates resources; A distributed protocol example is the IEEE 802. 4 Token Bus; Random access protocols: n The nodes are uncoordinated and the protocols operate in a fully distributed manner; for example, in the ALOHA protocol, a node wanting to transmit a new packet does it immediately, accepting the risk of collisions at the receiver; the receiver is required to send an immediate acknowledgement for a properly received packet; if the transmitter does not receives the ack. it tries a retransmission after a random time; 21

Smart Sensors and Sensor Networks MAC protocols for WSNs n Requirements: q q q New requirements are imposed by WSNs, the main one being the energy efficiency; Typical performance figures like fairness, throughput or delay have a minor role in SNs; fairness is not important since the nodes in a WSN do not represent individuals competing for bandwidth, but they collaborate to achieve a common goal; Other important requirements: n n n Scalability: is obvious when considering very dense sensor networks with dozens or hundreds of nodes in mutual range; Robustness against frequent topology changes: are caused by nodes powering down temporarily to replenish their batteries by energy scavenging, mobility, deployment of new nodes or death of existing nodes; Low complexity operation: q q Sensor nodes are simple, cheap and have limited hardware resources; therefore computational expensive operations like complex scheduling algorithms should be avoided; Very tight time synchronization would require frequent resynchronization of neighboring nodes, which can consume significant energy; 22

Smart Sensors and Sensor Networks n Energy problems on the MAC layer: q q Transmitting is costly, receive costs often are as transmit costs, idling can be cheaper but also about as expensive as receiving and sleeping costs almost nothing but results in a deaf node; The following energy problems are related to MAC protocols: n n Collisions: incur useless receive costs at the destination node, useless transmit costs at the source node and expending energy upon packet retransmission; collisions should be avoided either by design or by appropriate collision avoidance/ hidden – terminal procedures in CSMA; however, for applications with low load, the collisions are a minor problem; Overhearing: the wireless medium is broadcast type and all the sources’ neighbors that are in receive state hear a packet and drop it even if it is not destined to them; they overhear the packet; overhearing costs energy; overhearing can be desirable, for example, when collecting neighborhood information or estimating the current traffic load for management purposes; Protocol overhead: is induced by MAC – related control frames (RTS and CTS packets, request packets in some protocols, per – packet overhead etc. ); Idle listening: a node in idle state still consumes energy; switching off is a solution; since mode change costs energy, the rate of change is important; 23

Smart Sensors and Sensor Networks n Low duty cycle protocols and wakeup concepts: q q q Low duty cycle protocols try to avoid spending much time in the idle state and to reduce the communication activities of a sensor node to a minimum; In an ideal case, the sleep state is left only when a node is about to transmit or receive packets; In several protocols, a periodic wakeup scheme is used; one flavor is the cycled receiver approach: n n A node A listens onto the channel during its listen period and goes back into sleep mode when no other node communicates with it; A potential transmitter B must know about A’s listen periods and send its packet at the right time; this is a so-called rendezvous; A rendezvous can be implemented by letting node A to send a beacon at the beginning of its listen period or letting node B to send frequent request packets until one of them is sensed by node A; If node A wants to send a packet, it must also know the target’s listen period; 24

Smart Sensors and Sensor Networks q q q A wakeup period = sleep period + listen period; The node’s duty cycle = listen period/ wakeup period; Observations: n n q By choosing a small duty cycle, the transceiver is in sleep mode most of the time, avoiding idle listening and conserving energy; By choosing a small duty cycle, the traffic directed from neighboring nodes to a given node concentrates on a small time window (the listen period) and in heavy load situations significant competition can occur; Choosing a long sleep period induces a significant per – hop latency, since a prospective transmitter node has to wait an average of half a sleep period before the receiver can accept packets; in the multihop case, the per – hop latencies add up and create significant end – to – end latencies; Sleep phases should not be too short lest the start – up costs outweigh the benefits; In other protocols, there is also a periodic wakeup but nodes can both transmit and receive during their wakeup phases; when all nodes have their wakeup phases at the same time, there is no need for a rendezvous; 25

Smart Sensors and Sensor Networks q Wakeup radio concepts: n n n The ideal situation is to avoid idle state; A wakeup receiver is necessary: it does not need power but can detect when a packet starts to arrive; for example it suffices for it to raise an event to notify other components of an incoming packet; upon such an event, the main receiver can be turned on and perform the reception of the packet; The wakeup radio concept tries to attend the ideal situation by using the wakeup receiver idea; One of the proposed MAC protocol assumes the presence of several parallel data channels, separated either in frequency (FDMA) either in codes (CDMA); a node wishing to transmit a data packet randomly picks one of the channels and performs a carrier – sensing operation; if the channel is busy, the operation is repeated; after a certain number of tries the node backs off for a random time and starts again; if the channel is idle, the node sends a wakeup signal to the receiver indicating also the channel to use; the receiver wakes up its main data receiver, tunes to the indicated channel and data transfer can proceed; afterwards, the main receiver is sent back to its sleep mode; Advantages: q q Only the low – power wakeup transceiver has to be switched on all the time; The scheme is naturally traffic adaptive; the MAC is more and more active as the traffic load increases; 26

