UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Data communications using coherent minimum frequency shift keying on intrabuilding polyphase power.. Chiu, Frank Kwok King 1985

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
UBC_1986_A7 C48.pdf [ 13.55MB ]
[if-you-see-this-DO-NOT-CLICK]
Metadata
JSON: 1.0064794.json
JSON-LD: 1.0064794+ld.json
RDF/XML (Pretty): 1.0064794.xml
RDF/JSON: 1.0064794+rdf.json
Turtle: 1.0064794+rdf-turtle.txt
N-Triples: 1.0064794+rdf-ntriples.txt
Original Record: 1.0064794 +original-record.json
Full Text
1.0064794.txt
Citation
1.0064794.ris

Full Text

DATA COMMUNICATIONS USING COHERENT MINIMUM FREQUENCY SHIFT KEYING ON INTRABUILDING POLYPHASE POWER LINE NETWORKS by Frank Kwok King Chiu B.A.Sc University of British Columbia, 1983 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES (Department of Electrical Engineering) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA December, 1985 (E) Frank Kwok King Chiu, 1985 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of £/ec^c*J? Zmp^eerjy, The University of British Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 Date ABSTRACT The suitability of Coherent Minimum Frequency Shift Keying (CMFSK) modulation for data communications on polyphase intrabuilding power distribution circuits is examined. An actual modem was designed and implemented. Average bit error rate (BER) versus received Eb/No measurements were taken for an industrial, commercial, and residential power line environments at 1.2kbps, 4.8kbps, and 19.2kbps data rate. The 19.2kbps BER measurements indicate that a majority of errors are caused by impulses occurring in the power lines, while other errors are caused by momentary reductions of received Eb/No. Occurance of errors coincides mostly with impulses on the power line which are highly periodic with the ac mains voltage. In addition, the BER measurements reveal that CMFSK modulation at 1.2kbps and 4.8kbps data rate is less affected by impulse noise than at 19.2kbps. This finding is attributed to the increased resistance to impulse noise effects as the bit duration is increased. A baseband spectrum spreading technique is proposed and successfully tested to implement low data rate transmissions. Spread spectrum signalling overcomes potential narrow band impairments by sending a wide1 band signal over the power lines. In addition, the reduced power spectral density of the spread spectrum transmission reduces narrow band interference to other power line communications users as well as AM radios and allows higher output power to compensate for path attenuations. — i i i — TABLES OF CONTENTS ABSTRACT i i TABLES OF CONTENTS iii LIST OF TABLES v LIST OF ILLUSTRATIONS vi i ACKNOWLEDGMENTS x 1. INTRODUCTION 1 1.1 Communications over Electric Power Lines .... 1 1.2 Outline of Thesis 4 2. DISCUSSION OF COMMERCIAL POWER LINE MODEMS 6 2.1 BSR X-10, A Remote Control for Lights and Applicances 6 2.2 NON-WIRE Power Line Modem 7 2.3 ExpertNets Power Line Modem 8 2.4 Consultant's Choice Power Line Modem 8 2.5 Commercial Power Line Modem Technology3. POWER LINE TRANSMISSION CHARACTERISTICS ....... 11 3.1 Industrial Building Power Line Transmission Characteristics 13.2 Residential Building Power Line Transmission Characteristics 25 4. POWER LINE NOISE 9 4.1 Power Line Noise Characteristics . 29 4.2 Cross Sectional Power Line Noise Characteristics . 33 5. CMFSK POWER LINE MODEM DESIGN CONSIDERATIONS 35 5.1 Power Line Modem Design Criterion ... 35 5.2 Design Objectives of a CMFSK Power Line Modem . . 37 5.3 Design of a CMFSK Power Line Modem 39 5.4 CMFSK Modem Transmitter Description 41 5.5 CMFSK Modem Receiver Description .6. PRELIMINARY CMFSK TESTS TO SELECT SIGNALLING PARAMETERS 47 6.1 CMFSK Bit Error Rates in AWGn 46.2 Spectral Comparison of an PN Sequence and its CMFSK Output 46.3 CMFSK Carrier Frequency Selection ... 51 — i v — 6.4 BER Measurement -for CMFSK in a 'Remote' Reception 56 7. BER MEASUREMENT RESULTS IN VARIOUS ENVIRONMENTS AND AT VARIOUS DATA RATES 59 7.1 Interfloor/Interphase Industrial BER measurements at 1.2kbps, 4.8kbps, and 19.2kbps 59 7.2 Interfloor/Interphase Apartment BER measurements at 19.2kbps 60 7.3 Interfloor/Interphase Residential BER measurements at 19.2kbps 1 7.4 Cross Sectional BER Performance at 19.2kbps in an Industrial Building 67.5 Classification of Power Line Induced Errors ... 74 7.6. Discussion on Power Line Impulse Noise Suppression Techni ques 78. SPECTRUM SPREADING 83 8.1 Discussion of Spread Spectrum in Power Line Data Communications ...... 83 8.2 Potential Spread Spectrum Applications in Office and Residential Power Line Communications .... 87 8.3 Benefits of Spread Spectrum in Low Data Rate Communications 89 8.4 Application of Databand Spread Spectrum in the CMFSK Power Line Modem 91 8.5 Code Synchronization using a Digital Matched Filter 93 8.6 Tolerance to Power Line Small Error Bursts by Interleaving ..... 94 8.7 Experimental Results for Interleaved Spread Spectrum Data Transmissions 96 9. CONCLUSIONS 105 9. 1 Summary9.2 Cost Estimate of the CMFSK Modem 107 9.3 Recommendations for Further Research 10REFERENCES 110 BIBLIOGRAPHY 5 APPENDIX A. Power Line Noise Spectral Density Determination 116 APPENDIX B. Schematics of CMFSK Transmitter and Receiver 123 V LIST OF TABLES Table 4.1 Industrial Power Line Noise Spectral Density . . 30 Table 4.2 Residential Power Line Noise Spectral Density . 30 Table 4.3 Industrial Power Line Noise Spectral Density on Four High Activity and Four Low Activity Days . 34 Table 5.1 Coupled Power into the Power Line @ 120kHz ... 43 Table 6.1 1.2kbps, 4.8kbps, and 19.2kbps BER results in Additive White Gaussian Noise 48 Table 6.2 'Remote' Interphase 19.2kbps BER versus Eb/No at Transmitter Output of IV, 2V, 3V, and 4V at Two Different Carrier Frequencies of 60kHz and 120kHz 55 Table 7.1 Interfloor/Interphase 19.2kbps BER versus Eb/No at Transmitter Output of IV, 2V, 3V, and 4V in an Industrial Building 62 Table 7.2 Interfloor/Interphase 4.8kbps BER versus Eb/No at Transmitter Output of .IV, i.2V, .4V, and 2.5V in an Industrial Building 64 Table 7.3 Interf1oor/Interphase 1.2kbps BER versus Eb/No at Transmitter Output of 150mV and 200mV in an Industrial Building ...... 66 Table 7.4 Interfloor/Interphase 19.2kbps BER versus Eb/No at Various Transmitter Output in an Apartment Complex .....*.. 67 Table 7.5 Interf1oor/Interphase 19.2kbps BER versus Eb/No at Various Transmitter Output in a Residential House with a Hair Dryer 'ON' and 'OFF' 68 Table 7.6 Cross Sectional 19.2kbps BER Measurement ... 75 Table 7.7 19.2kbps BER Measurement of a 4-Level FSK and the CMFSK Modem 79 Table 8.1 31—Bit Interleaver/De-interleaver Connections . 98 Table A.l Parameters of 6th Order Band Pass Filters . . . 118 Table A.2 Parameters of 4th Order Band Pass Filters ... 118 Table A.3 Frequency Response of Power Line Coupling Terminated at 100k ohm 118 - vi -Table A.4 Frequency Response of Power Line Coupling Terminated at 12 ohm 119 Table A.5 Frequency Response of Power Line Coupling Terminated at 3.3 ohm ............. 120 Table A.6 Measured Noise Voltage in LOW, MID, and HIBH Frequency Band in an Industrial Building ... 121 Table A.7 Measured Noise Voltage in LOW, MID, and HIBH Frequency Band in a Residential House 122 - vi i — LIST OF ILLUSTRATIONS Fig. 3.1 'Local' and 'Remote' Transmission Curves on a Weekday Between 8:30am and 4:30pm in an Industrial Building 14 Fig. 3.2 'Local' and 'Remote' Transmission Curves on a Weekday Between 8:30am and 4:30pm in an Industrial Building ..... 15 Fig. 3.3 'Local' and 'Remote' Transmission Curves on a Weekday After 6:00pm in an Industrial Building . 