Modulation schemes in ambient backscatter communication

Transcription

1 IT Examensarbete 30 hp Oktober 2017 Modulation schemes in ambient backscatter communication Oliver Harms Institutionen för informationsteknologi Department of Information Technology

2

3 Abstract Modulation schemes in ambient backscatter communication Oliver Harms Teknisk- naturvetenskaplig fakultet UTH-enheten Besöksadress: Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0 Postadress: Box Uppsala Telefon: Telefax: Hemsida: This thesis presents a study of different modulation schemes in the context of backscatter communication. Backscatter communication is a way of wireless communication where no active signal is transmitted. Instead surrounding signals are modified to transmit data. The goal of this thesis is to explore in how far different modulation schemes in combination with off-the-shelf hardware can be used to tackle the current data rate and distance limitations of backscatter systems. This thesis compares the modulation schemes on-off keying (OOK) and frequency-shift keying (FSK) using a constant carrier signal as well as a digital television signal. For the use of a constant carrier signal it is shown that high ranges of up to 225 meters in a line-of-sight environment and up to 30 meters in a non line-of sight environment are reachable extending the current distance limitations by far and even the use of high data rates lead to a range of 175 meters. Moreover, this thesis shows the feasibility of replacing the constant carrier with a television signal and achieves ranges of over a meter in surroundings of television signals with a signal strength of not more than -70 dbm. Handledare: Ambuj Varshney Ämnesgranskare: Christian Rohner Examinator: Arnold Neville Pears IT Tryckt av: Reprocentralen ITC

4

5 Contents 1 Introduction Motivation Problem Statement & Goals Methodology Part I Part II Results Thesis Structure Related Work 5 3 Background Backscatter Communication Signal Strength in Backscatter Communication Hardware Software Defined Radio Beaglebone Black Backscatter Module CC Arduino Zero Modulation Schemes OOK FSK MSK Design & Implementation Carrier generator Backscatter unit Implementation Receiver unit Design Implementation Configuration calculation Experiments & Results Experiment Setup Minimal shifting frequency V

6 5.3 Comparison of deviations Comparison of OOK and FSK Line-of-sight experiment Non line-of-sight experiment Television Signals DVB-T Signal Design & Implementation changes Signal generator Backscatter unit Receiver Comparison of Carrier Signals Experiment Setup FSK comparison OOK comparison OOK comparison of ranges achievable with a TV carrier Conclusion & Future Work Part I Part II Future Work VI

7 List of Figures 3.1 Distances between physical components Signal strength in a backscatter system CC2500 Packet Format Modulation of an OOK signal Modulation of an FSK signal Modulation of an MSK signal C program Assembly programs Arduino program Noise Floor around carrier frequency Bit Error Rate for FSK with different deviations Comparison of different baud rates for OOK and FSK Line-of-sight experiment for 2.9 kbaud FSK with backscatter tag close to the carrier generator Line-of-sight experiment for 2.9 kbaud FSK with backscatter tag close to the receiver Line-of-sight experiment for 197 kbaud FSK with backscatter tag close to the carrier generator Map with locations for the non line-of-sight experiment Non line-of-sight experiment for 2.9 kbaud FSK Recorded TV Signal Replayed TV Signal Shifted TV Signal for OOK and FSK Comparison of carrier signals using FSK Comparison of carrier signals using OOK Comparison of carrier signals using OOK for different distances VII

8

9 1 Introduction Wireless communication is more and more important in today s connected world. It is used to transmit information without the need of a wired connection. However, actively generating radio signals as it is done in most wireless communication systems uses a lot of energy which is not compatible with new areas of wireless communication. Especially, internet of things sensors should be able to run on a single battery for a long time or function even entirely without batteries. To achieve this, a new way of wireless communication called ambient backscatter communication is actively researched. Ambient backscatter communication does not actively generate radio signals. Instead it uses ambient signals and modifies those to transmit data. There are different ways of backscatter communication already available, but they need expensive hardware and are limited to a short range. This thesis explores how different modulation schemes supported by off-the-shelf hardware can be used in backscatter communication and how they perform regarding range and data rate. This chapter gives a motivation of building such a backscatter system, the goals of the project and a summary of the main results. 1.1 Motivation Ambient backscatter is an approach of wireless communication which uses existing radio frequency (RF) signals, such as radio, television and WiFi, to transmit data without the expenses of generating RF signals. Data is encoded by reflecting or not reflecting radio signals, thus modulating data on an existing signal. The waiver of generating RF signals is a benefit compared to traditional wireless communication systems. The main benefit is the highly reduced complexity of the system. Instead of a complex radio, only a microcontroller operating a switch and an antenna is needed. Moreover, a backscatter system has a significantly lower energy consumption compared to a system generating RF signals and might run on batteries for years or even allow harvesting all needed energy. This makes a backscatter system usable in a wide field of applications such as outdoor sensor nodes or body monitoring equipment where a replacement of a battery is difficult or even impossible. Different applications in which ambient backscatter communication is a good approach compared to traditional wireless communication are: sensors deployed in hardly reachable environments medical implants [11, 23] 1

