A Software Defined Radio Based Data Link Design for VHF Band Wireless Sensor Networks G. Celika,∗, A. Aitalievaa , H. Celebia , M. Uysalb a Department b Department
of Computer Engineering, Gebze Technical University, Kocaeli, Turkey of Electrical and Electronics Engineering, Ozyegin University, Istanbul, Turkey
Abstract In this paper, we present an overall data link design for a tactical wireless sensor network operating at VHF band considering software defined radio (SDR) approach. An empirical channel modelling for the target environment is developed, and the large and small scale statistics of the wireless channel are obtained. Then, the physical layer properties including modulation, channel coding, frequency hopping are designed according to the power and bandwidth efficiency requirements. Additionally, the medium access control (MAC) and automatic repeat request (ARQ) protocols are designed and the complete end-to-end system is tested in the field using the SDR platform. The measurement results are promising for the proposed datalink of the VHF wireless sensor networks operating in the dense hilly environments. Keywords: Channel modeling, data link design, tactical communications, sensor networks, software defined radio
1. Introduction Tactical communication systems are designed to provide secure and reliable mission-critical communication capabilities for military. In recent years, wireless sensors are being increasingly used for tactical communications [1]. Especially, designing a reliable data link that collects the data from wireless sensors in the harsh channel environments and conveys them to center node is critical. Prior to designing of a data link for such wireless sensor networks, the channel propagation needs to be modeled by performing the measurement in the field [1, 2]. Due to physical phenomena like reflection, refraction, diffraction, and scattering the wireless signal that is traversed the path from a transmitter to a receiver faces distortion or impairments in various ways [1, 2, 3]. The multipath components (MPCs) with different phases, delays, and attenuations can be added at the receiver side constructively or destructively, which ∗ Corresponding
author Email addresses:
[email protected] (G. Celik ),
[email protected] (A. Aitalieva),
[email protected] (H. Celebi),
[email protected] (M. Uysal)
Preprint submitted to Elsevier
July 14, 2018
leads to the fading phenomenon [4, 5]. The phases of the MPCs depend on the position of the transmitter, the receiver and the surrounding objects, and the amplitude of the total signal changes with time [6, 7]. Small scale fading statistics are obtained using this aggregated MPCs. The large scale fading statistics consist of path loss of the signal as a function of distance and shadowing variance. This implies that the signal propagation over wireless channel varies according to the environment, therefore the aim of the channel modeling is to characterize the effects of the channel environment on the communications signal [5, 6, 7]. In the literature, the software defined radio (SDR) application on VHF band is also investigated in [8, 9, 10, 11]. The developed applications are run at different carrier frequencies and modulation schemes. However, the scenario proposed in our paper is not studied in the aforementioned papers. The contributions of this study are two folds: 1) empirical channel propagation modeling of VHF band (144 − 147 MHZ) in the dense hilly environment, 2) SDR implementations of the PHY and MAC layers for the data link of tactical wireless sensor networks. In this paper, we consider a tactical wireless sensor network operating at VHF 144 − 147 MHz band. The studies on VHF channel propagation modeling mostly cover the urban as well as the rural and forested environments [12, 13, 14, 15, 16, 17]. However, to the best of our knowledge, very few publications are available in the literature that discuss the issue of channel modeling at the VHF band in hilly, irregular terrain with fixed terminals. For instance, VHF propagation measurement at a frequency of 110.6 MHz at low altitude over hilly and forested terrain have been performed to develop a computer-based propagation model which could predict path loss by giving the terrain profile between the transmitter and receiver [12]. But, the propagation of waves in the hilly and forested area faces different propagation mechanisms than only hilly terrain. Additionally, taking measurement at different altitude have impacts on the propagation characteristics of signal. Moreover, the measurement that have been carried out at 210 MHz frequency in the city and the outskirts of Bern, which is one of the hilly regions of Switzerland, in order to find the multipath delay spread of terrains [18]. Related works in the field of empirical and ray-tracing based channel modeling for urban, rural and suburban regions and indoor can be referred to [19, 20, 21, 22, 23, 24, 25, 26, 27]. However, none of these studies coincide with our work geographically and the operating frequency band. In [28], ray-tracing based channel modeling for VHF band (145 MHz) is conducted using Wireless Insite software tools. Additionally, there is not any study in the literature on channel modeling of 145 MHz frequency at hilly and irregular terrain. Hence, modeling of the channel environment for wireless sensor networks operating at VHF frequency band plays an important role in developing tactical wireless sensor networks. Large and small scale channel statistics can be used for the tactical communications system design[18, 29, 30, 31]. Based on the channel characteristics, modulation and channel coding schemes of the system are determined. Note that the fundamental requirements of tactical wireless sensor networks are 2
