TSS: An Energy Efficient Communication Scheme for Low Power Wireless Networks

Transcription

1 TSS: An Energy Efficient Communication Scheme for Low Power Wireless Networks Rabindranath Ghosh St.Thomas College of Engg. and Tech., Kolkata, India Koushik Sinha Honeywell Tech. Solutions Bangalore, India sinha Debasish Datta Dept. of EECE IIT Kharagpur, India Bhabani P. Sinha ACM Unit, Indian Statistical Institute, Kolkata, India Abstract Wireless sensor networks typically require low cost devices and low power operations. Hence, such networks usually employ radios with only simple modulation techniques such as ASK, OOK and FSK [10]. We propose a new energy efficient communication scheme for wireless sensor networks that is based on the ternary number system encoding of data. An efficient algorithm for conversion from binary to ternary and vice versa is used that does not involve any division or multiplication but only addition. Our Ternary with Silent Symbol (TSS) communication scheme is similar in concepts to the RBNSiZeComm protocol proposed in [1] [3]. However, in contrast to RBNSiZeComm, TSS simultaneously saves energy at both the transmitter and receiver due to shortening of the transmission duration. We also propose a transceiver design that uses a hybrid modulation scheme utilizing FSK and ASK so to keep the cost/complexity of the radio devices low. With a non-coherent detection based receiver and assuming equal likelihood of all possible binary strings of a given length, there is a savings of about 20% in energy on an average at the transmitter compared to binary FSK, for additive white gaussian noise (AWGN) channels. Simultaneously, there is a savings of 36.9% at the receiver too, due to the reduced length of ternary encoded transmitted data, compared to binary encoding. Coupled with the low cost and low complexity of the transceiver, these results establish the effectiveness of TSS as a suitable candidate for communication in low power wireless sensor networks. Index Terms Energy-efficient communication, wireless networks, ternary encoding, silent communication, sensor networks. I. INTRODUCTION Wireless sensor networks (WSNs) typically utilize highly energy constrained, low cost sensor devices that are deployed (often in an ad hoc manner) in areas that are difficult to access and with little or no network infrastructure. In most scenarios, such battery powered sensor devices are expected to operate over prolonged periods of time. Communication being a major source of power drain in such networks, energy efficient communication protocols that can be implemented with low hardware and software cost/complexity are thus of paramount importance in WSNs to reduce the device recharging cycle periods and hence provide connectivity for longer durations at a stretch. In practice, most existing transmission schemes not only utilize non-zero voltage levels for both 0 and 1 so as to distinguish between a silent and a busy channel, they also keep both the transmitter and the receiver switched on for the entire duration of the transmission of a data frame. Communication strategies that require energy expenditure for transmitting both 0 and 1 bit values are known as energy based transmission (EbT) schemes. In other words, if the energy required per bit transmitted is e b, the total energy consumed to transmit an n-bit data would be n e b. Most current research efforts on reducing energy consumption have focussed on the MAC layer design [11], [12], [18], [20], optimizing data transmissions by reducing collisions and retransmissions [13], [14], [21] and through intelligent selection of paths or special architectures for sending data [10], [12], [15] [17], [20]. In all such schemes, the underlying communication strategy of sending a string of binary bits is energy based transmission. In contrast to EbT based communication schemes, a new communication strategy called Communication through Silence (CtS) was proposed in [4] that involves the use of silent periods as opposed to energy based transmissions. CtS, however, suffers from the disadvantage of being exponential in communication time. An alternative strategy, called Variable- Base Tacit Communication (VarBaTaC) was proposed in [5] that uses a variable radix-based information coding coupled with CtS for communication. However, for an n-bit binary string, the duration of transmission is in general significantly longer than n. Neither [4] nor [5] talk about the amount of energy saved by CtS and VarBaTaC for noisy channels and considering real-life device characteristics. Borrowing from the concepts of CtS and VarBaTaC, [2], [3] proposed a new energy efficient scheme