10 Internet of Things (IoT) devices [12, 17] One major drawback of existing backscatter systems is its limitations in range and data rate. Therefore, the major motivation of this thesis is to explore alternative modulation schemes to those currently used and analyse their potential to address current limitations in range and data rate. 1.2 Problem Statement & Goals As the development on ambient backscatter continues the data rate and range still have to improve to be widely used. Most of the backscatter systems already used modulate its information on the carrier wave making it complicated for the receiver to extract the backscattered information from the received signal and leading to low data rates and ranges. Some of the systems presented in recent papers already achieve high data rates such as Passive WiFi [12] and Inter-Technology Backscatter [11] but still have a pretty low communication range. They achieve the higher data rate by separating the carrier signal and the backscattered signal making it easier for the receiver to decode the information and allowing higher data rates; however, complex modulation schemes are used to create WiFi compliant signals. The main focus of this thesis is to use off-the-shelf radios to build a system not requiring any expensive hardware and study different modulation schemes like on-off keying and frequency-shift keying in terms of range and data rate. 1.3 Methodology To achieve the goals of this thesis the following methodology is used which can be divided into two parts. In the first part a constant carrier is used as ambient signal and the focus of this part is the implementation and the study of the modulation schemes (frequencyshift keying and on-off keying) itself. In the second part the use of a television signal as carrier is introduced and the focus lays in the comparison of the two different carrier signals regarding error rate and range Part I The methodology of the first part is the implementation and testing of an ambient backscatter system using off-the-shelf hardware and a controlled carrier including the following steps: On-off keying (OOK) implementation Frequency-shift keying (FSK) implementation Experiments regarding comparison of the modulation schemes Experiments regarding range and data rate 2

11 1.3.2 Part II The methodology of the second part is to compare the performance of the implemented backscatter system of part one for different carrier signals including the following steps: Study of the signal spectrum of a television signal Experiments using a television signal as carrier Comparison of the carrier signals regarding error rate and range 1.4 Results The results of Part I show a generally better performance for FSK with outdoor ranges of up to 225 meters at a baud rate 2.9 kbaud and up to 175 meters at 195 kbaud and indoor ranges of up to 30 meters at 2.9 kbaud. This shows that the used hardware and the implementation of this system is capable of extending current limitations in backscatter systems. For Part II the results show the general usability of television signals as carrier signal for the presented backscatter system with ranges of over a meter with a bit error rate well below This also shows an improvement regarding distance and error rate compared to previous systems. 1.5 Thesis Structure Chapter 1 (Introduction) containstheintroductiontothisthesisanditsmotivation. Moreover, the problem, the goals and the main results of this thesis are addressed. In Chapter 2 (Related Work )different publications with related topics to this project are summarized including their results. Chapter 3 (Background) providesusefulinformationsforthereaderofthisthesis. It contains an introduction into backscatter communication and its range constraints as well as informations on the hardware used for this thesis. Furthermore, the considered modulation schemes are explained. In Chapter 4 (Design & Implementation) thedesignofthedifferent components of the backscatter system and its implementations are presented. The presented components are the carrier generator, the backscatter unit and the receiver unit. Moreover, a mathematical description of the configuration algorithm is given which is used to find a suitable configuration for the different modulation schemes working for both the backscatter unit and the receiver unit. Chapter 5 (Experiments & Results) containstheevaluationofthefirstpartofthe thesis. It contains the experiments and their results to find the best working configuration of the system. Moreover, it presents the experiments performed in different environments and its results regarding range and data rate. 3

12 Chapter 6 (Television Signals)is the first chapter of the second part of this thesis. This chapter contains a description of digital television signals as well as the required design and implementation changes to be able to use the previously used backscatter system with television signals as carrier. Chapter 7 (Comparison of Carrier Signals) contains the evaluation of the second part of this thesis. The experiments as well as its results comparing the two different carrier signals regarding error rate and range are presented. In Chapter 8 (Conclusion & Future Work) theresultsofthetwopartsofthisthesis are summarized and future areas of research regarding the work done in this thesis are presented. 4

13 2 Related Work The topic of this thesis is related to some recent work in ambient backscatter communication. Recent publications use different ambient signals as well as frequency shifting to build backscatter systems for multiple different application areas. A focus of most publications is to achieve higher ranges and data rates. Some of these recent publications are summarised in this chapter. In Ambient Backscatter: Wireless Communication out of Thin Air [14] the authors present a battery free backscatter system which uses as its only source of power and as its carrier signal widely available television signals. Their system enables communication between backscatter tags with a bit rate of 1 kbps over a distance of 2.5 meters outdoors and 1.5 meters indoors. In Wi-Fi Backscatter: Internet Connectivity for RF-Powered Devices [13] the authors present a backscatter system enabled to communicate directly with off-the-shelf WiFi infrastructure and thus connecting a backscatter system to the internet. It reuses existing WiFi signals as power source and modifies those signals for transmitting data. The transmitted data, with a data rate of 1 kbps, can be decoded based on variations in signal strength at a distance of up to 2.1 meters. In Every Smart Phone is a Backscatter Reader: Modulated Backscatter Compatibility with Bluetooth 4.0 Low Energy (BLE) Devices [9] the authors present a backscatter tag which is able to create Bluetooth 4.0 Low Energy (BLE) packets that are indistinguishable from conventional BLE advertising packets and therefore decodable by every modern smartphone. For backscattering those signals at a data rate of 1 Mbps a constant carrier with a signal strength of 15 dbm is used giving the backscattered signal a receivable range of 9.4 meters with an energy consumption of the backscatter tag over 100 times lower than the energy consumption of a conventional BLE transmitter. In Passive Wi-Fi: Bringing Low Power to Wi-Fi Transmissions [12] the authors present a backscatter system creating b WiFi packets decodable on any WiFi device. The presented system is able to coexist with other devices in the ISM band due to a network device doing carrier sense and transmitting the carrier for the backscatter tags. The work achieved a range of feet in non line-of-sight and line-of-sight environments with 4 to 5 orders of magnitude lower energy consumption compared to existing WiFi chipsets. In Inter-Technology Backscatter: Towards Internet Connectivity for Implanted Devices [11] the authors present a backscatter system backscattering transmissions of one wireless technology (Bluetooth) to create signals compatible to those of another wireless technology (WiFi or Zigbee). This technology called interscatter is intended to be used in implanted devices like smart contact lenses communicating with a smartphone. Therefore Bluetooth advertising packets are used as carrier signal and a single sided 5