power efficiency, minimization of out-of-band emission, i.e., spectral efficiency, and bit error rate performance to achieve target desired quality of service (QoS) as well as low receiver complexity [32]. In this study, the communication environment is selected as a hilly and mountainous environment where remarkable path loss, shadowing effects and large delay are experienced. As the modulation and channel coding scheme, continuous phase modulation (CPM) and low density parity check (LDPC) channel coding combination is selected. The choice of proper combination of LDPC and CPM is crucial to satisfy the target tactical requirements of spectral efficiency, power efficiency and complexity of the system. Therefore, this specific combination is chosen as a suitable tradeoff among these three criteria. Besides the CPM-LDPC combination, frequency hopping (FH) is also applied to the design in order to reject interference from external sources. In order to show the performance of the proposed data link for the VHF tactical wireless sensor networks, the system is implemented in the SDR platform and the measurements are performed in the field. The overall communication system is implemented in an SDR platform, which is NI USRP 2920 device. In the field tests, the tactical communication system consists of a center node and wireless sensors are considered. The sensors collect data from the environment and send the data to the center node through the designed data link. In order to allow packet based multiple access, time division multiple access (TDMA) based medium access control (MAC) protocol which controls the packet delivery using automatic repeat request (ARQ) is designed. The paper is organized as follows: In Section 2, empirical channel modelling of VHF tactical wireless sensor networks is presented. In Section 3, the design of physical and MAC layers are provided. This includes LabView design for ARQ and MAC layer protocols. This is followed by the demonstration of the field test results of overall network using NI 2920 devices. Finally, the conclusion remarks are presented in Section 4. 2. Empirical Channel Propagation Modeling The common methodology for designing a wireless system (i.e. data link in our case) for a given frequency band operating at target channel environment is as follows. Initially, the channel propagation modeling for the given frequency band is performed in order to understand the statistical characteristics of radio channel for the desired operating environment. Once the statistical propagation parameters are obtained, then these parameters are utilized for the selection of transmission and reception techniques (modulation, coding, etc.) and algorithms at transmitter and receiver side. Therefore, in this study, we first aimed to obtain the statistical parameters of target channel environment, which is presented in this section. 2.1. Channel Measurement Setup and Environment VHF tactical wireless networks are mainly deployed in harsh environments such as valleys and dense rocky hills and mountains. Therefore, we have selected 3
such region (Inozu Valley) in Ankara, Turkey for the channel measurements. In our application, VHF tactical wireless sensor networks require long-range communications (on the order of kilometers). Hence, we have selected a time-domain channel modeling method, which is preferred for the long-range channel modeling [1]. The channel measurement setup for this method is shown in Figure 1. In the measurement setup, Rohde & Schwarz SMBV100A and FSV7 equipments are considered as transmitter and receiver, respectively [33, 34].The RF power amplifier of 50 Watts (47 dBm) is used at the transmitter side in order to increase the transmission range. The transmitter and receiver need to be synchronized in frequency, phase and time for conductiong the channel propagation measurements. The GPS clock reference is used to achieve synchronization between devices in time and frequency. For this purpose, SpectraCom EC1S GPS Clock Reference is employed which produces 10 MHz reference signal and 1 PPS (one pulse per second) signals from the received GPS signals. The generator SMBV100A transmits signal when the device is triggered by the 1 PPS signal from TRIG inputs, and at the same time signal analyzer FSV7 starts to sample when TRIG gets the trigger signal. As a result, both devices start to transmit and receive synchronously. The block diagram of the measurement setup and the list of the used equipments are shown in Figure 1 and Table 1, respectively.
Figure 1: Channel measurement setup.
The transmitter is located at the highest point of the measurement envi-
4
Table 1: Channel measurement setup equipments.
Equipment Rohde & Schwarz SMBV100A Rohde & Schwarz FSV7 Rohde & Schwarz FSV7 B-22 Mini Circuits LZY-1+ SpectraCom EC1S GPS Clock RFS Dipole Antenna
Usage Vector Signal Generator Signal and Spectrum Analyzer Preamplifier 50W (47 dBm) RF PA Reference GPS Clock 145 MHz TX/RX antenna
ronment which has altitude of approximately 1590 meters. The receiver points are selected in such a way that they have decreasing altitudes along the valley. Note that these points are located on the road which goes through the valley. The devices used at the receiver side are installed on a SUV vehicle so that the receiver can be easily moved from one point to another one during the measurements. The transmitter and the receiver sides are shown in Figure 2 and Figure 3, respectively.
Figure 2: Transmitter side.