called RBNSiZeComm for wireless sensor networks that recodes a binary coded data using a redundant radix based number representation and then uses silent periods to communicate the bit value of 0. The authors in [2], [3] showed that by using the redundant binary number system (RBNS) that utilizes the digits from the set {-1,0,1} to represent a number with radix 2, it is possible to significantly reduce the number of non-zero digits that need to be transmitted. They showed that theoretically, for a noiseless channel RBNSiZeComm provides on an average, 1 n +2 4n savings in transmission energy. Considering different application data types, for noiseless channels and utilizing the device characteristics of some commercial radios used for low data rate, low power applications, their simulation results showed a savings of about 69% on an average at the transmitter. [1] presents a non-coherent detection based transceiver design for the RBNSiZeComm protocol and shows that the energy /08/$ IEEE 85

2 savings is about 41% on an average at the transmitter, for noisy channels. Their simulation results show that for different applications data types, RBNSiZeComm can extend the battery life of radio devices from about 33% to 61% on an average for noisy channels, with their proposed transceiver design. However, no savings is generated at the receiver by their proposed scheme. A. Our Contribution One interesting open issue related to the RBNSiZeComm protocol proposed in [2], [3] was whether a recoding of binary data to higher radix such as ternary and then using the silent symbol strategy from [2], [3] would yield comparable savings in energy. In this work, we propose a new communication scheme based on recoding data from binary to the ternary radix and the silent symbol strategy, with the aim of generating energy savings simultaneously at the transmitter and the receiver. An efficient algorithm for conversion from binary to ternary and vice versa is used that does not involve any division or multiplication but only addition. A comprehensive analysis of the energy-efficiency of a communication strategy needs to address the MAC protocols along with adequate physical layer awareness. However, we do not address the issues related to MAC protocols in this work and make an effort to improve the energy efficiency with physical-layer centric approach. We study the performance of the proposed ternary with silent symbol (TSS) protocol in the context of low power wireless networks for noisy channels with a new transceiver design. In order to understand the energy-efficiency aspect of the proposed scheme (TSS combined with ASK and PSK/FSK), we therefore examine first the candidate source coding scheme, i.e., TSS. Thereafter, we estimate the energy spent in the baseband electronic circuits (device consumption) followed by the evaluation of the energy consumed during RF transmission for a given BER specification with appropriate analytical models. While there are numerous coded modulation techniques in the literature [22], applications requiring low power and low cost radio devices (e.g., sensor networks) usually employ radios with only simple modulation techniques such as ASK, OOK and FSK [10]. Our proposed transceiver design uses a hybrid modulation scheme utilizing FSK and ASK to keep both the cost and complexity low. With a non-coherent detection based receiver and assuming equal likelihood of all possible binary strings of a given length, there is a 20% savings in energy on an average at the transmitter compared to binary FSK, for additive white gaussian noise (AWGN) channels. In addition, there is a savings of about 36.9% in battery energy at the receiver resulting from a reduction in the length of the transmitted data (and hence transmission duration) as a result of recoding to ternary radix from binary data. Thus, while the energy savings generated by TSS at the transmitter are inferior to that of RBNSiZeComm (which is around 41% on an average for noisy channels [1]), TSS scores over RBNSiZeComm in saving energy at the receiver too. These results hence demonstrate that TSS may be useful for low power wireless networks, particularly for multi-hop communications. II. BASIC IDEA AND ALGORITHM DESCRIPTIONS Let a given binary message B be represented by an n-bit binary string b n 1 b n 2...b 2 b 1 b 0. Let T be its equivalent m- digit ternary representation given by T = t m 1 t m 2...t 2 t 1 t 0. We assume that n is even, otherwise we pad the message with a 0 bit at the leftmost (msb) position. For conversion from binary to ternary, we successively scan every two bits of the binary message starting from its msb position and then