14 backscatter signal meeting the b WiFi standard is created which can be received by a smartphone. In Enabling Practical Backscatter Communication for On-body Sensors [23] the authors present ultra-low power on-body sensors using backscatter communication working together with commercial WiFi and Bluetooth radios. They use a system where a smartphone acts as carrier generator and another wearable device like a wristband receives the information of the on-body sensor and extend that to a system with multiple transmitters or receivers. They deal with the problems of errors introduced by body movement and the energy consumption used for shifting the signal. Their results show a coverable distance of up to 4.8 meters with a data rate of 50 kbps consuming not more than 45 µw of energy. In Augmenting IoT Networks with Backscatter-Enabled Passive Sensor Tags [17] the authors present a concept of adding battery free passive sensing capabilities to existing IoT deployments without the need to modify those deployments. The sensor tags are designed to collect sensor readings and transmit those to nearby active IoT devices using backscatter communication. The carrier for those tags is generated by one of the existing devices which removes the need for an external carrier generator. The strength of the produced carrier signal is maximally 0 dbm creating a range of the backscattered signal of 20 cm which is sufficient for the use case of close distance communication. In HitchHike: Practical Backscatter Using Commodity WiFi [24] the authors present a low-power backscatter system using codeword translation to transmit valid b WiFi signals using existing WiFi infrastructure. The system introduces a one-sided signal shift to remove the normally found copy of a backscatter system with frequency shift. The HitchHike system transmits with a data rate of 1 Mbps and covers distances of up to 54 meters in a line-of-sight environment using as little energy as 33µW. 6

15 3 Background This chapter starts with an introduction how backscatter communication works. Afterwards, the hardware used for the project is introduced, followed by a description of three different modulation schemes possible with the presented hardware. 3.1 Backscatter Communication Backscatter communication is a form of wireless communication not requiring to actively generate RF signals. Instead the radar cross-section (RCS) of the antenna of the backscatter tag is modulated to either absorb or reflect a carrier signal. This modulation of the RCS is performed by changing the impedance of the antenna circuit between two states. To communicate on a frequency different to the frequency of the carrier signal, a shifting operation in addition to the RCS modulation is performed. To achieve this, the carrier signal is multiplied with a square wave generated by the backscatter tag. The frequency of this square wave is the required offset f, betweenthecarrierfrequencyf c and the desired frequency. The square wave can be written as a Fourier series as shown in Equation 3.1. S tag ( ft)= 4 1X n=1,3,5,.. 1 sin(2 n ft) (3.1) n Multiplying the carrier signal S c = sin(2 f c t) with this square wave leads to the resulting signal r(t) in Equation 3.2 [12]. r(t) = S c S tag ( ft) = sin(2 f c t) S tag ( ft) = 4 1X 1 n sin(2 n ft) sin(2 f ct) n=1,3,5,.. = 2 1X 1 n {cos(2 (f c n f)t) cos(2 (f c + n f)t)} (3.2) n=1,3,5,.. Using the first harmonic (n =1) of Equation 3.1 results in two signals created, by the performed multiplication in Equation 3.2, at an offset of f on both sides of the carrier signal s frequency f c. Thus, the backscattered signal is shifted to the desired frequency f c + t and a copy of it to the frequency f c f.[12,24] 7

16 3.1.1 Signal Strength in Backscatter Communication In a backscatter system consisting of three components, a carrier generator, the backscatter tag a receiver, the communication range depends on two parameters. These two parameters are the distance between the carrier generator and the backscatter tag (d 1 ) as well as the distance between the backscatter tag and the receiver (d 2 ). Figure 3.1: Distances between physical components The signal strength at the receiver in free space (P r )canbemodelledusingfriispath loss as given in Equation 3.3. Pt G 2 t G r P r = K (3.3) 4 d d 2 24 The equation consists of three parts, where the first term in the first parenthesis contains the signal propagation from the carrier generator to the backscatter tag with P t being the transmit power of the carrier generator and G t its antenna gain. Similar to that, the third term in the second parenthesis describes the signal propagation from the backscatter tag to the receiver with the antenna gain at the receiver G r and the wavelength of the transmitted RF signal. The second term of the equation, the factor K, isaconstantaccountingforthereturnlossandantennagainoftheusedbackscatter tag. Figure 3.2: Signal strength in a backscatter system Figure 3.2 illustrates how the general curve of the Friis model looks like for fixed carrier generator and receiver positions and a variable placement of the backscatter tag. This shows that the highest achievable ranges in a backscatter system are those where the backscatter tag is located close to the carrier generator or to the receiver. [12, 16] 8