Figure 3: Receiver side.
5
The distance between the measurement points is selected as 500 meters. The measurement data are collected in this region over approximately 14 km where it is illustrated in Figure 4. The altitude profile of these receiver points are shown in Figure 5.
Figure 4: The representation of the measurement points.
Figure 5: Altitude profile of the points.
The channel measurement campaign is conducted as follows. Firstly, the measurement time of the channel, Tmeas is chosen lower than the coherence time of the channel, Tc , i.e. Tmeas < Tc , since the channel needs to be stationary
6
during the measurement [1, 2, 5].Secondly, the pulse width of the transmitted signal Tp , is selected lower than the delay spread of the channel, τc , i.e. T p < τc in order to avoid inter-symbol interference (ISI) [1, 2, 5]. Furthermore, the initial values of Tc and τc for the VHF band are set as 1s and 5µs, respectively, [29]. According to these limitations, the measurement time for each point is selected as 1 second and the period of the transmit signal is selected as 100µs with 4µs (Tp ) duty cycle. The transmit pulse used for the channel measurement is shown in Figure 6.
Figure 6: The transmit pulse used for the channel modeling.
At each receiver point, the measured signal is collected in the IQ form. Then, these recorded data are used to obtain the small and large scale statistics of the channel environment. 2.2. Channel Propagation Statistics The measured data collected in the environment are processed in MATLAB in order to model channel propagation characteristics in terms of path loss model. The received power v.s. distance is shown in Figure 7. It can be seen that as the distance between transmitter and receiver increases, the received signal power decreases. In order to estimate the path loss exponent (n) and shadowing parameter (σ) the received power at a distance d from the transmitter is calculated using the following equation [1], 0
Pr (d) = Pr (d0 ) − 10n log( 0
d ), d0
(1)
where Pr (d) is received signal strength at distance d distance, and Pr (d0 ) is the received signal strength at a reference distance d0 . 7
Figure 7: Received signal power vs. distance
In our case, d0 is determined as 40m, since wavelength λ is equal to 2.069m. Then, the minimum mean squared error (MMSE) estimation method is used for estimating the path loss exponent (n) as [1, 2], J(n) =
k X
0
Pr (d) − Pr (d)
2
,
(2)
i=1 0
where Pr is the computed value and Pr is the calculated value. The next step is to equalize J(n) = 0 and compute n value. The resultant n value is used to find the shadowing parameter (σ) [4, 5] σ=
J(n) 45
2 ,
(3)
Applying this method to the measurement data, the empirical path loss exponent and shadowing are calculated as n = 3.5138 and σ = 6.2520, respectively. In our case, the transmitted power Pt is 47dBm and by taking cables and connectors losses into account, Pt is calculated to be 45 dBm. In this case, the path loss Pl for each point is calculated by, 0
Pl = P t − Pr .
(4)
The log-distance path loss model P Lld (d) is given by, P Lld (d) = P Lf (d) + 10n log
8
d , d0
(5)
where P Lf (d)) is the path loss in free space.
Figure 8: Empirical path loss model.
The path loss for the computed values is plotted and compared with the empirical values in Figure 8. The characteristics of a multipath fading channel are often defined by a power delay profile (PDP). This is calculated by measuring the spatial average of |h(t, τ )|2 over local area, where h(t, τ ) is the channel impulse response (CIR) which is represented by a discrete number of impulses as follows, |h(t, τ )|2 =
N −1 X
αk2 δ 2 (τ − τk ).
(6)
k=0
The PDP at the time t0 for a probing pulse P (t) at the channel input is given as, 2
P (τ0 ) = |r(t0 )| =
N −1 X
αk2 (t0 ).
(7)
k=0
In order to determine the small scale statistics, IQ measurement samples obtained at 45 receiver points are used. First, the CIR for each point is obtained and then the PDP is computed and plotted, which is shown in Figure 9. According to these results, as the distance between the transmitter and the receiver increases the time delay of the received signal also slightly increases. It
9
also can be seen that while the range between transmitter and receiver increases there is a reduction in the signal strength.
Figure 9: Power delay profile at different receiver points
The small scale fading channel statistics are obtained for both original (correlated) data and decorrelated data cases. The purpose of decorrelation process is to decorrelate random variables by reducing correlation between them. In other words, observation matrix should have a diagonal covariance matrix. In our case, the decorrelation process is applied to the channel measurement data in order to reduce the correlation effects in the PDP[35]. If each sample of PDP is assumed to be random variable, then PDP includes 200 random variables. For each random variable, 9614 observations are available. If the columns are assumed to be random variable, the measurement matrix with dimension of 9614 × 200 is obtained. This operation can be performed by the following steps. If the observation matrix is X, then its covariance matrix is found by, P
= Cov(X) = E XX T .