convert the leftmost part of the binary message (starting from its msb to the currently scanned bit position) to its equivalent ternary representation. To do this, we replace the currently scanned two bits (bit-pair) by either a single ternary digit (0, 1 or 2 depending on whether the binary bit-pair is 00, 01 or 10), or by two ternary digits (10) 3 (if the binary bit-pair is 11), and then add it to four times the so far obtained ternary number T from all the previous bit positions (on the left of this bit-pair). Note that this addition will be in ternary number system, and the multiplying factor four is to adjust the weight of T with respect to the currently scanned bit-pair. Also, since this addition will be in ternary number system, the weight of four can be assigned to T by adding the two ternary numbers T 0 and 0T. Thus, the equivalent ternary representation of the part of the binary message from its msb to the currently scanned bit-pair will then be obtained as T 0 0T x, where represents a ternary addition operator, and x is the ternary number equal to 0, 1, or 2 (when the scanned bit-pair is 00, 01 or 10, respectively), or equal to (10) 3 (when the scanned bi-pair is 11). Example 1: Consider the binary number First, we get T =2from the leftmost two bits 10. Then, for the next two bits, the equivalent ternary digit is 1. So we add (in ternary) the three numbers (20) 3, (02) 3 and (1) 3 to get the changed value of T = (100) 3 (which is equal to 9 in decimal corresponding to the binary digits 1001 scanned so far). Then for the next two bits, we get the equivalent ternary representation as (10) 3, which is added to (1000) 3 (= T 0) and (0100) 3 (= 0T ) in ternary to get the changed value of T = (1110) 3 = (39) 10. Finally, for the bit-pair 00, we add (11100) 3 to (01110) 3 to get the equivalent ternary number as (12210) 3 = (156) 10. Algorithm Binary2Ternary outlines the steps required to convert a number from binary to ternary. In a similar manner, we can reconvert the received m- digit ternary message to its equivalent binary form by using algorithm Ternary2Binary, where again we scan the given ternary number from its most significant (leftmost) digit position in a digit by digit manner, and convert the part of the so far scanned ternary digits to its equivalent binary representation. In the algorithm, we denote a ternary digit by t i and the converted binary number by B. As in the previous Binary2Ternary algorithm, we need to adjust the weight of the binary representation of the previously scanned ternary digits, by attaching a weight of three to it, with respect to 86

3 Algorithm 1 Algorithm Binary2Ternary 1: procedure BINARY2TERNARY(IN bit vector B, integer n, OUT ternary number T ) 2: /* Initialization */ 3: if b n 1b n 2 =00then 4: T 0; 5: end if 6: if b n 1b n 2 =01then 7: T 1; 8: end if 9: if b n 1b n 2 =10then 10: T 2; 11: end if 12: if b n 1b n 2 =11then 13: T 10; 14: end if 15: for i = n 3 downto 1 step -2 do 16: if b ib i 1 =00then 17: T T 0 0T ; this is a ternary addition 18: end if 19: if b ib i 1 =01then 20: T T 0 0T 1; ternary addition 21: end if 22: if b ib i 1 =10then 23: T T 0 0T 2; ternary addition 24: end if 25: if b ib i 1 =11then 26: T T 0 0T 10; ternary addition 27: end if 28: end for 29: end procedure the currently scanned ternary digit, and this is achieved by adding B0, 0B and the equivalent binary representation of the currently scanned ternary digit (note that, this time it is a binary addition). Example 2: Let the given ternary number T = The leftmost digit is 2 whose binary representation is 10. Sowe initially form the binary number as B =10. The next ternary digit is 1. We add the three binary numbers 100(= B0), 010(= 0B) and 1 to get the changed value of B = 111. The next scanned ternary digit is 0. So we add 1110 to 0111 to get the changed value of B as Finally, for the rightmost ternary digit 2, we add , and 10 to get as the equivalent binary representation B of the given ternary number T. Binary to ternary conversion of data provides us with the advantage of reduction in the total number of symbols to be transmitted and hence saves on the total transmission duration. As both the transmitter and the receiver consume energy proportional to the duration for which they are switched on, regardless of the symbol transmitted, this recoding to a higher radix will lead to a savings in energy at both the transmitter and the receiver simultaneously. Additional energy savings can be obtained at the transmitter by adopting the silent symbol strategy proposed in [2], [3] for one of the 3 possible ternary digits. Our proposed ternary with silent symbol (TSS) communication strategy