17 3.2 Hardware This section describes the hardware components needed to build the backscatter system presented in this thesis. These are a Software Defined Radio as carrier generator, a Beaglebone Black in combination with a backscatter module and a receiver (CC2500) in combination with an Arduino Software Defined Radio A Software Defined Radio or SDR is a "radio in which some or all of the physical layer functions are Software Defined" [5]. That means that any device which can transmit or receive radio frequency signals with at least one parameter changeable by software is a Software Defined Radio. Examples of those parameters important for this thesis are frequency, transmit power and signal amplitude. The SDR used in this project is the USRP B200 by Ettus Research [20]. It is used to generate a simple carrier and to record and replay a television signal Beaglebone Black The Beaglebone Black is a development platform like the Raspberry Pi with GPIO connectors to interface with external hardware. The speciality of the Beaglebone and the reason for its usage in this project are the on-board Programmable Real-time Units [3]. The Programmable Real-time Units (PRU) are 32-bit microcontrollers which are part of the Texas Instruments AM3358 processor. The PRU is a microcontroller running at 200 MHz with single-cycle access to some of the pins and access to the memory of the processor. Because of the real-time behaviour of this chip and its high frequency, the Beaglebone Black is an optimal development platform for this project [4] Backscatter Module The backscatter module is a module developed by the Uppsala Networked Objects group (UNO) at Uppsala University. It has an external antenna which can be switched on and off by external hardware through the Analog Devices HMC190BMS8 RF switch which was used in [11] as well. The power consumption of this module is 0.3 µw CC2500 The CC2500 is a transceiver by Texas Instruments operating in the 2.4GHz-ISM band (2400MHz MHz) [21]. It is a widely used transceiver and especially highly configurable regarding frequency, baud rate 1 and modulation scheme which is of importance for this project. 1 rate of symbols transmitted per second 9

18 The CC2500 used in this project is part of a ccrf click module with an PCB trace antenna and running at a voltage of 3.3 volts. [15] In this project the CC2500 is used as a receiver for the data transmitted by the backscatter tag. To be able to decode the received data, the transmitter has to use a certain packet format given by the CC2500. This packet format is explained in the following paragraph. Packet Format A CC2500 packet consists always of a preamble, sync word and data field. If the chip is configured to receive packets of variable lengths, a length byte is inserted between the sync word and the data. Moreover, it is possible to insert an address byte for communication between multiple CC2500 modules and add a checksum to the end of the packet to check whether the received data is valid. The general structure of a packet is given in image 3.3. Figure 3.3: CC2500 Packet Format [21, p. 29] The data sheet provides additional informations for the different parts of this packet format. Beginning with the preamble, which has a length of 2 to 24 bytes of alternating 1 and 0 bits. Furthermore, the sync word consists of two predefined bytes which are transmitted 0 to 2 times. [21] In the following chapters, a packet consists always of 4 preamble bytes and 4 sync bytes which are the standard settings of configurations created with SmartRF Studio 7 2 for the CC2500. A length byte is not used because only packets with a fixed data length are transmitted and therefore the data length is configured using a separate register. The address field and checksum are not used either Arduino Zero The Arduino Zero is an Arduino board using a 32-bit ARM Cortex M0+ microcontroller. It is used to communicate with the CC2500 using SPI and is chosen due the matching voltage level of its pins and the CC2500 board. Moreover, it has a USB connector to be programmed easily and to exchange data with a connected computer. [1] 2 Software by Texas Instruments to find register configurations for their RF chips, like the CC

19 3.3 Modulation Schemes The CC2500 supports different modulation schemes like the binary modulation schemes OOK, FSK and MSK. Binary modulation scheme means that every bit is transmitted as its own symbol resulting in an equality of bit rate and baud rate for those modulations schemes. In the following sections, the three mentioned modulation schemes are described OOK On-off keying (OOK) is the simplest form of Amplitude-shift keying 3 (ASK). A one is encoded as the presence of a signal and a zero is encoded as the absence of a signal [19]. OOK uses only one frequency for transmitting data and requires therefore less bandwidth than other modulation schemes. Moreover, it needs less energy than the other presented modulation schemes because it only needs to emit a signal when a one is transmitted [7]. Figure 3.4 shows an example of an OOK signal. Figure 3.4: Modulation of an OOK signal [10, p. 6] FSK Frequency-shift keying (FSK) is a modulation scheme encoding different symbols as different frequencies. The simplest form of FSK is binary frequency-shift keying (BFSK or 2-FSK) which encodes a one as one frequency and a zero as another frequency. It is less susceptible to errors than OOK because a receiver looks for specific frequencies for the different symbols and therefore noise spikes cannot cause as many errors as for OOK. [22, 10] The version of FSK used in this project is binary FSK with a continuous phase (CPFSK). Continuous phase means that there are no phase jumps at the frequency transitions. An example of such a signal can be seen in Figure 3.5. v c1 (t) and v c2 (t) 3 One and zero are represented as different amplitudes of a carrier wave. 11