(8)
P
Note that covariance matrix can be written in terms of eigenvalues and eigenvectors as, X = EDE−1 , (9) where E has dimensions of 200 × 200 matrix and each row indicates an eigenvector and the corresponding eigenvalues. D refers to the eigenvalues of the covariance matrix. D is a diagonal matrix and it can be expressed by,
10
ET
X
E = D.
(10)
where we used E−1 = ET . By using matrix X, a new Y matrix with a diagonal covariance is obtained by, Y = ET X.
(11)
As a result, the correlated and decorrelated PDPs of the channel are shown in Figure 10 and Figure 11, respectively. In addition, the corresponding correlated and decorrelated channel transfer functions (CTFs) are illustrated in Figure 12 and Figure 13 respectively.
Figure 10: Average PDP for the original (correlated) case.
As seen in Figure 10 and Figure 11, the MPCs in the decorrelated PDP case can be distinguished clearly than in the original PDP case. In order to differentiate between MPCs and the noise, the noise threshold is chosen as −81.6681 dBm. By employing the PDP results, the small scale characteristics are determined for both original and decorrelated PDP cases. In this process, the maximum excess delay is tx − t0 where t0 is the first arriving delay and tx is the maximum delay of the signal in the PDP. The mean excess delay is the first moment of the PDP and it is defined by [6, 7], P P 2 P (τk )τk α τk . (12) τ¯ = Pk k 2 = Pk α k k k P (τk ) 11
Figure 11: Average PDP for the decorrelated case.
The root mean square (RMS) delay spread is the square root of the second central moment of the PDP and it is defined by [6, 7], q σ = (τ¯2 ) − (¯ τ )2 , (13) where, (τ¯2 ) is given by: P P 2 2 α τ P (τk )τk2 . (τ¯2 ) = Pk k 2k = Pk k αk k P (τk )
(14)
The coherence bandwidth is used to characterize the channel in the frequency domain, Bc, which is approximately equal to the inversely-proportional to the RMS delay spread (Bc ≈ σ1τ ). Note that all the channel small scale fading statistics obtained for decorrelation and original cases are shown in Table 2. Table 2: Empirical small scale fading statistics for both decorrelated and original cases.
Maximum excess delay Mean excess delay RMS delay spread Coherence bandwidth
Deccorrelated 26 us 2.48 us 3.26 us 306 kHz
12
Correlated 44.6 us 9.53 us 3.43 us 292 kHz
Figure 12: The CTF for the correlated case.
3. MAC and PHY Layer Implementations in SDR Platform 3.1. Utilization of Channel Parameters for PHY and MAC Design In the previous section, an extensive channel measurement campaign is conducted to obtain statistical channel parameters of interest in order to be utilized for PHY and MAC layer designs. The obtained statistical channel parameters are large and small scale fading statistics such as path loss exponent, shadowing variance, coherence time, coherence bandwidth, delay spread. These channel parameters are utilized for the selection and design of PHY and MAC techniques. For instance, the delay spread results are used for the mitigation of Inter-Symbol Interference (ISI) and the propagation delay results are utilized for the calculation of RTT. The rest of channel parameters utilized for the system designs are discussed throughout this section. We have determined PHY and MAC design using the channel statistics that we have obtained in the previous section. First of all, selection of single vs. multi-carrier waveforms for our communications system is vital. Since the measured coherence bandwidth of the channel (306 kHz, see Table 2) is greater than the signal bandwidth (200 kHz), the channel type in our case is frequency flat channel. Therefore, we decided to use single-carrier communications rather than multi-carrier communications in our design.
13
Figure 13: The CTF for the decorrelated case.
Another critical design option is to decide on whether using channel coding or not. In this regard, since the measured path loss is large (on the order of 90dB), low SNR values are measured at the receiver side. Considering this fact, we decided to use FEC techniques (LPDC) in our design for improving the performance of the communications systems. Therefore, we determined the optimal transmission parameters for the proposed LDPC-coded CPM system to satisfy given spectral efficiency subject to a constraint on the demodulator complexity [32]. 3.2. MAC and PHY Layers Design In this section, the ARQ based MAC layer design for this communication system is presented. we selected Stop-and-Wait (SAW) technique for the ARQ mechanism [36, 37]. In this protocol, the out-of-order is rejected by the receiver and the erroneous received packet are not accepted. For either case, ACK packet for the last successfully received data packet is sent. However, when a packet is received without error, the ACK packet for this data packet is sent to the transmitter terminal. The round-trip time (RTT) for the packet transfer is selected as RT T ≥ Tp + Tprx + Tptx , where Tp signal propagation time, Tprx and Tptx packet processing times for center node and sensor, respectively, including LDPC coding/decoding operations. In this study, the MAC protocol is designed in order to satisfy TDMA between center node and sensor nodes. For each time slot, exactly one of the 14
sensor nodes is communicating with center node and the other nodes remain sleep during this time. Source Address (32-bit) Destination Address (32-bit) Control Bits (16-bit)
Sequence Number (4-bit)
Payload Length (16-bit)
Payload (max. 4000-bit)
Figure 14: Frame format.