combines these two strategies to derive an energy efficient communication scheme that saves energy at both the transmitter and the receiver. Assuming equal likelihood of all the three symbols, without any loss of generality we assume 0 to be our silent symbol. In practice however, the relative frequencies of the three symbols will depend on the nature of the data to be transmitted and hence a better strategy would be to assign the most frequent symbol as the silent symbol - the relative frequencies of the symbols determined either statistically or for each message to be transmitted. The transmission protocol of TSS thus consists of the following two steps: Protocol TransmitTSS Step 1: Recode the binary data to ternary using the Binary2Ternary protocol. Step 2: Transmit the ternary data symbol by symbol, obtained by the recoding process from step 1 using the following rules: Step 2.1: If the symbol to be transmitted is a 0, then switch off the transmitter for the symbol period. Step 2.2: Otherwise, switch on the transmitter and transmit the symbol (1 or 2) in the the time slot. The receiver side protocol would be the reverse of the transmitter: 87

4 Algorithm 2 Algorithm Ternary2Binary 1: procedure TERNARY2BINARY(IN ternary number T, integer m, OUT binary number B) 2: /* Initialization */ 3: if t m 1 =0then 4: B 00; 5: end if 6: if t m 1 =1then 7: B 01; 8: end if 9: if t m 1 =2then 10: B 10; 11: end if 12: for i = m 2 to 0 step -1 do 13: if t i =0then 14: B B0+0B; this is a binary addition 15: end if 16: if t i =1then 17: B B0+0B +1; binary addition 18: end if 19: if t i =2then 20: B B0+0B +10; binary addition 21: end if 22: end for 23: end procedure Protocol ReceiveTSS Step 1: Receive the ternary encoded data symbol by symbol as follows: Step 1.1: If a signal is received in the given time slot (symbol period), then the received symbol is either a 1 or a 2, depending on the predetermined received signal to symbol mapping. Step 1.2: Otherwise, if the channel is quiet, then the received symbol is a 0. Step 2: Convert the received ternary message to binary using the Ternary2Binary protocol. III. PROPOSED IMPLEMENTATION We consider the theoretical analysis of the energy savings generated at the transmitter for noisy channels. We assume that the channel noise is additive white gaussian noise (AWGN). Without any loss of generality, we assume that the transmitter uses FSK modulation with two frequencies - f c and f c +Δf corresponding to the symbols 1 and 2, respectively, and is switched off during the 0 s. Effectively this will be a hybrid modulation scheme involving FSK and ASK. As a representative example for showing the energy savings, we use here a non-coherent detection based receiver presented in [1] with a schematic structure as shown in figure 1. While combining ternary encoding with the silent symbol transmission strategy increases the total energy savings at the transmitter, for the detection of the silence of any carrier waveform in presence of bandpass noise, the instantaneous power levels of both non-silent symbols may have to be increased appropriately. This is because of the relatively inferior (higher) SNR requirements of ASK for the same BER value. Thus, although apparently some energy is saved during silent symbols, the detection of symbols in both non-silent and silent intervals with acceptable BER performance might increase the transmit power for the non-silent symbols employing PSK/FSK modulation scheme. Consequently, while one can indeed have some energy saving during silence transmission, there might as well be a need to increase the instantaneous power level for the non-silent intervals in order to achieve an overall energy savings for the system. For a successful strategy, the difference (in db) between the energy saved due to silence and the additional amount of energy needed for the increased instantaneous power for FSK needs to be positive. An analysis of the net savings in energy generated by TSS at the transmitter is given below. In figure 1, r is the signal received at the input of the receiver which consists of the transmitted signal s as well as the channel noise n, i.e, r = s + n. The received signal r is first passed through two bandpass filters (BPF) as shown in the figure, with center frequencies at f c and f c +Δf (corresponding to 1 and 2, respectively). Without any loss of generality, let q A and q B be the outputs of the threshold detectors corresponding to the symbols 1 and 2 respectively. q A and q B are ORed to get the value q. Ifthe transmitted symbol is 0 and if e th is correctly chosen, then the outputs of each of the threshold detectors will be zero, making q equal to 0. On the other hand, if the transmitted symbol is 1or2, and the threshold value e th is correctly chosen, it will generate the desired output symbol 1 or 2 at the output ŝ of the receiver. Thus, the desired output ŝ of the receiver is given as follows : i) If q =0, then ŝ =0, corresponding to s A = s B =0 ii) If q =1then, 88