20 Figure 3.5: Modulation of an FSK signal [10, p. 9] are the two signals of frequencies used to generate the respective FSK signal v FSK (t) according to the bit pattern. In this project, contrary to Figure 3.5 the higher frequency is used for encoding a one instead of a zero MSK MSK is a special form of continuous-phase frequency-shift keying (CPFSK). The speciality is that the frequency difference between transmitting a zero and a one is equal to half the data rate resulting in a difference of the waveform for transmitting a one or a zero of half a period. This difference of half a period can be seen in Figure 3.6. [18] Due to the fixed relationship between the frequency deviation and the data rate for MSK and the accompanying incomparability with the other modulation schemes for fixed deviations and data rates, MSK was not further used in this project. Figure 3.6: Modulation of an MSK signal [18] 12

21 4 Design & Implementation This chapter addresses the design used in this project. It presents the physical connection of the components presented in Chapter 3 and the software implementation used within the subsystems, in which the design can be divided. Those subsystems are the carrier generator, the backscatter unit and the receiver unit. Figure 3.1 gives an overview on the physical placement of the subsystems. Moreover, a mathematical description to calculate a configuration usable for the communication between the backscatter unit and the receiver unit is presented. 4.1 Carrier generator The SDR described in the previous chapter is connected to a computer and is used to generate a carrier wave with a constant frequency. It transmits a carrier signal at a specific frequency within the 2.4 GHz ISM band. 4.2 Backscatter unit The backscatter unit consists of the Beaglebone Black and the backscatter module described in the previous chapter. The backscatter module is connected to the Beaglebone Black using two connections. It is connected to pin P8.11 of the Beaglebone Black, which is accessible as an output by the first PRU of the Beaglebone Black [6], and to ground to have a common ground Implementation The implementation of the backscatter logic consists of two parts. A C program for writing the data to send into the shared memory of the Beaglebone s processor and starting the second program which is a program written in assembly running on the PRU-core. There are each a C program and a PRU-assembly program for OOK and FSK. The structure of the C program can be seen in Figure 4.1. Depending on its purpose (FSK or OOK) the data written to the shared memory is slightly different. For OOK the first bytes contain the number of toggles for transmitting a one, the corresponding delay time and the delay time for transmitting a zero, whereas FSK starts directly with the actual packet data. For FSK, each bit in preamble, sync word and payload is represented by the number of toggles and the delay time between two toggles. For OOK the data consists of the actual bits which are saved as one byte each for being processable by the PRU program. 13

22 14 Figure 4.1: C program

23 (a) FSK program (b) OOK program Figure 4.2: Assembly programs 15

24 The assembly programs, of which the flow charts for FSK and OOK are given in Figure 4.2a and Figure 4.2b respectively, perform the main task of the backscatter unit by toggling the antenna according to the data in the shared memory. The assembly program for FSK executes for each toggle-delay pair its program routine. This means, the antenna is toggled, followed by the predefined delay, for the number of given toggles. For OOK, the number of toggles for transmitting a one, the corresponding delay time and the delay time for transmitting a zero are read first. Afterwards, the same routine as for FSK is executed when a 1 is read from memory and a second routine is executed when a 0 is read. This second routine delays the further execution for the delay time read in the beginning of the program execution. 4.3 Receiver unit The receiver unit consists of a CC2500 module and an Arduino Zero. In the following, ashortoverviewaboutthephysicaldesignisgivenfollowedbyanexplanationofthe program implementation Design The CC2500 module is connected to the Arduino according to the following table: CC2500 3V3 GND MISO MOSI CLK CS GDO0 Arduino Zero 3V3 GND ICSP-1 ICSP-4 ICSP-3 A2 D5 To communicate between the Arduino Zero and the CC2500, SPI is used, therefore the CC2500 connectors for MOSI, MISO, clock (CLK) and chip select (CS) are connected to those configured in the SPI library for the Arduino Zero. [2] The GDO0 pin, used for notifications of incoming data from the CC2500 can be connected to any digital input pin of the Arduino, here pin 5 was chosen Implementation The CC2500 is configured using different registers. The values for those are easily configurable using SmartRF Studio 7. Some of those values have to be configured in accordance with the results of the configuration algorithm presented in section 4.4. The most important registers are those setting the centre frequency, the deviation between the two frequencies used for FSK transmission, the baud rate, the receive filter bandwidth and the modulation scheme. The frequency is configured using the registers FREQ2, FREQ1 and FREQ0. The two most significant bits (of FREQ2) arealways01. The 24 bit register value is calculated using Formula 4.1 containing the centre frequency f c and the oscillator frequency f XOSC of the CC2500 which is 26 MHz. FREQ[23 : 0] = f c 2 16 f XOSC (4.1) 16