The frame format of the data and ACK packets are designed as shown in Figure 14. The format of the source and destination addresses are selected as ordinary IPv4 address, e.g. 192.168.1.1. The control bits are used to convey some information bits which are used in the MAC layer operations. For each frame, the sequence number of 4-bit is defined in order to represent the order of the packets. Since the payload size is variable, which is at most 4000-bit, the payload length in bits is defined in the frame format. Therefore, the frame contains at most 4100 bits. Since the center node and the sensors communicates in TDMA fashion, the communication order of the sensors is determined by the center node and sent to the sensor nodes. In this process, the sensor nodes send Iamhere message to the center node repeatedly until the OK message from the center node is received. The center node listens to Iamhere packets and sends OK to the corresponding sensor node. When Iamhere messages are received from all the sensor nodes, the order of these nodes are set in a packet and the center node broadcast this order packet to the sensor nodes. The sensor nodes extract their own orders from the received packet. The flowchart of the order determination processes for both center node and sensor nodes are shown in Figure 15 and Figure 16, respectively. After receiving the orders, the node, which have order 1, immediately starts to communicate with center node. The other nodes sleep until they take the turn which is illustrated in Figure 17. In this study, NI-USRP 2920 SDR platform is used in order to implement the MAC protocol. NI-USRP 2920 devices support 50 MHz - 2.2 GHz adjustable frequency range with I/Q sampling frequency between 200 kHz and 20 MHz. The physical and MAC layer designs including ARQ and LDPC coding/decoding methods are implemented using LabView software development kit. The CPM modulation and LDPC channel coding parameters that are determined in one of our previous studies is used in the physical layer design [32]. They are implemented in LabView using Modulation Toolkit Libraries of LabView environment. The designed system also includes frequency-hopping (FH) structure. The channels used in the FH system is determined according to pseudo-random (PN) codes to mitigate the effects of interference. The PN sequences at each communicating terminal produce the same channel value at the same time. This condition is satisfied in LabView by providing the same seed and the PN polynomial degree at each terminal. 4-bits linearly shifted feedback register (LSFR) are used to generate the PN bit sequence to obtain the channel values between 1 and 16. In the FH communications, it is required that the center node and sensor 15
Sensor Counter = 0
Sensor Counter < 3
No
Get the sensors on the list
Broadcast the orders of the sensors in TDMA
Yes Listen to the sensors
Go to Communication Stage
No
Is heartbeat message received from any of the sensor?
Yes Increment Sensor Counter
Send acknowledgement message to the corresponding sensor and add this to the sensor list
Figure 15: The flowchart of the order determination for the center node.
nodes are synchronized over the same channel. In the synchronization stage, one of the channels is set as receive channel at the center node and the sensor nodes transmit “SYNC FREQ” synchronization messages to the center node and switch to the receive mode. After receiving synchronization message, the center node replies to the corresponding sensor node by transmitting “SYNC ACK” message. If the sensor nodes receive this reply, the channel synchronization is achieved at this time. After that point, both communicating nodes go to the communication stage and hop to the same channel at the same time. The synchronization steps for both terminals are illustrated in Figure 18. In the communication stage, the center node and the sensor nodes communicate as shown in Figure 19. The data and ACK packets are sent over different channels. In other words, the channels hop for each packet. At the end of the channel list, the hopping pattern takes the initial position. The communication stages including MAC protocol for both terminals are shown in Figure 20 and Figure 21, respectively. The selected parameters for CPM and LDPC are given in Table 3 [32]. In addition, the root raised cosine pulse shape filter is used to
16
Send heartbeat message to Center
Wait for the response
Is acknowledgement message received?
No
Yes Wait for the order number
Is order message received?