5 BPF(A) f c ED e A ê A s A r = s + n BPF(B) f c + Δf ED e B Comparator/ Estimator ê B s B q q A e th q B e th Fig. 1. Representative hybrid FSK/ASK receiver for TSS scheme ŝ = A. Error Analysis { 1 if e A >e B, (when s A =1, s B =0) 2 if e B >e A, (when s B =1, s A =0) The computation of the BER in AWGN channels for the TSS protocol is the same as for the RBNSiZeComm protocol proposed in [1]. We reproduce here the analysis from [1] for the sake of completeness and readability: Let P 0, P 1 and P 2 be the probabilities of occurrences of the symbols 0, 1 and 2, respectively in the transmitted ternary message. Hence, the bit error rate BER can be written as, BER = P 0 (1 P {q =0 0}) + P 1 [1 P {q =1and e A >e B, e A 1}] + P 2 [1 P {q =1and e B >e A, e B 2}] (2) where P {q =0 0} is the probability that the symbol 0 is detected at the receiver, given that 0 was transmitted. Now, (1) P {q =0 0} = P {q A =0 0} P {q B =0 0} =[P {q A =0 0}] 2 (by symmetry) (3) Note that P {q A = 0 0} follows Rayleigh distribution. Assuming a zero mean and standard deviation of σ, we can write, P {q A =0 0} = eth e A 0 σ 2 e e 2 A /2σ2 de A =(1 e e2 th /2σ2 ) u(e A ) Hence, we have P {q =0 0} =(1 e e2 th /2σ2 ) 2. Now, let X = P {q =1and e A >e B, e A 1}. Then we can say, X = P {q =1 1}P {e A >e B, e A 1}) = X 1 X 2 (say), where X 1 = P {q =1 1} and X 2 = P {e A >e B 1, e A 1}. From figure 1, it follows that X 1 can be written as, X 1 = 0 [1 P {q A =0 1}P {q B =0 1}]p 1 (e A ) de A, where P {q B =0 1} is governed by a Rayleigh distribution, while P {q A =0 1} by a Rician distribution. Thus, P {q A =0 1} = P {q B =0 1} = eth 0 ( ) e e 2 A σ 2 e A +s2 A sa e A 2σ 2 I 0 σ 2 de A and, eth e B 0 σ 2 e e =1 e e2 th /2σ2 Solving for X 1 and X 2 we get, X 1 = 0( eth e A 0 σ 2 e 2 B /2σ2 de B [ ) 1 (1 e e2th /2σ2 I 0 (s A e A σ 2 ) X 2 = 0 e 2 A +s2 A 2σ 2 ) (s A e A ) ea e 2 I 0 dea σ 2 σ 2 e A +s2 A 2σ 2 ] de A (4) [( 1 e e2 A /2σ2 ) ea ] (s A e A ) I 0 u(ea σ 2 ) e 2 σ 2 e A +s2 A 2σ 2 de A (5) [ and, BER = P 0 1 (1 e e 2 th /2σ2 ) 2] + P 1 (1 X 1 X 2 )+P 2 (1 X 1 X 2 ) (6) 89

6 Assuming P 0 =0.33 and P 1 + P 2 =0.67, wehave, BER =0.67 [ 1 (1 e e2 th /2σ2 ) 2] +0.33(1 X 1 X 2 ) (7) B. Energy Requirement Calculations To compute the bit error rate for different SNR values from the above equation, we set s A = 1, i.e., signal power = s 2 A /2=1/2 and SNR value = 1/2σ2. The computations to find the optimum value of threshold voltage e th that gives the minimum BER for a given SNR value were implemented in Matlab. The results of these computations are shown in table I. Note that from equation 7, we get the required peak transmitter power for a given BER when the transmitter is transmitting a symbol. The SNR values for binary FSK modulation are determined from the relation, BER = 1 2 e γ/2, (8) where SNR in db = 10 log 10 γ. It follows from table I that for a given BER in the range of 10 4 to 10 8, the peak transmitter power in TSS is about 2.83db higher than that in non-coherent binary FSK detection (figure 2). However, in our proposed TSS scheme, the transmitter will be ON only during the non-zero symbols (1 or 2), and switched off during the zero symbols in the ternary encoded message. Hence the required average transmitter power for a given BER will be less than the value as demanded by equation 7. We now consider a possible implementation scheme: C. TSS Implementation We consider that the symbol duration of the ternary encoded message in TSS is the same as that of binary FSK. Thus, the transmission duration of the ternary encoded message will smaller than the equivalent binary message by virtue of the encoding. The average transmitter power for TSS will thus be reduced from the above peak transmitter power values due to silent symbol based transmission of one of the 3 symbols with a probability of 1/3. Noting that 10 log 10 (1 1/3) db =1.76 db, the average transmitter power in TSS will be less by 1.76db than the corresponding SNR values given above in table I, so that the average transmitter power P 1 for TSS communication in this case will be higher than the corresponding average power P b for binary FSK by ( ) db = 1.07 db. Thus we have, 10 log 10 ( P1 P b )=1.07 Hence, P1 P b = = Figure 2 depicts the comparison of peak and average scaled transmitter power for a given BER, between TSS and binary