16 Fig. 3.4 'Local' and 'Remote' Transmission Curves on a Weekend Day 17 Fig. 3.5 Pictures of Received 'Local' and 'Remote' Signals at Various Frequencies 18 Fig. 3.6 0° and 180° Phase Transmission Curves in a Residential House 26 Fig. 3.7 Transmission Difference between the 0° and 180° Phase With and Without a Resistive Signal Bypass 27 Fig. 3.8 Extended 0° and 180° Phase Transmission Curves of the Residential House 28 Fig. 4.1 Power Line Noise in Three Frequency Bands ... 31 Fig. 4.2 Power Line Noise Spectral Density 32 Fig. 5.1 Block Diagram of CMFSK Transmitter/Receiver . . 40 Fig. 5.2 CMFSK Modem Power Output Measurement Method . . 42 Fig. 5.3 Block Diagram of Digital Integrate/Dump .... 44 Fig. 6.1 CMFSK Modem White Noise BER Measurement Method . 48 Fig. 6.2 CMFSK Modem White Noise 19.2kbps BER Performance 49 Fig. 6.3 Spectrum of a 19.2kbps, Length 2047 PN Sequence . 50 Fig. 6.4 Spectrum of a 19.2kbps PN code CMFSK Output . . 50 Fig. 6.5 Interphase 19.2kbps BER versus Eb/No curve at a Carrier Frequency of 60kHz 52 Fig. 6.6 Interphase 19.2kbps BER versus Eb/No curve at a Carrier Frequency of 120kHz .... 53 - viii -Fig. 6.7 Spectrum of CMFSK output Superimposed on Power Line Noise 54 Fig. 6.8 'Remote' Interphase 19.2kbps BER versus Eb/No at a Carrier Frequency of 120kHz 57 Fig. 7.1 19.2kbps BER versus Eb/No Measurement in an Industrial Building .............. 69 Fig. 7.2 4.8kbps BER versus Eb/No Measurement in an Industrial Building 70 Fig. 7.3 1.2kbps BER versus Eb/No Measurement in an Industrial Building 1 Fig. 7.4 19.2kbps BER versus Eb/No Measurement in an Apartment Complex 72 Fig. 7.5 19.2kbps BER versus Eb/No Measurement in an Residential House 73 Fig. 7.6 Impulse Noise Induced Bit Errors ....... 76 Fig. 7.7 Comparison of 19.2kbps BER of a 4-level FSK and the CMFSK Modem 80 Fig. 7.8 Block Diagram of 4-Level FSK Modulator 81 Fig. 7.9 Block Diagram of 4-Level FSK Demodulator .... 82 Fig. 8.1 Proposed High Frequency Bypassing and Isolation of a Power Line Network 88 Fig. 8.2 Proposed Across-Phase Modem Relay ....... 88 Fig. 8.3 Spectral Comparison of BSR X-10 and CMFSK Modem . 90 Fig. 8.4 Block Diagram of Databand Spread Spectrum ... 92 Fig. 8.5 Block Diagram of Digital Matched Filter .... 95 Fig. 8.6 Block Diagram of Inter1eaver/De-inter1eaver ... 97 Fig. 8.7 Block Diagram of Simple Majority Logic Decision, Interleaved Majority Logic Decision, and Correlation Decision Methods 100 Fig. 8.8 Comparison of 1.2kbps BER Performance of the Interleaved Majority Decision, Simple Majority Decision, Correlation Decision, and Narrow Band Methods 101 — ix -Fig. 8.9 Comparison o-f 4.8kbps BER Performance of the Interleaved Majority Decision, Simple Majority Decision, Correlation Decision, and Narrow Band Methods 102 Fig. A.l A Simple Spectrum Analyzer used in Power Line Noise Spectral Density Measurement 117 Fig. B.1 Detailed Transmitter Block Diagram 124 Fig. B.2 Detailed Receiver Block Diagram ... 125 Fig. B.3 Schematics of Transmitter 126 Fig. B.4 Schematics of Receiver - Input Coupling, 4th Order BPF, and FSK Demodulator 127 Fig. B.5 Schematics of Receiver - Hard Li miter, Edge Detector, Pulse Stretcher, and Bit Clock Recovery 128 Fig. B.6 Schematics of Receiver - Digital Integrate/Dump, Threshold Decision, and Data/Data Clock Output . 129 ACKNOWLEDGEMENTS I would like to thank Dr. R.W. Donaldson -For his supervision and Mr. M. Chan -for his assistance in the modem BER measurements. In addition, I would like to thank Dr. C.S.K. Leung -for providing the equipment which made the measurements possible. Financial support during this work was provided by NSERC Postgraduate Scholarships. 1 1. INTRODUCTION 1.1 Communications over Electric Power Lines Over the years power distribution lines have been used by the power industry -for telecommunication purposes for remote metering, automatic load management, and other applications C1D—C53. The aim is to achieve automatic control of electric power distribution. Studies have been performed to measure the transmission and noise characteristics found in power lines C53-C103. These studies demonstrate that an usable power line bandwidth lies between 1kHz and 10kHz. Above 10kHz, the power distribution line cables, power transformers, associated distribution feeders, power factor correction capacitors, and other equipment effectively form a sharp rolloff low pass filter. The most frequently encountered noise is related to and synchronized with the 60Hz mains voltage. It is called harmonic noise and its spectrum shows discrete lines at multiples of 60Hz. In order to satisfy the control and monitoring functions required by an ultility company, data rates of less than 100bps are sufficient. There are two common modulation schemes employed in a power line carrier (PLC) or distribution line carrier (DLC) communication systems. They are PSK and DPSK modulation methods. In either scheme, the carrier is locked or synchronized to a harmonic of ac line frequency. In this way, the spectrum of the transmission (30Hz for 15bps, 150Hz for 75bps) is centered between two harmonics. The receiver is also synchronized to the ac mains voltage, and therefore, can provide excellent harmonic - 2 -noise rejection. The resultant narrow band signal detection is performed in essentially white noise; and with adequate carrier recovery, both the PSK and DPSK systems are superior to other modulation techniques in terms of achieving the lowest average bit error rate (BER) at the same received Eb/No. Clinard CI13 and O'Neal C53 described the parameters of several experimental and commercial PLC systems. Most of these systems have data rates below 100bps and can easily satisfy present communication requirements. It has been suggested that dedicated networks, such as fibre optics and satellite links, would carry an increasing part of future communication needs for power—distribution automation C2D. Recent developments in local area networks and proliferation of microprocessor controlled equipment motivate use of the power lines as a communication medium for achieving intrabuilding local area networking. There are three obvious advantages: 1. Elimination of a custom installed communication networks, which could be costly if coverage exceeds several rooms and floors. 2. Ease of access into the network through a standard electrical plug and outlet receptacle. 3. Inherent network capability within every building that has power lines. A power line network would find applications in an office, a factory floor, and a home C123-C143. Uses include home bus control, security monitoring, and interactions with centralized and distributed data bases. In these applications, short - 3 -messages ensure adequate response time in heavy traffic. Long messages, resulting from batch processing or file transfer, could prevent other users from using the network and would therefore be restricted from network access until traffic has subsided. The physical layer of a power line network requires a power line modem. Modem design is difficult because of the inherent incompatabi1ity of the power lines to carry high frequencies, and the uncontrolled generation of impulse noise by phase-controlled electrical equipment. Furthermore, the transmitter power is regulated by appropriate governing bodies [15,16]. Limitations on transmitted power prevents a modem from achieving a high transmission rate by transmitting at high powers; because such transmissions could cause harmful interference to other power line communication systems. Several commercial power line modems are available for use over existing power lines. They are modems from BSR, NONWIRE, ExpertNets, and Consultant's Choice, all from U.S.A. In addition to these commercial products, several research papers reported similar developments. Ochsner [17] described a spread spectrum FSK system capable of 1200bps transmission rate. Van der Sracht et al. L181 reported the result of a 60bps ac zero crossing synchronized spread spectrum modem. Hirosaki et al. [193 described a 9600bps spread spectrum modem suitable for home use. Hariton et al. C203 reported a modem IC designed for transmitting and receiving signals over ac power lines using non-coherent FSK and on—off ASK modulation scheme. Not all power line modems can - 4 -operate satisfactorily across phases and under different power line environments. 