No
Yes Sleep order (Order number x Slot Time)
Go to Communication Stage
Figure 16: The flowchart of the order determination for the sensor nodes.
shape the symbols. The parameters of the pulse shape is given in Table 4. In the packet-based communications, the frame synchronization bits are transmitted per each frame to determine the start of the frame data obtained by the receiver. In this study, 20-bit length synchronization sequence generated by PN sequence is used. These bits are added at beginning of each frame. In addition, 100 guard bits are added at the beginning of each frame to mitigate the multipath effects. The overall frame format is given in Figure 22. The synchronization bits are known at the receiver side. Using the CPM 17
Figure 17: The illustration of TDMA communications between center node and sensor nodes.
Table 3: CPM and LDPC parameters.
Parameters CPM symbol length (L) CPM alphabet size (M) CPM modulation index (h) LDPC coding rate (R)
Values 4 2 0.2 0.5
Table 4: Pulse shaping filter parameters
Parameters Alpha Filter length TX filter type
Values 0.5 8 Root raised cosine
Generator Synchronization Parameters block in LabView at the receiver side, it can be determined during the demodulation process whether the received signal includes the synchronization bits or not. The synchronized signal are then subject to LDPC decoding after the demodulation. The channel matrix for LDPC is selected from one of the WiMax parity check matrix with size 144x576. 3.3. The Field Tests for Performance Evaluation The designed complete end-to-end SDR communication system is tested with a simple file transfer in the open area. The center node and the sensor node are located at the separation distance up to 350 meters as shown in Figure 23 and Figure 24, respectively. This distance is determined based on maximum output power of the NI-USRP 2920 (up to 15W using external power amplifier) [38] and the measured receiver sensitivity of NI-USRP 2920 in the field, which is -80dBm
18
Determine channels using PN sequence generator and push to the channel queue
Determine channels using PN sequence generator and push to the channel queue
Pop a channel from queue, set this as SYNC channel and add to the end of queue
Using the first channel on channel queue, start to listen to message
Send message over SYNC channel
Is message received?
No
Wait for timeout Yes Wait for the response from the center over SYNC channel
Is received?
Send message to the corresponding sensor Pop the used channel and push to the end of the channel queue
No
Yes
Go to Communication Stage
Go to Communication Stage
(a)
(b)
Figure 18: Synchronization stages for (a) center node and (b) sensor nodes .
considering 200 kHz signal bandwidth to achieve BER of 10−3 . The center node and the sensor node is selected as receiver and transmitter, respectively. In this case, an image is used as test image and it is transferred over the designed data link. The carrier frequency of the nodes vary between 144 − 147 MHz since the FH communications is used over 200 kHz channels. In other words, the carrier frequencies hop among 15 channels for each packet transmission. When both terminals start to communicate each other, they are in the order determination stage. Since there is one sensor node, the order determination
19
Channel
f1
f2
f3
f4
f5
f1
Sensor Transmit Packet
Sensor
Channel
Wait for Packet
f1
Transmit ACK
f2
Wait for Packet
f3
Transmit ACK
f4
Wait for Packet
f5
Transmit ACK
f1
Wait for Packet
Transmit Packet
Wait for Packet
Transmit Packet
Wait for Packet
Figure 19: The FH communications between the center node and the sensor nodes.
stage is omitted by both terminal. After this stage, the synchronization stage is performed. The center node operates at a frequency and the sensor node tries to synchronize to this frequency (find the channel that is used by the center node) by adjusting the carrier frequency according to the hopping pattern. After the synchronization is achieved, both terminal switch to the communication stage and stay at this stage unless the synchronization fails. In the communication stage, the sensor node picks a data packet from the queue and transmits it and waits for the ACK packets within a timeout period. If the data packet is received at the center node, the ACK packet including the sequence number of the received data packet is transmitted to the sensor node, immediately. The received data packet is stored into the queue. When the last data packet is received at the center node, the image is constructed using the data packets in the data queue as show in Figure 25. The LabView built-in VI components are used to generate the image.
20
End of Slot?
Sleep 2 Slots Yes
No
Synchronized?
No
Go to Synhchronization Stage
Yes Pop a channel from channel queue, set this as RX channel and add to the end of queue Energy on the RX channel over threshold? Pop a channel from channel queue, set this as TX channel and add to the end of queue
Yes
Receive raw signal and send to the Demodulator
No
Switch to RX mode, pop a packet from packet queue and transmit
Is frame synchronization achieved on the demodulated signal?
Decode the data Yes
No
No
Does received data packet belong to me and the transmitted packet ?
Yes
No
Send packet for the received packet number over TX channel
3 slots time unsuccessful packet/ACK transmission?
Yes
Go to Synhchronization Stage
Figure 20: The communications stage process for the center node.
21
End of Slot?
Sleep 2 Slots Yes
No
Go to Synhchronization Stage
Synchronized? Yes
No Pop a channel from channel queue, set this as TX channel and add to the end of queue
Wait for ACK
Pop a channel from channel queue, set this as RX channel and add to the end of queue
No
Energy on the RX channel over threshold?