FSK. However, noting that in our implementation scheme the transmission duration T 1 will be reduced from the transmission time T b in case of binary FSK transmission by a factor of log 3 log 2 =1.6, wehave T b T 1 =1.6. Thus, the total (scaled) energy E 1 in TSS is related to the energy E b in binary FSK as, E 1 = P 1 T 1 = E b P b T b 1.6 =0.799, (9) log 10 (BER) Fig. 2. BER 7 Peak and Average Transmitter Power Comparison Transmitter Power (scaled) in db Peak TSS Power Avg. TSS Power FSK Comparison of peak and average transmitter power (scaled) for given TABLE I COMPARISON OF PEAK TRANSMITTER POWERS Required SNR in db BER value Corresponding TSS Binary e th value (volts) Protocol FSK which translates to a savings of energy by an amount of 20%. Simultaneously, there is a savings of energy at the receiver too: Let n 1 = log 2 n and n 2 = log 3 n be the number of symbols required to transmit a value n represented in binary and ternary respectively. Thus, the fractional reduction in the transmission duration by encoding in ternary is given by, n 1 n 2 =1 n 2 n 1 n 1 =1 log e 2 = (10) log e 3 Due to the reduction in transmission duration, the receiver needs to be switched on for a shorter duration compared to a binary encoded message transmission. We hence have the following result: Theorem 1: The proposed implementation scheme of TSS will result in an energy savings of approximately 20% at the transmitter and 36.9% at the receiver simultaneously, compared to binary FSK. Observation 1: Alternatively, if the ternary symbols were stretched to have the same transmission time as that of the original binary message, then the required bandwidth of the bandpass filters of the receiver will be reduced by a factor 90

7 of 1.6, reducing the gaussian noise power by the same factor, i.e., 10 log 10 (1.6) db =2.041 db. Thus, the required average transmitter power P 2 will be reduced from P 1 by db, i.e., P 2 will be less than P b by an amount of ( ) db =0.97 db. Thus, 10 log 10( P b P 2 ) = Hence, P b P 2 = = The transmission duration T 2 now is equal to T b. Hence, the transmitter energy E 2 in this case is related to E b as, E 2 E b = P 2 T 2 P b T b = =0.80, (11) which again means a savings of energy by an amount of 20%. However, no savings in energy would be generated at the receiver. IV. DEVICE ISSUES The discussions above assume that i) the time to turn on/off the radio is negligible compared to the time period of a transmitted pulse, ii) the underlying hardware supports the conversion and representation of binary number to ternary and vice versa and, iii) the energy spent on the conversion process is negligible compared to transmission/reception. Furthermore, there is the issue of maintaining receiver-transmitter synchronization for the duration of transmission of a data packet. A comprehensive study of the energy-efficiency of TSS will need to address the MAC protocols along with adequate physical-layer awareness. However, we do not address the issues related to MAC protocols in this work and make an effort to improve the energy efficiency with physical-layer centric approach. Given the simplicity of the Binary2Ternary and Ternary2Binary protocols, we believe that the energy consumed during the conversion process will not significantly impact the net energy savings generated by TSS. We present a discussion on some of the above mentioned issues below. A. Representation of Ternary Encoded Numbers In order to internally represent the three symbols (0, 1 and 2) at the MAC layer, we can use two bits to encode each ternary digit. As an example, the ternary digit 0 may be encoded using the bit-pair 00, while 1 and 2 using the bit-pairs 01 and 10 respectively. Thus, the MAC buffer for an n-bit binary data frame, would have to be 2n-bit wide to hold the equivalent ternary data. The system would read every bit pair to determine whether the corresponding transmitted ternary digit is to be a 0, 1 or a 2 and accordingly switch on/off the radio circuit. We also assume that the data in higher layers of the network stack are represented in binary only and a conversion of binary to ternary and vice versa occurs exclusively at the MAC layer. Only the digits 1 and 2 require transmissions, while the transmitter s radio circuit is switched off for the digit 0 and the receiver (assumed to have already been synchronized with the transmitter for this data packet transmission) interprets each silent symbol period as the ternary digit 0. B. Synchronization between Transmitter and Receiver We assume that the receiver and the transmitter are synchronized on the MAC packet header, similar to the protocol (which utilizes the preamble field in