1.2 Outline of Thesis In this thesis the suitability of Coherent Minimum Frequency Shift Keying for intrabuilding power line communications is investigated. An actual modem is implemented and tested, and CMFSK performance measurements for representative intrabuilding lines is determined. A spectrum spreading technique is developed to overcome narrow band impairments inherent with low data rates. The modem itself has several advantages over existing commercial products, including: 1. Variable data rates up to 19.2kbps. 2. Compact signal spectrum and fast sidelobe rolloff useful in lowering narrow band interference. 3. Transmission spectrum independent of data rate achieved by a databand spread spectrum technique. 4. Operation independent of ac mains voltage, eliminating the need to adjust to 50/60Hz, 2/3-phase power lines. In Chapter 2, several commercial modems are discussed to give an understanding of existing power line modem technology. In Chapter 3 and Chapter 4 respectively, power line transmission and power line noise characteristics are presented. In Chapter 5, an acceptable power line modem performance is presented, and the design philosophy for a CMFSK modem is outlined. - 5 -In Chapter 6, the performance of the CMFSK modem in AWGN is reported. The choice of operating frequency is explained and the compact spectrum of the CMFSK modulaion is-demonstrated. In Chapter 7, BER measurements at 1.2kbps, 4.8kbps, and 19.2kbps in industrial, commercial, and residential power line environments are given. Two major error mechanisms, namely impulse noise and momentary received Eb/No reduction, are reported. A majority of single-bit errors are caused by impulse noise, while multiple-bit errors are caused by either adjacent impulses or momentary received Eb/No reductions. In Chapter 8, the merits of using spread spectrum in a power line modem is discussed. The advantages include minimized co-site interference and overcoming of potential narrow band impairments. In addition, an interleaved PN code databand spread spectrum technique is reported, along with its ability to resist small error bursts. Results are given for 1.2kbps and 4.8kbps data rate at a code rate of 19.2kbps. A summary of the results together with recommendations for further work are found in Chapter 9. - 6 -2. DISCUSSION OF COMMERCIAL POWER LINE MODEMS Commercial power line modems are manufactured by BSR, NONWIRE, ExpertNet, and Consultant's Choice. The information is furnished by company literature or in reviews C213-C25D. A brief description on the modems and their communication protocols is also included. 2.1 BSR X-10. A Remote Control for Lights and Appliances This system consists of a central command unit (master) and a maximum of 255 light/appliance modules (slaves). A master can only transmit and a slave can only receive. Therefore, the master does not know whether a command sent to a particular slave has been executed or not. Transmissions and receptions are synchronized to the zero crossings of the power lines. This limits the BSR systems to split-single-phase supply. A single bit is sent out between two zero crossings, yielding a transmission rate of 120bps on a 60Hz power line. There is no forward error correction (FEC) and no automatic request for retransmission (ARQ) built into the modem. The modulation scheme is on-off ASK at a carrier of 120kHz. Average transmitter output varies from 800mV to 1.2V depending on the impedance of the power line to which it is coupled. There is no documented BER for this system but is believed to be small since received Eb/No should be very high given the level of output power and low data rate. This system is suitable for use in an apartment complex or a residential house. - 7 -The cost for the BSR X-10 system is about CAN$50 for the command console and CAN*25 for each remote module. 2.2 NON-WIRE Power Line Modem The NON-WIRE consists of a central control unit (master) and a maximum of 255 remote units (slaves). Master and slave can transmit to one another. However, a slave transmits only in response to queries from the master. An onboard microprocessor controlled by a firmware program performs error detection, and if possible, error correction on the received bit stream. The master will retransmit up to eight times if a slave does not acknowledge within a set time limit; this serves as a negative acknowledgement driven ARQ scheme. The transmission rate of the system is 1560bps. After the error check bits are removed, the throughput is reduced to 1.2kbps. The raw BER is lower than lxlO-=, and the system has at most one undetected error in two years of continuous operation. The modulation technique used is on-off ASK at a carrier frequency of 32kHz. The transmitted voltage imposed on the power line is 8v p-p. This system operates asynchronously with the power line ac voltage and could communicate across the different phases in either a 50/60Hz 2/3—phase power line network. This practice is not recommended and a separate 3—phase master is available to communicate across the different phases. Typical transmission range exceeds 2000 feet in a power line. The cost for either the master or slave alone is about CAN $1000. This includes the basic power line modem and an onboard microprocessor which implements the various host inter-face and network communication protocol functions. 2.3 ExpertNets Power Line Modem The ExpertNets system is similar to NONWIRE system at the protocol level. The modem uses a carrier frequency of 90kHz. The data rate is 700bps at an error rate less than i0~'T. The transmission range is typically 600 feet between units in standard 110v/220v house wirings. The output power of the modem is not known. Modem cost runs from US$100 to US$200 per unit. 2.4 Consultant's Choice Power Line Modem This modem boosts of 80,000bps transmission rate over ac power lines. The high rate is made possible by a modulation and detection technology used in low frequency and high noise environment. The modem uses an Intel 8039 microprocessor to perform error detection/correction and to implement the network protocol. A hybrid on-off ASK and FSK modulation scheme is used to combat common ac power line noises. Neither the BER nor the modem output voltage level is reported. 