25 Figure 4.3: Arduino program 17

26 The deviation is configured in register DEVIATN and calculated according to Formula 4.5. The value of i is configured using the three least significant bits (2, 1 and 0) of the register and the value of j using the bits 6, 5 and 4. The baud rate and the receive filter bandwidth are configured in the registers MDMCFG4 and MDMCFG3. The baud rate is calculated using Formula 4.3 with i representing the register MDMCFG3 and j representing the four least significant bits of the register MDMCFG4. The four most significant bits are used to configure the filter bandwidth. Formula 4.2 is used to calculate this bandwidth, with i being bit 5 and 4 of the register and j being bit 7 and 6. BW = f XOSC, for i, j 2 [0, 7] (4.2) 8 (4 + i) 2j The last important register is MDMCFG2. The bits 6, 5 and 4 of this register are used to configure the modulation scheme used and OOK is represented by the value 011 whereas FSK is represented by the value 000 [21]. The general algorithm for the receiver unit is shown in Figure 4.3. The main part of the algorithm consists of the following procedure. Whenever the CC2500 starts receiving a packet, the GDO0 pin is pulled up and when the complete packet is received, the pin is pulled down. This is used to notify the Arduino that a new packet is in the buffer. Afterwards the Arduino requests the data from the CC2500 and receives it. The corresponding RSSI value (signal strength) to this transmission is calculated as well as the RSSI value of the background noise (noise floor). In the end the data and the two RSSI values are transmitted to the connected computer for further processing. 4.4 Configuration calculation The formulas and equations used to calculate the settings for the backscatter unit are based on the following formulas given in the data sheet of the CC2500. The possible baud rates can be calculated with Formula 4.3: [21, p.26] R DATA = (256 + i) 2j 2 28 f XOSC, for i 2 [0, 255], j2 [0, 15] (4.3) And the possible deviations from the centre frequency with Formula 4.4: [21, p. 33] f dev = f XOSC 2 17 (8 + i) 2 j, for i, j 2 [0, 7] (4.4) The oscillator frequency f XOSC for the CC2500 module used in this thesis is 26 MHz. Therefore f XOSC will be replaced with in all further occurrences of the formulas above. Based on these formulas and the constraint that the delays of the Beaglebone can only be multiples of 10 ns with a minimum of 80 ns, Equation 4.5 can be used to calculate the possible delays for the two FSK frequencies for a deviation with an error of less than one percent. 18

27 t t 2 2 f dev < 0.01 (4.5) All times (t, t 1 and t 2 ) in this and the following equations and formulas are times in nanoseconds. The first term in the equation above is the formula for the achievable deviations using the Beaglebone whereas f dev represents the deviations possible for the CC2500 as described in Formula 4.4. Using those delays for which Equation 4.5 is true Formula 4.6 can be used to calculate the centre frequency by which the signal is shifted. f shift = t t 2 (4.6) 2 The possible baud rates for the delay times calculated with Equation 4.3 with minimum deviation from the baud rates provided by the CC2500 can be calculated with Equation 4.7. The modulo operator used in this formula and the next one is the modulo operator of the programming language Python, with support for floating point numbers R DATA t mod < " (4.7) The first term of this equation denotes the number of toggles, corresponding with a delay time t, needed to transmit one baud. This should not deviate much from a natural number. Otherwise, errors would be introduced because it would lead to an inaccurate baud rate. The equation has to be true for both delay times t 1 and t 2 using the same baud rate from Formula 4.3. The comparison value " might have different values based on the needed accuracy. Similar to Equation 4.7, Equation 4.8 is used to find reasonable baud rate configurations for OOK. OOK has one more constraint in comparison to FSK. For OOK the number of toggles, corresponding with a delay time t, needed to transmit a one has to be an even number to have the antenna again in low state after that number of toggles R DATA t mod 2 1 < " (4.8) 19

28

29 5 Experiments & Results In this chapter, the experiment setup as well as the different experiments and its results are presented. First of all, the setup used for the different experiments including the data sent to achieve comparable results is explained. Afterwards the minimal shifting frequency for sufficiently rejecting the carrier is determined followed by the determination of the best deviation needed for FSK. After comparing the two modulation schemes extended evaluations in line-of-sight and non line-of-sight environments are presented for the use of FSK. 5.1 Experiment Setup The physical setup for all experiments out of the first one (Chapter 5.2) is as shown in Figure 3.1. The position of the SDR is fixed within an experiment whereas the placement of the backscatter tag and the receiver is changed modifying the distances d 1 and d 2. Each performed experiment is divided into multiple experimental runs. Each of these experimental runs consists of sending 100 predefined random packets consisting of a one byte packet number followed by 63 bytes of random data. All experiments from section 5.3 on consist of three experimental runs with different orientations of the receiver to find the average error. For the experiments from section 5.5 onwards, several experimental runs were performed for each location and orientation to reach a higher accuracy. The carrier frequency for all of the following experiments was set to 2480 MHz for not interfering with surrounding signals. The transmit power of the SDR and the backscatter module differ between the experiments due to a change of the used antenna. For the first experiments a simple omni-directional 2.4 GHz antenna was used resulting in a transmit power of 12 dbm. From section 5.5 on instead, the SDR signal was amplified and omni-directional antennas with a higher range were used to achieve higher ranges. The amplified SDR signal has a signal strength of 26 dbm. The overall trends of the first experiments are not effected by this choice. 5.2 Minimal shifting frequency The first experiment was done to find out which frequency shift is minimally needed that the backscattered signal is minimally affected by noise created by the SDR. For this experiment only the SDR and the receiver unit were used and placed about 30 cm away from each other. The receiver was set to listen on a fixed frequency while the 21