Yes
Receive raw signal and send to the Demodulator
Switch to TX mode, pop a packet from packet queue and transmit No
Is frame synchronization achieved on the demodulated signal?
Yes
Yes Decode the data
Does received ACK packet belong to me and the transmitted packet ?
No
No
Push the packet to the packet queue in order to retransmit
3 slots time unsuccessful packet/ACK transmission?
Yes
Go to Synhchronization Stage
Figure 21: The communication stage process for sensor nodes. Guard Bits (100-bit)
Frame Synchronization Bits (20-bit) Message Bits (4100-bit)
Figure 22: Overall frame format.
22
Figure 23: Field Tests: center node antenna.
Figure 24: Field Tests: Sensor node antenna.
23
Figure 25: GUI for the center node (Image is fully received and re-constructed).
24
4. Conclusions In this paper, the overall design of a data link for tactical VHF wireless sensor networks including empirical channel propagation modeling, physical and MAC layers design are presented. First of all, the empirical channel modelling for the desired communications system is developed and the large and small scale statistics of the channel are obtained. Then, the CPM-LDPC parameters are selected in order to satisfy the power and bandwidth efficiency. Second of all, the MAC and ARQ protocols are designed and implemented in the SDR platform. Then, the complete end-to-end SDR wireless sensor network system is tested in the field using NI-USRP 2920 devices. The results show that the designed data link is successfully carried an test image in the dense hilly environment with the distance up to 350 meters between two terminals. The performance results demonstrate that the proposed SDR system has great potential to operate in such harsh environments for VHF tactical wireless sensor network applications. 5. Acknowledgements This study is partially supported by Turkish Government Ministry of Science, Industry, and Technology and ASELSAN Corporation under SANTEZ project grant program. 6. References [1] Molisch AF. (2011). Wireless Communications. (2nd edn). John Wiley & Sons, Ltd. [2] Michelson DG, & Ghassemzadeh SS. (2009). Measurement and Modeling of Wireless Channels. In Ed. Vahid Tarokh, New Directions in Wireless Communications Research. Springer Science+Business Media. [3] Pasons J.D. (2000). The mobile radio propagation channel, 2nd Edition, John Wiley & Sons Ltd. [4] Cho YS, Kim J,Yang WY, & Kang CG. (2010). MIMO-OFDM wireless communications with MATLAB. John Wiley & Sons, Ltd. [5] Fontan F.P, & Espineira P.M. (2008), Modeling the wireless propagation channel: A simulation approach with MATLAB. Wiley Series on Wireless Communications and Mobile Computing. [6] Iftikhar W., & Raichl J. (2008). Channel Sounding. Masters thesis, Halmstad University. [7] Pedersen T. (2009). Contributions in Radio Channel Sounding, Modeling, and Estimation. Ph.D.Tesis, Aalborg University.
25
[8] Nambissan T. J., Nikhil T.V. Vinodkumar V.,(2015).A VHF radio for software defined radio applications. International Conference on Emerging Trends in Engineering, Science and Technology (ICETEST). [9] Bartram P., Bridges C.P., Bowman D. Shirville G. (2017). Software defined radio baseband processing for ESA ESEO mission. IEEE Aerospace Conference, 4-11 March 2017, MT, USA. [10] Chamaillard B., Lastera M. & Roque D. (2018). A flexible VHF-band aeronautical datalink receiver based on software defined radio. IEEE Aerospace and Electronic Systems Magazine 33(1) pp.58-61 January 2018. [11] Capela G., Rodrigues A., Sanguino J. & Bolas, E, (2015). Opportunistic usage of the maritime VHF band using a software defined cognitive radio. 1st URSI Atlantic Radio Science Conference (URSI AT-RASC), 2015. [12] Swarts F., & Ferreira H.C. (1991).A Comparison Between Statonary And Mobile Digital VHF radio Channels And Channel Models, Communications and Signal Processing. COMSIG 1991 Proceedings., South African Symposium, 74 79, Pretoria, 30 August 1991. [13] Berger G.L., & Safer H.(1997). Channel sounder for the tactical VHF-range. MILCOM 97 Proceedings, pp. 1474 1478, Monterey, CA, 02-05 November 1997. [14] Meng Y. S., Lee Y. H., & Ng B. C. (2010).Path loss modeling for nearground VHF radio-wave propagation through forests with tree-canopy reflection effect, Progress In Electromagnetics Research. [15] Meeks M. L. (1983). VHF Propagation over Hilly, Forested Terrain. IEEE Transactions On Antennas And Propagation, 31(3), 483-489. [16] Pugh J.A., Bultitude R.J.C, & Vigneron P.J. (2007). Propagation Measurements and Modelling for Multiband Communications on Tactical VHF Channels. IEEE Military Communications Conference, pp. 17, 29-31 October. [17] Maurcio Henrique Costa Dias, A. Rotava, F. G. Andrade, R. A. Alem, M. A. K. Melo, and J. C. A. Santos (2011). Path loss measurements of HF/VHF land links in a Brazilian Atlantic rainforest urban site. IEEE Antennas And Wireless Propagation Letters. [18] Andreas Zogg , (1987). Multipath Delay Spread in a Hilly Region at 210 MHz, IEEE Transactions On Vehicular Technology. [19] H-J Li, C-C Chen, T-Y Liu, H-C Lin, ( 2000). Applicability of Ray-Tracing Technique for the Prediction of Outdoor Characteristics. IEEE Trans. Veh. Tech.