2.5 Commercial Power Line Modem Technology The above survey shows the dominant modulation technology is on-off ASK with carrier frequencies from 32kHz to 120kHz. In addition, the ac zero crossings are used in the BSR X-10 system for transmitter/receiver sychronization. - 9 -There are modulation schemes that are superior to ASK, i.e. •frequency shift keying (FSK) and phase shift keying (PSK) , in terms of lower achievable BER at the same received Eb/No. Both schemes, however, are more complicated to implement in hardware. The simplicity of on-off ASK makes large scale production feasible, which translates to low cost power line modems and low cost power line local area networks. The cost of a power line local area network will be one determinant of its success in power line data communication applications. Coaxial cable or fibre optics based communication systems will have a much higher transmission capacity (1Mbps to 100Mbps) and higher reliability. At present, the labour cost of installing a dedicated network whose coverage is as wide and as universal as the power lines is often very expensive. However, this does not apply to new constructions where an dedicated coaxial or fibre optics cable can be laid along side the power lines at a minimum cost. It is mandatory that the price of a power line network be kept low enough that it can be purchased at a small incremental cost to complement a faster dedicated data communication network. The low cost requirement makes on-off ASK an attractive modulation scheme in commercial power line modems since it is not hardware intensive and is essentially adjustment-free. This is because an ASK signal can be generated by a crystal controlled oscillator and can be detected using a high Q band pass filter. This is unlike other systems such as PSK or FSK which requires - 10 -more hardware. The selection of carrier frequency is determined by the noise and transmission characteristics of the power line. In general, modems having high transmission rates or which are expected to operate in residential houses and apartments use a carrier frequency from 90kHz to 120kHz. However, a lower carrier frequency has to be used whenever there is excessive attenuations at the high frequencies. A low carrier frequency places the signal spectrum in high power line noise environments, and the data rate is often reduced to ensure an acceptable received Eb/No. 3. POWER LINE TRANSMISSION CHARACTERISTICS Intrabuilding power line transmission characteristics have been reported by Chan C263 and Ochsner C17D. Their reports show that a typical power line acts like a low pass filter, with a cutoff frequency varying from 70kHz to 150kHz depending on the particular load profile. In addition, selective frequency fades are also reported. Their presence is a possible combination of special line loading and excitation of standing waves in the power lines. 3.1 Industrial Building Power Line Transmission  Characteristics A set of power line transmission measurements was performed in an industrial building (UBC McLeod building). A transmitter was fixed at remote phase B (see Section 7.1) and the receiver was alternately plugged into each of the three phases of a 3—phase bench power supply outlet bar in Room 442. The set of transmission curves are plotted in Fig. 3.1. These curves were obtained on a weekday between 8:30am and 4:30pm when the power line was heavily loaded. This set of remote transmission curves illustrates that: 1. Transmission loss from 20dB to 40dB is common with the receiver and transmitter on the same phase. 2. No discernable difference appears in transmission characteristics among the different phases in a remote transmission. 3. No sharp narrow band frequency dropouts >20dB are observed between 30kHz and 150kHz. - 12 -It is expected that (1) would prevail in a power line since it is not intended to carry high -frequency signals. Item <2) seems to contradict other -findings C41, in which transmission difference up to 16dB was reported. This is not really a discrepancy as another set of curves taken for close-in and across-phase transmissions will demonstrate. In this case, a transmitter was fixed in phase B of the bench supply mentioned above and the receiver was alternately plugged each of the three phases. The transmission curves obtained are shown in Fig. 3.1, which show the 16dB transmission difference and some selective frequency fades around 120kHz. This set of curves are designated as 'local' because the received SNR is higher than those in the 'remote' case. Notice that the 'local' transmission curves represent better power line reception conditions, because the received SNR is higher than those found in the 'remote' set of curves. The 'remote' transmission curves show that a typical power line bandwidth is quite narrow. In addition, it shows that the signal levels at remote phases are roughly equal. This could be a result of signal leak-throughs from the originating phase into adjacent phases via electrical equipment connected to more than one phase. It has been observed that an off—phase signal often became stronger than an in-phase signal, possibly because of some particular loading. In general, signal leak-throughs tend to equalize the signal levels in adjacent phases of a power line, thereby lowering the energy in the originating phase and raising that in adjacent phases. It is possible that these leak-throughs - 13 -or bypasses reduce the deepness o-f a fade, which explains (3). To observe the effects of power line loading on transmission characteristics, more transmission measurements were performed. Three more set of curves are obtained and plotted in Fig. 3.2 (another weekday between 8:30am and 4:30pm), Fig. 3.3 (after 6:00pm on a weekday), and Fig. 3.4 (weekend afternoon). The last two plots characterize the transmission conditions of a lightly loaded power line, and are expected to show less transmission loss. Comparisons between Fig. 3.1 & Fig. 3.2 (loaded) and Fig. 3.3 & Fig. 3.4 (lightly loaded) demonstrate that a lightly loaded power line suffers from less attenuation, especially in the high frequency range. In addition, a lightly loaded power line seems to support standing waves, possibly as a result of untapped energy being confined in the line. Photographs of received signal at various frequencies from 'local' and 'remote' transmissions are shown in Fig. 3.5. In these pictures, the effects of power line impedance modulation can be clearly seen from the periodic amplitude fades or dropouts in the received signals C17,203. 'Local' signal reception seems to be affected the most and exhibits severe amplitude fades at two frequencies, 53kHz and 106kHz, one being twice the frequency of the other one. These pictures are taken on a weekday between 8:30am and 4:30pm. The bandpass filter used in enhancing the 'remote' 100kHz and 120kHz signals in Fig. 3.5 has a center frequency of 140kHz with a 50kHz bandwidth. oJB *> I Jo»i **»t fcKflt liMt *>Mt 9*1% wMt IIpW, U*fo IhW* jytfjfe I5*** Fig. 3.1 'Local' and 'Remote' Transmission Curves on a Weekday Between 8:30am and 4:30pm in an Industrial Building -2o# '3* -401 r!-iii! ft it:' 'Si' 1 I: <telki Fig. 3.2 'Local' and 'Remote' Transmission Curves Weekday Between 8:30am and 4:30pm in an Industrial Building -i.it III Iii! ihSillif [liiJIi^iJ'ltfe ••* -hil iii|pj|jjl|||p yun pin-* 2>f-H 9'*** tot** I I'M* i»H» Fig. 3.3 'Local' and 'Remote' Transmission Curves on a Weekday After 6:00pm in an Industrial Building >4 I ' * ' 3""" 1 "M* ««JWt /J-»!«t Fig. 3.4 'Local' and 'Remote' Transmission Curves on a Weekend Day - 18 -'Remote' Upper traces 60Hz ac mains voltage as time reference Lower trace: Received signals with amplitude fading Fig. 3.5 Pictures of Received 'Local' and 'Remote' Signals at Various Frequencies: 40kHz 'Local' 'Remote' Fig. 3.5 Cont'd: 60kHz - 20 -- 21 -'Remote' Fig. 3.5 Cont'd: 100kHz - 22 -* Remote' Fig. 3.5 Cont'd: 120kHz - 23 -Fig. 3.5 Cont'd: BPF 100kHz (upper picture) and BPF 120kHz (lower picture) - 24 106kHz 53 kHz Fig. 3.5 Cont'd: Severe Amplitude Fade at 53kHz and 106kHz - 25 -3.2 Residential Building Power Line Transmission  Characteristics Sets o-f transmission curves are obtained for a residential house. They demonstrate that there is a wider bandwidth on a residential power line and also there is less attenuation. In addition, the effect of a signal bypass is reported. The two-story house is fed by a 110V/2-phase power circuit. Fig. 3.6 shows the in-phase and across-phase transmission curves. A wide power line bandwidth appears and there are no selective frequency fades. Fig 3.7 shows the effect of a resistive signal bypass on across-phase transmissions. The figure demonstrates that there is