Smart Sensors and Sensor Networks n Disadvantages: q q q Difficult hardware solution for such an ultralow power wakeup receiver; The range of the wakeup radio and the data radio should be the same; if the range of the wakeup radio is smaller than the range of the data radio, possibly not all neighbor nodes can be woken up; if the range of the wakeup radio is significantly larger, there can be a problem with local addressing schemes: these schemes do not use globally or networkwide – unique addresses but only locally unique addresses, such that no node has two or more one – hop neighbors with the same address; on the other hand, a node’s MAC address should be unique within its two – hop neighborhood; since the packets exchanged in the neighbor discovery phase have to use the data channel, the two – hop neighborhood as seen on the data channel might be different from the two – hop neighborhood on the wakeup channel; This schemes critically relies on the wakeup channel’s ability to transport useful information like node addresses and channel identifications; this might not always be feasible for transceiver complexity reasons and additionally requires methods to handle collisions and transmission errors on the wakeup channel; if the wakeup channel does not support this feature, the transmitter wakes up all its neighbors when it emits a wakeup signal, creating an overhearing situation for most of them; if the transmitting node is about to transmit a long data packet, it might be worthwhile to prepend the data packet with a short filter packet announcing the receiving node’s address; all the other nodes can go back to sleep mode after receiving the filter packet; instead of using an extra packet, all nodes can read the bits of the data packet until the destination address and if this address is not identical with its own address, the node can go back into sleep mode; 27

Smart Sensors and Sensor Networks n Contention – based protocols q q q A given transmit opportunity towards a receiver node can in principle be taken by any of its neighbors; If only one neighbor tries its luck, the packet goes through the channel; if several neighbors try their luck, these have to compete with each other and in unlucky case, for example due to hidden – terminal situations, a collision might occur, wasting energy for both transmitter and receiver; Contention – based protocols: n n q Based on a periodic wakeup scheme; Without idle listening avoidance and without restrictions as to when a node can receive a packet; Contention – based protocols are appropriate in case of a network that is idle for long times and starts to become active when triggered by an important external event; upon the triggering event, all nodes wish to transmit simultaneously, potentially creating lots of collisions; if the nodes want to send their packets periodically, the danger of collision is repeated; 28

Smart Sensors and Sensor Networks n Schedule – based protocols q Advantages: n n q They do not explicitly address idle listening avoidance but do so implicitly, for example schemes that explicitly assign transmission and reception opportunities to nodes and let them sleep all other times; Transmission schedules can be computed such that no collisions occur at receivers and hence no special mechanisms are needed to avoid hidden – terminal situations; Disadvantages: n n The setup and maintenance of schedules involves signaling traffic, especially when faced with variable topologies; If a TDMA variant is employed, time is divided in small slots and both transmitter and receiver have to agree to slot boundaries to actually meet each other to avoid overlaps with other slots, which would lead to collisions; maintaining time synchronization involves extra signaling traffic; for cheap sensors, having cheap oscillators, it is expected large drifts leading to frequent resynchronization; Such schedules are not easily adapted to different load situations; it is difficult for a node to give up unused time slots to its neighbors; The schedule of a node may require a significant amount of memory; 29

Smart Sensors and Sensor Networks Link – layer protocols n The DLL has the task of ensuring a reliable communication link between neighboring nodes; n The DLL sits on the top of the packet transmission and reception service offered by the MAC layer and offers its services to the network layer and other higher layers; n Tasks: q q Error control: the effect of errors must be compensated; the efficiency and energy consumption of the different error – control mechanisms depends on the error patterns on the link; Framing: user data is formatted in packets or frames; the format and size of packets can have significant impact on performance metrics like throughput and energy consumption; Flow control: introduces signaling to let the transmitter slow down transmission when the receiver is not able to accept packets; it is not an issue in SNs; Link management: it involves discovery, setup, maintenance and teardown of links of neighbors; link quality; 30

Smart Sensors and Sensor Networks n Error control: q q q The data transport service provided by a link layer can be characterized in terms of the following attributes: error – free, in – sequence, duplicate – free and loss – free; The information that the receiver node’s link layer delivers to its users should contain no errors, the transmitted bits should be reproduced exactly; Errors are outlined by bit errors and packet losses; properties in WSNs: n n q Both are bursty, that is they tend to occur in clusters with error – free periods between the clusters; the empirical distribution of clusters and error – free periods have a large coefficient of variation; The error behavior even for static transmitter and receiver is time varying, and in the instantaneous bit – error rates can be sometimes quite high, 10 -4 – 10 -2; the same is true for packet – loss rates, which can reach values beyond 50 %; ARQ (Automated Repeat Request) techniques n The transmitter node’s link layer accepts a data packet, creates a link – layer packet by prepending a header and a checksum, and transmits this packet to the receiver; the receiver checks the packet’s integrity with the help of the checksum and provides feedback to the transmitter regarding the success of packet transmission; on receiving negative feedback, the transmitter performs a retransmission; 31