30 Figure 5.1: Noise Floor around carrier frequency frequency of the carrier generator was changed in steps of 250 khz from 10 MHz below the receiver frequency to 10 MHz above the receiver frequency. For each frequency, the carrier generator was set to, the RSSI value at the receiver was recorded. These RSSI values represent the noise floor introduced by the carrier signal for different offsets between carrier frequency and receiver frequency. To receive a backscatter signal at the receiver, the strength of this signal has to be stronger than the noise floor. Therefore a low noise floor value is appreciated. Figure 5.1 shows that the noise floor for shifts bigger than 2 MHz changes only slightly. Therefore, the difference between a 2 MHz shift or a larger shift is only minimal which leads to a minimal frequency shift of 2 MHz used for further experiments. 5.3 Comparison of deviations Calculating all possible deviations for which Equation 4.5 is true leads to different shifting frequencies all above 2 MHz. Therefore all of them could be used. To be able to compare the different deviations, all deviations with delay times corresponding to a frequency shift of 2 to 2.5 MHz were taken into account. Due to the fact that the number of toggles for the two different frequencies for FSK has to differ by at least one, the maximum baud rate for a deviation of 41 khz is 164 kbaud. Therefore three different baud rates below 164 kbaud were chosen to have comparable results for the different deviations. These baud rates are around 2.4 kbaud, 75 kbaud and 150 kbaud. The used baud rates for each configuration were calculated in accordance with equation 4.7 with a value of 0.3 for ". The distances between the SDR and the backscatter tag and between the backscatter tag and the receiver were both chosen to be 1 meter for this experiment. The results of this experiment in Figure 5.2 shows that the best deviation is 95 khz Deviations above 95 khz show overall worse results than small deviations. Moreover, the smallest tested deviation shows worse results than a deviation of 95 khz at high baud 22

31 Figure 5.2: Bit Error Rate for FSK with different deviations rates. This is due to the difference in the number of toggles between the two frequencies. This difference is only one for the smallest baud rate which means that the difference between the frequencies of the two symbols is only half a period. For this configuration of half a period difference it is hard to find the best fitting baud rate because the exact one is not available on the CC2500. Therefore it is better to not use a difference between the toggle numbers of one. That means that the highest tested baud rate will provide better results with deviations larger than 41 khz. The following experiments are all done with the baud rate configuration of 95 khz and a frequency shift of MHz. 5.4 Comparison of OOK and FSK To compare the general performance of OOK and FSK a couple of experiments were performed inside a lab. For those experiments, the backscatter tag was placed one, two or three meters away from the carrier generator and the receiver was placed in different distances from the carrier generator, building a straight line with the carrier generator and the backscatter tag. The maximum distance coverable between the carrier generator and the receiver was 8 meters. The results of these experiments are shown in Figure 5.3. The evaluation of the results clearly shows that lower baud rates perform generally better than higher baud rates. One of the reasons for that is the sensitivity of the CC2500 which is higher for signals of low baud rates, resulting in a higher range for those signals. Furthermore, FSK performs always better than OOK which was expected due to FSK being less susceptible to noise than OOK but is also due to a higher sensitivity of the CC2500 for FSK than for OOK which can be experimentally seen. The comparison of the three figures also shows that the performance is better for a placement of the backscatter tag close to the signal generator which is in accordance with Chapter

32 (a) 1 meter distance between SDR and backscatter tag (b) 2 meters distance between SDR and backscatter tag (c) 3 meters distance between SDR and backscatter tag Figure 5.3: Comparison of different baud rates for OOK and FSK with different distances between SDR and backscatter tag 24

33 The peak with a high bit error rate for all baud rates and modulation schemes, especially at 5 meters distance in Figure 5.3a, is due to the lab environment with a large metal tube at the ceiling. At this distance the backscatter tag was placed approximately below that tube. The reflections of this tube are likely to create interferences leading to way more errors than in the neighbouring configurations. Because of the much better performance of FSK compared with OOK, the following experiments were done for FSK only. 5.5 Line-of-sight experiment The next experiments were done to ascertain the performance of FSK in line-of-sight environments. To reach the best possible performance and avoid interferences with other signals, the line-of-sight experiments were performed outdoors between the university and a forest. Figure 5.4: Line-of-sight experiment for 2.9 kbaud FSK with backscatter tag close to the carrier generator The first experiment analyses the possible range of FSK with a baud rate of 2.9 kbaud for different distances between the carrier generator and the backscatter tag. As visible from Figure 5.4 ranges of up to 225 meters are reachable. Moreover, it can be seen that even at these high distances the bit error rate is below 10 4 for a placement of the backscatter tag close to the carrier generator. Higher distances between the carrier generator and the backscatter tag lead to higher bit error rates but for the most coverable distances the bit error rate is well below Figure 5.5 shows the results of the second experiments studying the performance of FSK with the same baud rate as above for small distances between the backscatter tag and the receiver. With small distances between the backscatter tag and the receiver distances of up to 200 meters from the carrier generator are coverable. For higher distances between the backscatter tag and the receiver smaller distances between the receiver and 25