consistent improvement from 30kHz to 150kHz. The signal bypass is provided by a resistive path connecting across the two phases of the power line when a range element is turned on. The set of transmission curves is extended beyond 150kHz to 500kHz in Fig. 3.8. They show a 20dB frequency fade at 300kHz in the lBO43 phase. This phase exhibited another frequency fade at 600kHz, which was not shown in these curves. By providing a resistive signal bypass, 5dB to lOdB reductions in the depth of these fades were observed. I'M* ••Ml S*H\ MHX hw* *w* "0M*- U*M» ,J*»» Fig. 3.6 and ISO0 Phase Transmission Curves in a Residential House I M I loMt 4oKH* 5oW> cV>*H» MH 9oMl lt6M* ""^ l2okU* tof* Fig. 3.7 Transmission Difference between the 0° and 180° Phase With and Without a Resistive Signal Bypass OiB -loiB JoKW* to<Ht *bft)t (oMi 7»jt«, «<H*i 3aW»/«*/* '8»<fc JooWrf* Motfj JHW« Fig. 3.8 Extended 0° and 180° Phase Transmission Curves of the Residential House - 29 -4. POWER LINE NOISE 4.1 Power Line Noise Characteristics The noise -found on a power line is mainly periodic in nature and shows discrete spectral lines. This particular nature of power line noise has important implications for narrow band transmissions, because these noise spectral spikes can be avoided by placing the signal spectrum in between them, as is the case in several PLC communication systems. However, this approach is limited to low speed transmissions only. Measurements of power line noise spectral density are found in C9,10,27,28,293. A set of power line noise data was collected in the industrial building and residential house. Results of the calculated noise spectral density are compiled in Table 4.1 and Table 4.2. In addition, photographs of the power line noise in three frequency bands were taken in the industrial building and are shown in Fig. 4.1. These pictures indicate the existence of high amplitude decaying oscillatory outputs from the BPFs caused by impulsive noise spikes at the input. Furthermore, they also show that the background noise is also susceptible to periodic power line impedance modulation effects (see Section 3.1). One other observation is the different duration periods of these impulse effects in the three frequency bands, about lOOus in the low frequency band (20kHz-30kHz), 50us in the mid frequency band (50kHz-70kHz), and 25us to 30us in the high frequency band (120kHz-180kHz). The duration of an impulse at the input is around lOus to 15us. The duration of an impulse noise spike and - 30 -Table 4.1 Industrial Building Power Line Noise Spectral Oensi ty fa Phase A Phase B Phase C Remote 25kHz 85.0 43.4 43.4 91.2 60kHz 14. 1 22. 1 15.0 16.9 140kHz .373 .290 .381 .211 * Remote B is an outlet used for remote reception and transmission as defined in Section 3.1. note: 1. Unit in 10-AV=_mm/kHz. 2. Coupling terminating resistance, R»r», is 12 ohm. 3. Refer to Appendix A for noise density determination. Table 4.2 Residential Building Power Line Noise Spectral Density f0 Phase 0° Phase 180e* 25kHz 1.65 3.28 60kHz 1.22 .802 140kHz .0119 .0305 note: 1. Unit in 10-*V21_m./kHz. 2. Coupling terminating resistance, R»r», is 12 ohm. 3. Refer to Appendix A for noise density determination. HID BPF C 5b*Wz % 7o/cfe) Fig. 4.1 Power Line Noise in Three Frequency Bands - 32 -HIGH EPF C IZokHz -to fQotHz.) Fig. 4.1 Cont'd 4oKHz 14-6 kHz 1 Fig. 4.2 Power Line Noise Spectral Density - 33 -its induced e-f-fects on the BPFs may have implications in high speed transmissions. As the period of a bit decreases, the probability of its information being destroyed by an impulse increases. Table 4.1 and Table 4.2 show a downward trend in power line noise spectral density as frequency increases. This is in agreement with reported findings. The data also shows that an industrial power line is noisier than a residential one. A picture of the industrial power line noise spectral density is shown in Fig. 4.2, in which several intensified noise regions appear. In addition, there exists an approximate 20dB difference in spectral density between the extreme ends of the recorded spectrum. 4.2 Cross Sectional Power Line Noise Characteristics Power Line noise is influenced by particular line loading. Measurements of noise spectral density were performed in the UBC McLeod Building on four high power line activity and on four low power line activity days. A high activity day is any non-holiday weekdays between 8:30am and 4:30pm. A low activity day includes any weekend days. Measurement results are found in Table 4.3, which shows a trend of increased noise on low activity days and decreased noise on high activity days. This trend is attributed to reduced power line attenuations on low activity days. But it has been observed that power lines on a weekend period may have noise values similar to that of a high activity day, simply because the building's power lines are loaded. - 34 -Table 4.3 Industrial Building Power Line Noise Spectral Density on Four High Activity and Four Low Activity Days Phase A HIGH Activity Days Phase B Phase C Remote B* 60kHz .435 1.43 2.72 5. 17 60kHz 6.55 2.02 3. 11 3.31 60kHz 6.26 1.29 3.52 6.26 60kHz 1.57 .513 .870 .828 140kHz .314 .034 . 156 .113 140kHz . 143 .087 .119 . 143 140kHz . 131 .055 .097 .087 140kHz .072 .031 .023 .034 LOW Activity Days 60kHz 4.92 3.52 4.43 5.70 60kHz 11.64 5. 17 4.43 6.26 60 kHz 10. 14 3.52 3.96 3.52 60kHz 1.51 1. 19 1.56 1.29 140kHz .298 .026 .087 .934 140kHz .298 . 191 .271 1.38 140kHz .387 .206 .346 .855 140kHz .064 .037 .020 .034 • Remote B is an outlet used for remote reception and transmi ssion de-fined in Section , 3.1. note: 1. Unit in 10-* V2, /kHz. Coupling terminating resistance, R*n, is 3.3 ohm. 3. Refer to Appendix A for noise density measurement. - 35 -5. CMFSK POWER LINE MODEM DESIGN CONSIDERATIONS 5.1 Power Line Modem Design Criterion An acceptable modem design should meet the following requi rements: 1. Acceptable channel bit error rate performance at high transmission rate, i.e. 10-3 or better at 19.2kbps. 2. Minimum spectral spillover to reduce harmful interference to other power line communication systems and AM radios. 3. Proven technology giving high reliability, low implementation difficulty, and low cost. 4. Ability to operate in 50/60Hz 2/3-phase 110V/220V power line networks with minimum user adjustment. 5. Ability to operate continuously in a power outage. 6. Ability to co—exist with existing power line communication systems in the same power line network. 7. Meets or exceeds all applicable emission regulations. Some researchers C17D-C19D have applied spread spectrum methods to power line modems. Their intention is to overcome narrow band frequency fades and to achieve Code Division Multiple Access. In Chapter 8, the advantages and disadvantages of using spread spectrum signalling over power lines are discussed in more detail. Here a brief summary will suffice. The primary usefulness of spread spectrum in our application lies with its spectral density reduction capability. Coupled with the CMFSK modulation with its compact spectrum and fast sidelobe rolloff, spread spectrum is able to transform remanant sidelobe spillages into 'AWGN'—like noises. Therefore, spread - 36 -spectrum allows the use of enhanced transmitter power to overcome path attenuation without generating unacceptably high levels of interference. One should keep in mind that the ultimate goal is to concentrate as much energy as possible in the band of interest, and to transform unavoidable spectral spillages into white noise. This approach may seem redundant since the power line itself will eventually eliminate the high frequency components of a signal. Although this may be the case in remote transmissions, much high frequency interference can