the MAC frame header for synchronization between the sender and the receiver). The transmission time of the payload of the packet is assumed to be small enough so as not to lose synchronization between the receiver and the transmitter. C. Effect of Device Characteristics For most devices, neither the time to switch on/off the radio nor the energy consumed by the device when the radio is switched off, is negligible. As such, the maximum data rate at which we can transmit using the TSS protocol is dependent on the time it takes to turn the radio circuit on or off. For most commercially available radio devices, the power drawn in the transmit or receive state is considerably more than the active state - i.e., when the radio is switched off [10]. However, the penalty paid for this low power operational state is the switching time from the active to transmit state and vice versa. In particular, the energy expenditure for transmitting the equivalent ternary data frame using the TSS can be represented as the sum of the following three terms, 1) A base energy value that exists throughout the duration of the transmission, which is equal to the energy required when the device is in the active state. 2) Extra energy in addition to the base energy required to transmit a 1 or 2 in TX state. 3) Extra energy in addition to the base energy required during the edge transitions between a symbol 0 and the symbols 1 and 2, which corresponds to switching from the active to TX state and vice versa. For the sake of simplicity of computation, we can consider the radio circuit as a single stage low pass RC circuit for the transition from the active state to transmit state and vice versa, given the fact that for most devices this transition time is of the order of microseconds. We assume that the radio is deactivated on the trailing edge of a 1 or 2 that is followed by a 0, while it is activated on the rising edge a 1 or 2 following a 0. Under these assumptions, it is not unreasonable to assume that the rise time of the RBN symbol pulse is almost equal to the fall time of the pulse and in turn to the turn-on time of the radio circuit from the active to the transmit state. It was shown in [2], [3] that under the above assumptions, the effect of the radio device characteristics on the performance of their RBNSiZeComm protocol can be neglected for several of the commercially available low cost, low power radios popularly used in wireless sensor networks. As our TSS protocol also uses 3 symbols similar to the RBNSiZeComm protocol, we can hence conclude that their results would also be applicable for TSS. V. DISCUSSION ON SOME POTENTIAL APPLICATION SCENARIOS The evolution of 3G enabled cellular telephony has lead to the enabling of many new applications, remote healthcare 91

8 systems being one of them. For example, smartphones are fast emerging as the platform of choice for wireless remote healthcare. Cheap medical sensors attached to the body to create a body area network can talk with such phones - Bluetooth being the communication protocol of choice [6]. The smartphones then relay this information over the cellular network to a central repository/processing center. Real time collection and transmission of physiological data however, imposes a huge burden on the energy constrained communication devices. As a typical example, at low data rates of 60 bytes/s and transmission frequency of about once every 0.5 1sec (the usual required data rate for transmitting physiological data like ECG signals and other vital body parameters), would result in a cell phone s battery being drained completely within hours - even while transmitting data over a strong connection [6]. In other words, this implies a recharge cycle of every half a day for a user. By using TSS, one can considerably improve upon this recharge cycle. Likewise, considering wireless sensor networks deployed in farms, vineyards, etc., the collected information through sensor networks include various climate data such as temperature, relative humidity, solar radiation and soil composition, etc. These data are typically transmitted as ASCII text messages [7] [9]. Considering a typical deployment scenario for a field of size 60m 60m, radio range of 0 80m with multi-hop communication, data rate of 300 packets/min = 0.5 packets/s and duty cycle of 15% at 20mA with a battery of 7.2Ah capacity, a radio battery will last for only 100 days [7] [9]. Given the difficulty in replacing the batteries of sensor devices deployed in such network deployment scenarios, using an energy efficient communication scheme such as TSS that provides savings in both transmission and reception of data would