34 Figure 5.5: Line-of-sight experiment for 2.9 kbaud FSK with backscatter tag close to the receiver the carrier generator are coverable which is in accordance with the model described in Chapter The coverable ranges for the same tag distances as in Figure 5.4 are almost the same. The bit error rate is especially for the 200 meter range higher than in the previous experiment but still for the most configurations below Figure 5.6: Line-of-sight experiment for 197 kbaud FSK with backscatter tag close to the carrier generator The last figure (Figure 5.6) for line-of-sight experiments shows the experiments done for the use of a high baud rate (197 kbaud). Using this baud rate, a shorter maximum range of 175 meters is coverable. Moreover, the bit error rate is significantly higher for the same distance configurations compared with the usage of a low baud rate of 2.9 kbaud. Generally it is possible to reach high distances with a high baud rate but 26

35 this should only be considered if a high amount of data has to be transmitted and in combination with a strong error correction algorithm. 5.6 Non line-of-sight experiment The previous experiment showed the distances possible in an outdoor environment. Here the focus lays on the performance of 2.9 kbaud FSK in a non line-of-sight indoor environment. Therefore the carrier generator and the backscatter tag are placed in the same room with different distances between each other and the receiver is placed in a different room as seen in Figure 5.7. Figure 5.7: Map with locations for the non line-of-sight experiment Figure 5.8 shows the results of this experiment. The vertical lines show the positions of the walls in the test environment. This experiment shows as well as the previous ones better performance for short distances between the carrier generator and the backscatter tag. Especially, for a distance of 6 meters between the carrier generator and the backscatter tag the signal-to-noise ratio is close to 0 resulting in a maximal range of 20 meters communicating through 4 walls whereas a maximal range of 30 meters, communicating through 8 walls is possible for a carrier generator-tag distance of 1 meter. The bit error rate for this experiment is for almost all cases below Figure 5.8: Non line-of-sight experiment for 2.9 kbaud FSK 27

36

37 6 Television Signals The second part of this thesis deals with a study whether television signals can be used as carrier signal for the previously described communication system. A use of TV signals instead of a constant carrier signal would make the backscatter system being usable in a much wider variety of situations and save a lot of energy and cost by not needing extra hardware for generating a specific carrier wave. This chapter deals with properties of DVB-T television signals and the design and implementation changes needed to use those signals as carrier signals. 6.1 DVB-T Signal Terrestrial television signals are available in the most places and transmitted continuously. Therefore they are theoretically usable as a carrier signal. The terrestrial TV signals available in Europe and many other places worldwide are DVB-T signals using the signal format OFDM 1 with a maximum channel width of 8 MHz. [14, 8] Figure 6.1: Recorded TV Signal A 15 second sample of such a signal with a centre frequency of 594 MHz was recorded with the SDR used in this thesis to replay it as carrier signal. The 20 MHz wide signal was recorded using the SDR set to a sample rate of 20 million, in combination with an antenna suitable for Sub-GHz signals. An averaged FFT-plot of the recorded signal is shown in Figure Orthogonal Frequency Division Multiplex: large number of close spaced carriers, orthogonal to each other to avoid interferences. [8] 29

38 6.2 Design & Implementation changes In this section the relevant changes performed to use a TV signal as a carrier with the previously used hardware are described Signal generator For replaying the signal using the SDR, a program using GNURadio was developed. The program consists of a file source connected to a USRP sink for outputting the signal. The parameters needed are the sample rate which is the same as for recording the signal and the centre frequency around which the signal should be replayed. The centre frequency was chosen to be inside the 2.4 GHz ISM band to be able to use the same receiver as in the previous part of this thesis and to be able to compare the performance of the different modulation schemes for different signals used as carrier. The frequency chosen is 2476 MHz to achieve that the right edge of the replayed TV signal is at approximately 2480 MHz which is the same frequency used for the carrier signal in the previous part of this thesis. The FFT-plot in Figure 6.2 shows the average replayed signal. Figure 6.2: Replayed TV Signal Backscatter unit For the OOK program running on the Beagelobone Black no changes were needed. Only for FSK some changes were necessary. The approach of shifting the carrier signal by small different amounts like in the previous chapters was not possible because the higher frequency shift for transmitting a one would lead to a signal not only at the higher of the two frequencies the receiver is listening on but also at the lower frequency due to the width of the TV signal. Therefore the program was changed to allow smaller delay times than 80 ns and make a larger frequency shift possible. The chosen frequency shift for transmitting a one was chosen to be 10 MHz. The shifting frequency for transmitting a zero was set to 2.5 MHz to be still above the minimum shifting frequency and have the 30

39 (a) OOK: TV Signal (blue) and shifted Signal (red) (b) FSK: TV Signal (blue), shifted Signal transmitting 0 (red), shifted Signal transmitting 1 (green) Figure 6.3: Shifted TV Signal for OOK and FSK edges of the shifted signals at the right positions to reuse the previously used deviation of 95 khz. An FFT-plot with the carrier signal and the shifted signals for FSK and OOK can be seen in Figure Receiver For the receiver program only minor changes had to be made. First of all the centre frequencies for using a TV Signal as carrier had to be adjusted. The other adjustment is the filter bandwidth of the CC2500 which was set to its maximum value of khz instead of 325 khz for using FSK with a TV Signal as carrier. This adjustment was made to better receive the FSK signal after testing with the old value first. 31