reach near—by or co-site users. This is similar to the near/far field problem discussed in Chapter 8. The use of ac zero crossings in synchronizing a transmitter/ receiver pair is found in BSR X-10 and in C43. The benefit is clear in low transmission rate system where small timing errors, in the form of in-phase zero crossing jitter [IB], can be tolerated before system performance is adversely affected. Measured results for off—phase zero crossing jitter were not included in C183. Our observations indicate that off-phase zero crossing jitter is much more serious than in-phase jitter. This is to be expected since in—phase zero crossings are derived from the same 60Hz signal but off-phase zero crossings are derived from three different 60Hz ac voltages. In high speed data transmissions, these zero crossings may not be accurate enough to provide adequate synchronization timing. However, they can still provide coarse timing reference. - 37 -Sixty—Hz synchronization methods -fail during a power outage. It is argued that there is no expected communications needs in such a situation, and therefore, outage problems can be ignored. However, this approach is rather restrictive on a power line local area network capability. Power supply networks in large buildings may be backed up by generators, digital systems may be temporarily backed up by batteries to perform pre-programmed shutdown operations, and power line security/monitor networks may be required to remain operational during power outages. Any modem design for use on power distribution networks must consider the problems mentioned in this and preceding sections. In the following sections, the design of a coherent minimum FSK modem is described along with its BER measurements at different bit rates and different power line environments. 5.2 Design Objectives of a CMFSK Power Line Modem From the above discussion, a compact spectrum modulation scheme with fast sidelobe rolloff is desirable in controlling interference when output power must be raised to compensate for transmission loss. Several compact spectrum modulation schemes exist and are discussed in 130,313. A suitable modulation scheme has a constant envelope and continuous phase transition output waveform. The spectrum for ASK, FSK, and PSK modulation schemes has a first null at 1/T, T being the bit duration. A more compact spectrum is realized with coherent minimum FSK (CMFSK). It has a - 38 -1st null at .75/T and a f-"* spectral rolloff rate, compared to f-2 in either PSK or ASK. Its drawback is an extra 3dB received energy is required to achieve the same BER performance of a perfectly sychronized PSK system operating in white noise. In practice, the 3dB energy advantage a PSK system enjoys over a coherent FSK system is sometimes obscured by imperfect carrier recovery C30,32,33,343 and by the presence of impulsive noise. The impulse noise effect on a PSK/DPSK and FSK system was reported in 126,35,363. Chan [263 presented the BER statistics for this author's CMFSK modem and for a perfectly synchronized PSK modem. He stated that the 3dB advantage of the PSK system did not seem to enhance its performance over the CMFSK modem. However, Chan reported more double and triple bit errors from the CMFSK modem and more single bit errors from the PSK modem at the same BER, which was attributed to the PSK inherent 3dB advantage. However, this finding is not supported by comparing the Eb/No values corresponding to the same BER from either modem. A possible interpretation of Chan's observation is that the CMFSK modem is more tolerant of low level impulse noise, but makes more errors when overcome by them. This possible impulse-tolerant ability of a FSK system may be because signal detection only involves the frequency information, while on the other hand, a PSK system relies on both the carrier and phase information of an incoming signal. This gives PSK the 3dB energy advantage in AWGN, but not necessarily so in impulse or non-Gaussian noise. The BER versus Eb/No curve of the PSK modem, without relying on perfect synchronization, should be established to investigate if - 39 -there is any significant energy advantage of PSK over FSK in impulse noise impaired signal detection. The coherent minimum FSK modulation scheme seems to be a good compromise between a compact spectrum, steep sidelobes, and efficient use of received signal energy. Its implementation is facilitated by proven modulation/demodulation techniques from existing integrated circuit chips. The goal of control led interference can be achieved by applying spread spectrum methods to the baseband signal without excessive spectral spillovers from the desired signal frequency band. Spread spectrum is thereby effective in reducing narrow band interference to other power line communications users and AM radios, while also allowing a higher transmitter output to compensate for path attenuation. 5.3 Design of a CMFSK Power Line Modem The heart of the modem lies with two EXAR ICs, XR-2206 and XR-2211. The XR-2206 is a function generator used in the transmitter section for generation of the two FSK frequencies. The XR-2211 is a FSK demodulator used in the receiver. The ICs are chosen for their frequency stability against power supply and temperature variations. The block diagrams of the transmitter and receiver are found in Fig. 5.1. More detailed schematics are found in Appendix B. - 40 -DATA xg-2206 a 5 6 7 8 HH C > 1 5^15^2 F5K Modulator 0') Tr*.v\smftier -0 C£ JJ7V04C -D> 0— 117WC cl <3 ZnieqirAire /Dump z XR-2211 13 /4 /2 /I HH ^2 FSK Demodulator RPrW DATA c3 VR1 XK-22/Z. C/ock Kecoi/ery (ii) Receiver Fig. 5.1 Block Diagram of CMFSK Transmitter/Receiver - 41 -5.4 CMFSK Modem Transmitter Description Referring to the transmitter block diagram, the two FSK frequencies are determined by CI, Rl, and R2. The FSK signal from the XR-2206 is fed to a linear power amplifier. The amplified signal is then coupled into the power line via pulse transformer Tl and ac blocking capacitor C2. The input to the power amplifier is controlled by VR1, and therefore, the power delivered to the power line can be varied. Measurements were performed to determine the output power at various transmitter output voltage levels (see Fig. 5.2). The results are contained in Table 5.1, which shows a maximum output of 0.8W. The power line impedance was also determined to be 10.5+.5 ohms at 120kHz, which agrees well with other reported values 137,383. 5.5 CMFSK Modem Receiver Description The receiver consists of an input power line coupling network, a bandpass filter, a FSK demodulator, a bit clock recovery, and a digital integrate/dump (see Fig. 5.3). The input power line coupling network is made up of an ac blocking capacitor CI, pulse transformer Tl, and resistor Rl. The frequency response of this coupling network is flat from 30kHz to beyond 200kHz (+0.0dB, -2.0dB). The bandpass filter (BPF) is made up of a high pass filter followed by a low pass filter. Its —3dB bandwidth is about 50kHz and is centered at 140kHz. The equivalent noise bandwidth is deduced to be 60kHz from the BER measurements in AWGN. - 42 -Spra^ue Poise Transferrer 11Z210Q where 2 x Vou* x V„ PF = Power Factor = magnitude I cos |^ 3 Pout = Power Delivered into the Power Line (Vou^ X X PF) - lout2 X Rmt = (Vout x Va x PF) - Vj^ x Rint R R2 Rmt = Internal Resistance o-f the Power Line Coupling = measured to be 3.0+.3 ohms. Fig. 5.2 CMFSK Modem Power Output Measurement Method - 43 -Table 5.1 Coupled Power into the Power Line @ 120kHz v4„ vout v„ Pc3«Jlt 1.00+.01V 0.93+.01V 70+2mV 50mW 2.00+.02V 1.86+.02V 138+2mV 200mW 3.00+.02V 2.80+.02V 202+2mV 440mW 4.00+.05V 3.72+.05V 290+2mV 790mW note: 1. P0«t = Power delivered into the power line. 2. Power line impedance is deduced to be 10.5+.5 ohms @ 120kHz using this table and Fig. 5.2. 