Smart Sensors and Sensor Networks n n There are 3 standard ARQ protocols; Alternating bit (Send and Wait): q q q q n The transmitter buffers one packet, sends it, and sets a timer; The receiver either receives the packet and sends a positive ack or nothing is received and the receiver keeps quiet or transmits a negative ack; In case of a positive ack the transmitter’s buffer is freed and the next packet is loaded Otherwise, the packet is retransmitted; The transmitter stamps each new packet with sequence numbers alternating between 0 and 1; retransmitted packets are copies of the original packet and have thus the same sequence number; The sequence numbers allow the receiver to detect duplicates, which result if the positive ack, not the data packet, is lost; It can provide loss – free, duplicate – free and in – sequence delivery of data; Goback N: q q Alternating bit is inefficient in case when multiple packets are in transit on the same link; Goback N allows the transmitter to have multiple unacknowledged frames; The transmitter keeps a buffer for up to N packets, called its window; Each packet in the window has its own timer, started upon the packet’s transmission; The receiver accepts frames only in sequence and drops frames that are correctly received but do not have the expected sequence number, typically because some previous frame had been lost; 32

Smart Sensors and Sensor Networks q q q n Selective Repeat/ Selective Reject: q q n n Therefore, the receiver only needs buffer space for a single frame; One common strategy for acknowledgements is to let the receiver always acknowledge the last packet arrived in sequence; If at the transmitter the timer for the oldest frame expires because the corresponding acknowledgement has not been received, this frame and all other frames in the window are retransmitted; Unlike to Goback N protocol, in Selective Repeat the receiver has also N buffers and uses them to buffer frames arriving out of sequence; To achieve in – sequence delivery of data to the user, the receiver keeps out – of – sequence packets in the data buffer until the missing packets have arrived; The receiver can use both positive and negative acknowledgments; The transmitter retransmits only those packets for which no acknowledgment has been received within the timeout period; Send and Wait and Selective Repeat have the important property that only erroneous packets are retransmitted while Goback N potentially retransmits correctly received packets, which is a waste of energy; In practice, often, the number of retransmissions/ packet is bounded to avoid spending too much energy in hopeless cases; in this case, a loss – free service can not be guaranteed; such a protocol is also said to be semi reliable; 33

Smart Sensors and Sensor Networks q FEC (Forward Error Correction) techniques: n The transmitter accepts a stream or a block of user data bits or source bits, adds suitable redundancy and transmits the result toward the receiver; n Depending on the amount and structure of the redundancy, the receiver might be able to correct some bit errors; FEC can be used as an open loop technique, meaning there is no feedback from the receiver; accordingly, the transmitter uses the same coding method all the time; The lack of feedback is energy saving, since feedback is usually provided through acknowledgement packets; these would require the transmitting node to switch its transceiver into receive mode and wait for the acknowledgement; therefore both the reception costs (for the acknowledgement) and the costs for the receiver turnaround are saved; Furthermore, since typically the data packets in WSNs tend to be small, the acknowledgement packets make up a significant share of the total energy to transmit a packet; n n n 34

Smart Sensors and Sensor Networks q Power control: n n q The reliability of packet transmission over a link can be increased by increasing the radiated output power of the transmitter; Increasing this power increases the energy per bit, Eb/N 0 and the SNR, thus decreasing the need for retransmissions; It was shown that for a single – hop ad hoc network there is an optimal transmit power (equivalent to a BER target) balancing the radiated energy and the need for retransmissions for a given packet length; In larger networks, with multihop communication, things are different; if one node increases its transmit power, it also increases the interference seen by other nodes and thus effectively the bit error rates; Error concealment: n n n The idea is to not correct all transmission errors but to live with them to some extent and to take other measures to let the influence of errors disappear for the application; This relaxes the reliability requirements and energy consumption, but at the price of higher computational efforts; Error concealment is not primarily a link layer technique but needs to incorporate application information; 35

Smart Sensors and Sensor Networks Link management n Link quality: q It can be expressed in terms of packet loss rates; features: n n q For a given transmit power, there is no deterministic relationship between distance and link quality; nodes at the same distance from the transmitter can experience widely varying packet loss rates; The region around a node having a certain packet loss rate does not have the shape of a circle, but is irregular; There is a significant degree of asymmetric links; in such a link, packets sent from node A to node B are received by B with few losses but conversely A receives B’s packets with much higher loss probability; the fraction of asymmetric links grows with the distance, taking values between 5 and 15% of all links; The packet loss rate is time variable even when the neighbors in question are stationary; although the mean loss rate for a given distance over time is more or less fixed, there can be significant short – term variations; If a node wants to estimate the quality of a link toward a neighboring node, it has to do so by receiving packets from the neighbor and computing loss rates; because of the variability, it is not sufficient. 36