26
[20] McKown J. W., & Hamilton R. L. Jr., (1991).Ray tracing as a design tool for radio networks. IEEE Network, 27-30, November. [21] X. Zhao, J. Kivinen, P. Vainikainen, K. Skog, (2002). Propagation Characteristics for Wideband Outdoor Mobile Communications at 5.3 GHz. IEEE Journ. Selected Areas Comm., April. [22] A. Domazetovic, L. J. Greenstein, N.B. Mandayam, I. Seskar, (2005), Propagation Models for Short-Range Wireless Channels With Predictable Path Geometries. IEEE Trans.Comm., July. [23] Al-Zaidi R., Woods J.C., Al-Khalidi M. & Hu H., (2018). Building Novel VHF-based Wireless Sensor Networks for the Internet of Marine Things. IEEE Sensors Journal, 17(8), 2018. [24] Bolli S. & Khan M.Z.A (2017). A novel RMS delay spread model for VHF/UHF bands. Progress In Electromagnetics Research Symposium, St. Petersburg, Russia, 22-25 May 2017. [25] Fischer J., Grossmann M., Felber W., Landmann M. & Heuberger A, (2012). A delay spread distribution model based on measurements of the mobile-to-mobile channel in the frequency band from 30 MHz to 400 MHz. IEEE International Conference on Wireless Information Technology and Systems (ICWITS). [26] Mohammed A. H. & Ozdemir M. K., (2017). A directional FM channel model for contemporary wireless systems. Canadian Journal of Electrical and Computer Engineering. 39(4), pp. 311-321. [27] Molina-Garcia-Pardo J.M., Lienard M., Degauqe P., (2009). Propagation in tunnels: Experimental investigations and channel modeling in a wide frequency band for MIMO applications. EURASIP Journal on Wireless Communications and Networking. [28] Aitelieva A., Celik G., Celebi H (2015). Ray tracing-based channel modeling for VHF frequency band. IEEE Signal Processing and Communications Applications (SIU 2015), Malatya, Turkey. [29] Lee K., & Ellingson S. (2008). Terrestrial propagation at LWA frequencies. [30] Andersen J.B., Rappaport T., & Yoshida S. (1995). Propagation measurements and models for wireless communications channels. IEEE Communications Magazine, 33(1), pp.42-49. [31] Gupta P., Kumar P.R. (2000). The capacity of wireless networks. IEEE Transactions on Information Theory, 46(2), 388-404. [32] Foruhandeh M., Uysal M., Altunbas I., & Guven T. (2014). Optimal choice of transmission parameters for LDPC-coded CPM. IEEE MILCOM, pp. 368371. 27
[33] Rohde & Schwarz GmbH & Co. KG. (2013). R&S SMBV100A Vector Signal Generator Operating Manual. Rohde & Schwarz GmbH & Co. KG. [34] Rohde & Schwarz GmbH & Co. KG. (2012). R&S , FSV Signal and Spectrum Analyzer Operating Manual. Rohde & Schwarz GmbH & Co. KG. [35] Dony R.D. (2001).Karhunen- Loeve Transform. In K. R. Rao and P.C. Yip. Boca Raton Editors, The Transform and Data Compression Handbook, CRC Press LLC. [36] Burton H. O., & Sullivan D. D. (1972). Errors and error control. Proceedings of the IEEE, 60(11), pp. 1293-1301. [37] Lu, D. L., & Chang, J. F. (1993). Performance of ARQ protocols in nonindependent channel errors. IEEE Transactions on Communications, 41(5), pp. 721-730. [38] National Instruments (2017). USRP-2920 http://www.ni.com/pdf/manuals/375839c.pdf
28
Specifications,