3. Refer to Fig. 5.2 for power measurement technique. GX DATA CLK A A A A A A /|\ 54rnplithy CLfc -£> J>AT/\ Fig. 5.3 Block Diagram of Digital Integrate/Dump - 45 -The XR-2211 FSK demodulator incorporates a phase locked loop (PLL) to compare the incoming -frequency with a reference frequency. The FSK demodulation process generates an analog voltage at pin 11, which corresponds to either a sent '1* or '0' plus received noise. The reference frequency is controlled by C2 and VR1 and the bandwidth of PLL response is set by R2 and C3 (see [393 for more information on FSK modulation/demodulation). The recovered analog data pulses are then low pass filtered to further attenuate out-of-band noise. At this point, the data pulses could instead be analog signals, thereby providing voice communications capability via the power lines. This method is used in most low cost residential power line FM intercoms. The analog data pulses, after hardlimiting, drive a phase locked loop (XR-2212) controlled bit clock recovery circuit, which supplies timing signals to the modem's digital circuitry. The principle of bit clock recovery is discussed in [40,413. In essence, a differentiated bit stream has a discrete spectral line at the bit rate frequency. This line is then locked onto by a PLL, which provides the appropriate bit timing signals. A more detailed description on the operation of a clock recovery circuit is found in [423. The digital integrate/dump uses a 1-bit A/D conversion to extract 16 samples of either '1' or '0' from a raw data pulse. At the end of the sampling interval, the 16 samples are summed to yield a threshold decision on the data pulse. Its advantage is the elimination of an 'integrate' capacitor which is prone to - 46 -a large voltage step change caused by an unsuppressed impulse (see Fig 5.7 in C4D). If a digital integrate/dump had been used in that instance, the impulse would likely have modified one or two samples out of sixteen, resulting in decreased 'integrator' sensitivity to impulse effects. Another benefit is the easy adaptation to other bit rates by a simple change in the sampling clock rate. Finally, this digital integrate/dump can be easily incorporated into an IC chip. Degradation due to the discrete nature of this circuit does occur because of hard decisions on each sample; however the degradation in terms of higher BER was not apparent when this digital integrate/dump was compared to an analog integrate/dump. This is attributed to the many samples used in forming the threshold decision. - 47 -6. PRELIMINARY CMFSK TESTS TO SELECT SIGNALLING PARAMETERS 6.1 CMFSK Bit Error Rates in AWGN A white noise test was performed to verify modem operation at 19.2kbps transmission rate and offset frequency of +4.8kHz. The carrier frequency was chosen to be 120kHz. The test setup is as shown in Fig. 6.1 and the white noise BER versus Eb/No results are plotted in Fig. 6.2. Additional BER test results at 1.2kbps and 4.8kbps data rate were obtained and recorded in Table 6.1. Using these results, the equivalent noise bandwidth was estimated to be 60kHz, which is larger than the measured —3dB bandwidth of 52kHz. In white noise the BER of the CMFSK modem is given by: BER = Q C (Eb/No)1"3 3 CO where QZxl = ( 27T)-1'2 exp(-yse/2) dy Eb = average bit energy = Pb / bit rate = V2.i0 /bit rate No = noise density = V2no,„ / Eqv BW Pb = received signal power Eqv BW = Equivalent noise bandwidth The solid curve shown in Fig. 6.2 represents the ideal BER performance using 60kHz as the equivalent noise bandwidth. 6.2 Spectral Comparison of a PN Sequence and its CMFSK Output , The spectrum of a 19.2kbps, length 2047 PN code is shown in Fig. 6.3. The main lobe is centered at d.c. and is 40kHz wide. If the spectrum is centered at 120kHz, it would represent the - 48 -Table 6.1 1.2kbps, 4.8kbps, and 19.2kbps BER Results in Additive White Gaussian Noise Bit Rate Eb/No BER Predicted BER 1.2Kbps 4.95dB 7.2x10"= 3.8xl0~= 1.2kbps 8.47dB 5.7x10-=* 4.0x10-=* 1.2kbps 10.97dB 3.2x10"* 2.1x10-* 1.2kbps 12.55dB 2. 9x10-° 1.Ixl0-B 1.2kbps 13.58dB 1.4x10-= 9.0x10-^ 4.8kbps 4.95dB 7.5x10-= 3.8x10-= 4.8kbps 8.47dB 6.8x10-=* 4.0xl0-3 4.8kbps 10.97dB 2.2x10-* 2.1x10-* 4.8kbps 12.18dB 1.5xl0-«» 2.4x10-" 4.8kbps 12.91dB 3.0x10-* 4.9x10~* 4.8kbps 16.99dB < 10-° 7.6xl0-13 19.2kbps 4.95dB 6.7x10-* 3.8x10-= 19.2kbps 8.47dB 5.0x10-=* 4.Ox10~3 19.2kbps 10.97dB 1.9x10-* 2.1x10-* 19.2kbps 11.99dB Z.4xl0-~ 3.4x10-° 19.2kbps 12.91dB 1.2x10-" 4.9x10-" 19.2kbps 13.74dB 2.6x10— 5.8x10-^ 19.2kbps 14.49dB 2x10-'' 5.7x10-" flojse Generator BER <j-F5K henno^nla^ot—I HP /645/A DATA ERROR AM/\LY2ER DATA Fig. 6.1 CMFSK Modem White Noise BER Measurement Method Fig. 6.2 CMFSK Modem White Noise 19.2kbps BER Performance - 50 -Fig. G.4 Spectrum of a 19.2kbps PN code CMFSK Output output spectrum of a PSK system with a 120kHz carrier. The spectrum of the CMFSK output is shown in Fig. 6.4. The main lobe is centered at 120kHz and is 30kHz wide. A comparison of the two spectra confirms the narrower spectrum and faster sidelobe roll off characteristics of CMFSK over an equivalent PSK system. 6.3 CMFSK Carrier Frequency Selection The carrier frequency dictates where the signal spectrum will lie. Its selection is a compromise between low noise and low attenuation loss. Two frequencies are chosen, 60kHz (low attenuation) and 120kHz (low noise), to determine which carrier frequency is better. The BER was measured at the two different frequencies for both a 'local* and a 'remote' reception. The 'local' reception results at 60kHz are plotted in Fig. 6.5. The 'local' reception results at 120kHz are plotted in Fig. 6.6. The 'remote' reception results are tabulated in Table 6.2. The tests were performed on weekdays between 8:30am and 4:30pm. The BER results were recorded after about 100 bit errors had occurred. The exceptions were when the BER measurements surpassed 10-* BER range and were stopped irregardless of the number of errors. It is observed that the 60kHz carrier frequency suffers from more power line impairments, i.e. deeper amplitude fades. Therefore, the 120kHz carrier frequency is chosen for its overall better performance. A spectrum analyzer was used to record the background noise and to superimpose it on top of the spectrum of received signal plus noise. The result, shown in Fig. 6.7, shows the spectrum of Fig. 6.5 Interphase 19.2kbps BER versus Eb/No curve at a Carrier Frequency of 60kHz Fig. 6.6 Interphase 19.2kbps BER versus Eb/No curve at a Carrier Frequency of 120kHz - 54 -^'5ial + A/oise Fig. 6.7 Spectrum of CMFSK output Superimposed on Power Line Noise - 55 -Table 6.2 'Remote' Interphase 19.2kbps BER versus Eb/No at Transmitter Output o-f IV, 2V, 3V, and 4V at Two Different Carrier Frequencies of 60kHz and 120kHz Trx Output Trx Freq Rcr Locn Rcr Eb/No BER (Rm 442) l.OV 60kHz Bench 'A' 12.2dB 2.0x10"= l.OV 60kHz Bench B' 19.7dB 5.2xl0-» l.OV 60kHz Bench 'C* 14.3dB 7.0x10-= l.OV 120kHz Bench A' ll.OdB 1.0x10-= l.OV 120kHz Bench *B' 17.5dB 3.5x10-* l.OV 120kHz Bench 'C 13.2dB 3.3x10-= 2.0V 60kHz Bench 'A' 18.3dB 4.0xl0~:s 2.0V 60kHz Bench 'B' 25.5dB 8.0x10-* 2.0V 60kHz Bench 'C 20.0dB 3.2x10-= 2.0V 120kHz Bench *A' 15.6dB 7.5x10-= 2.0V 120kHz Bench B' 24.0dB 4.0x10-° 2.0V 120kHz Bench C* 18.4dB 2.3x10-' 3.0V 60kHz Bench 'A* 22.2dB l.OxlO-3 3.0V 60kHz Bench *B' 29.8dB 9x10-* 3.0V 60 kHz Bench 'C* 24.3dB 4.0x10-= 3.0V 120kHz Bench 'A' 19.6dB 4.2x10-' 3.0V 120kHz Bench *B* 26.3dB 1.0x10-* 3.0V 120kHz Bench *C* 20.5dB 7.5x10-* 4.0V 60kHz Bench 'A' 24.3dB 2.5x10-* 4.0V 60kHz Bench 'B' 30.7dB 3x10-* 4.0V 60kHz Bench *C 25.3dB 3.7x10-= 4.0V 120kHz Bench 'A' 26.0dB 6.5x10-* 4.0V 120kHz Bench 'B' 32.0dB l.OxlO-* 4.0V 120kHz Bench *C 25.8dB 1.6x10-* note: 1. Transmitter on remote phase B. 2. Bench *C signal receptions suffer from severe periodic amplitude dropouts at 60kHz. - 56 -the 120kHz signal is concentrated in a relatively quiet frequency band. This indicates that given the same transmitter power, a high carrier frequency will be received at higher SNR ignoring the effects of path attenuation. However, bit error performance should not be affected by the choice of carrier frequency as long as received Eb/No are the same. The effect of carrier frequency on BER can be seen in the photographs of power line noise in three frequency bands shown in Fig. 4.1, in which they show an abundance of rich impulses in lower frequencies. Observations of these impulses on a scope show approximately the same number of impulses in each frequency band, but the amplitude and duration of the impulses are higher and longer in the lower frequencies. Therefore, a data signal on a low frequency carrier may suffer excessive BER deterioration from a noise impulse. 