greatly benefit the farmer. VI. CONCLUSION We have presented in this paper a new low energy communication scheme that can generate energy savings simultaneously at the transmitter and the receiver, unlike the RBNSiZeComm protocol in [1] [3]. Using low cost and low complexity implementation scheme based on a hybrid modulation utilizing FSK and ASK for the TSS protocol proposed we show that for AWGN noisy channels, there is an average savings of 20% in battery energy at the transmitter for equal likelihood of all possible binary strings of a given length. Simultaneously, there is a savings of 36.9% energy at the receiver. An efficient algorithm involving only addition (and no multiplication or division) for conversion from binary to ternary and vice versa is used in order to keep the energy consumed for the radix conversion process low at both the transmitter and the receiver. Coupled with the low cost and low complexity of transceiver, these savings clearly demonstrate the usefulness of TSS for low power wireless sensor networks, particularly for multihop communications. REFERENCES [1] K. Sinha, An energy efficient communication scheme for applications based on low power wireless networks, to appear in Proc. 6th IEEE Consumer Communications and Networking Conference (CCNC), Las Vegas, USA, Jan , [2] K. Sinha, A new energy efficient MAC protocol based on redundant radix for wireless networks, Proc. Recent Trends in Information Systems (RETIS), Calcutta, pp , [3] K. Sinha and B. P. Sinha, A new energy-efficient wireless communication technique using redundant radix representation, Tech. Rep., Indian Stat. Inst., ISI/ACMU-07/01, [4] Y. Zhu and R. Sivakumar, Challenges: communication through silence in wireless sensor networks, Proc. Intl.Conf. on Mobile Comp. and Networking (MobiCom), pp , [5] Y. P. Chen, D. Wang and J. Zhang, Variable-base tacit communication: a new energy efficient communication scheme for sensor networks, Proc. 1st Int. Conf. on Integrated Internet Ad hoc and Sensor Networks (InterSense), Nice, France, [6] D. Jea, J. Liu, T. Schmid and M. Srivastava, Hassle free fitness monitoring, Intl. Workshop on Sys. and Networking Support for Healthcare and Assisted Living Environments (HealthNet), June, [7] J. Burrell, T. Brooke and R. Beckwith, Vineyard computing: sensor networks in agricultural production, IEEE Per. Comp., vol. 3(1), pp , [8] R. Beckwith, D. Teibel and P. Bowen, Report from the field: results from an agricultural wireless sensor network, Proc. IEEE Intl. Conf. Local Comp. Networks, [9] P. Mestre, E. Peres and C. Serôdio, Agricultural monitoring and control system using low-cost wireless sensors networks, Proc. 5th EFITA/WCCA, pp , [10] J. Polastre, R. Szewczyk and D. Culler, Telos: enabling ultralow power wireless research, Proc. Intl. Symp. on Inf. Processing in Sensor Networks, pp , [11] I. Demirkol, C. Ersoy and F. Alagoz, Mac protocols for wireless sensor networks, a survey, IEEE Comm. Mag., [12] W. B. Heinzelman, A. Chandrakasan and H. Balakrishnan, Energy efficient communication protocol for wireless microsensor networks, Proc. Intl. Conf. on Sys. Sc., [13] K. Sinha, S. Ghose and P. K. Srimani, Fast deterministic broadcast and gossiping algorithms for mobile ad hoc networks, J. of Par. & Dist. Comp. (JPDC), vol. 68(7), pp , [14] M. Cagalj, J.-P. Hubaux and C. C. Enz, Energy-efficient broadcasting in all-wireless networks, Wireless Networks, vol. 11, nos. 1/2, pp , [15] I. F. Akyilidiz, W. Su, Y. Sankarasubramaniam and E. Cayirci, Wireless sensor networks: a survey, Computer Networks, vol. 38, no. 4, pp , March [16] J. N. Al-Karaki and A. E. Kamal, Routing techniques in wireless sensor networks: a survey, IEEE Wireless Communications, vol. 11, no. 6, December [17] C. Intanagonwiwat, R. Gonvindan, D. Estrin, J. S. Heidemann and F. Silva, Directed diffusion for wireless sensor networking, IEEE/ACM Trans. on Networking, vol. 11(1), pp. 2 16, [18] W. Ye, J. Heidemann and D. Estrin, An energy-efficient MAC Protocol for Wireless Sensor Networks, Proc. IEEE Infocom, pp , [19] Y. Li, W. Ye and J. Heidemann, Energy and latency control in low duty cycle MAC protocols, Proc. IEEE Wireless Communications and Networking Conference, pp. PHY 30 4, [20] C. Enz, A. El-Hoiydi, J.-D. Decotignie and V. Pereis, WiseNET: an ultralow-power wireless sensor network solution, IEEE Computer, vol. 37, no. 8, pp , [21] Y. P. Chen A. L. Liestman and J. Liu, Energy-efficient data aggregation hierarchy for wireless sensor networks, Proc. 2nd Int. Conf. on Quality of Service in Heterogeneous Wired/Wireless Networks (QShine 05), Orlando, Florida, [22] J. G. Proakis, Digital Communication, McGraw-Hill, 4th ed.,