6.4 BER Measurement for CMFSK in a 'Remote' Reception A set of 'remote' reception BER data was collected to check for any significant difference between a 'local' and a 'remote' reception. The results, plotted in Fig. 6.8, show no appreciable difference in BER performance as long as the same Eb/No is received. 'Local* and 'remote' conditions are defined in Section 3.1. This result is useful because it was obtained from actual power line transmissions over an unknown and hostile signal path. In addition, the viability of the CMFSK modem in power line data communications is clearly demonstrated. The BER versus Eb/No curve can be broken down into two regions: Fig. 6.8 'Remote' Interphase 19.2kbps BER versus Eb/No at a Carrier Frequency of 120kHz - 58 -1. An exponential decay in BER from 10-* to 10-3, identified as the white noise region in the BER versus Eb/No curve. 2. A change from the exponential to a gradual linear dropoff when BER drops below lO-2, identified as the impulse noise region in the BER versus Eb/No curve. A power line modem suffers primarily from white noise effects at low Eb/No values. As the received Eb/No increases, impulse noise becomes the dominant cause of bit errors, leading to slow BER dropoff in high Eb/No values. This trend is found in telephone channels and other practical data links, where the predominant error mechanism is impulse noise. When there is very strong signals, only a high amplitude impulse can occasionally overwhelm the receiver and cause a bit error. Since power line impairments are periodic, errors are also periodic (see Section 7.5) . - 59 -7. BER MEASUREMENT RESULTS IN VARIOUS ENVIRONMENTS AND AT VARIOUS DATA RATES Our prototype CMFSK modem was operated under three different power line environments to obtain BER measurements as well as bit error statistics reported elsewhere C263. The three environments include an industrial building, an apartment complex, and a residential house. A majority of the BER measurements are from the 3-phase industrial building at 1.2kbps, 4.8kbps, and 19.2kbps data rate. The other buildings were used to collect BER measurements at 19.2kbps data rate only. 7.1 Interfloor/Interphase Industrial BER Measurements at  1.2kbps. 4.8kbps. and 19.2kbps The industrial building used was the McLeod Electrical Eng. building at UBC. It is four stories high with a main office, a power generation laboratory, a high voltage studies laboratory, and many other laboratory facilities. The power distribution circuit is supplied by a 110V/3-phase power network. Some heavy machinery such as a building ventilation fan runs continuously. The BER tests were taken across al1 the three phases of the power line network and across all the four floors of the building. The tests were run mostly between 8:30 am to 4:30 pm, Monday through Friday, when there was the most variety and the heaviest loading of the building's power lines. In the tests, the transmitter was placed on different floors and the receiver was confined to Room 442 and extracted the signal from four different outlets. Three of these outlets - 60 -belong to a 3—phase power supply outlet bar on one of the benches and the fourth outlet is underneath a sink in the same room. These outlets are identified as bench supply phase A, bench , supply phase B, bench supply phase C, and remote phase B, respectively. In Section 3.1, the 'remote' set of transmission curves was obtained by injecting a signal into remote phase B and recovering it in the three bench phases. The 'local' set of transmission curves was obtained by injecting a signal into phase B and recovering it in the three phases of the bench supply rail. Therefore, the 'local' phase B provides a reference against which other transmission curves can be compared with. It should be noted that the transmission paths between the "under—the—sink-outlet" and bench outlets are actually worse than those from the basement outlets to the 4th floor bench outlets. The BER measurements are compiled in Table 7.1 and plotted in Fig. 7.1. The same set of tests was repeated for data rates at 1.2kbps and 4.8kbps. The BER vs. Eb/No results are compiled in Table 7.2 and Table 7.3 and plotted in Fig. 7.2 and Fig. 7.3. These graphs reveal that the white noise region extends to lower BERs at lower data rates. This indicates the receiver has increased resistance to impulse noise at low data rates. 7.2 Interfloor/Interphase Apartment BER Measurements at  19.2kbps The three-story apartment has about 50 individual dwelling units. It provides accomodation to UBC students. The building is supplied by llOV/split-single-phase power. BER measurements - 61 -were taken for across-phase and across-floor conditions at 19.2kbps. Over the duration of the measurement period, it was observed that average background power line noise is lowest from mid morning to noon and highest from late afternoon to midnight. The BER measurements are compiled in Table 7.4 and plotted in Fig. 7.4. The white noise and impulse noise regions seem to be similar to those obtained in the industrial building. The main source of errors comes from the on/off transients of nearby appliances. A transient may cause either no errors or upwards of ten or more errors depending on its strength and duration. 7.3 Interfloor/Interphase Residential BER Measurements at  19.2kbps The two-story house is supplied with 1lOV/split-single-phase power. A set of remote across—phase 19.2kbps BER measurements were taken across two floors. Two sets of BER measurements were taken, one with a hair dryer turned on to its maximum of 1200W and the other with the hair dryer off. When the hair dryer was switched on, the average noise level increased from llmV to over 30mV while the noise peaks jumped from 130mV to over 850mV. The results are plotted in Fig. 7.5. The transmitted output power level is higher in the hair dryer "ON" case (see Table 7.5). 7.4 Cross Sectional BER Performance at 19.2kbps in an  Industrial Building In a practical application, a power line modem operates with a preset output voltage level. Therefore, it is important to know how the BER performance varies during continuous operation - 62 -Table 7.1 Interfloor/Interphase 19.2kbps BER versus Eb/No at Transmitter Output of IV, 2V, 3V, and 4V in an Industrial Building Trx Output Trx Locn Rcr Locn (Rm 442) Rcr Eb/No BER l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV l.OV 2.0V 2.0V 2.0V 2.0V 2.0V 2.0V 2.0V 2.0V 2.0V 2.0V 2-OV 2.0V 2.0V 2.0V 2.0V 2.0V Bsmt Bsmt Bsmt Bsmt 2nd floor 2nd floor 2nd f1oor 2nd floor 3rd floor 3rd floor 3rd f1oor 3rd floor 4th floor 4th floor 4th floor 4th floor Bsmt Bsmt Bsmt Bsmt 2nd floor 2nd floor 2nd floor 2nd floor 3rd floor 3rd floor 3rd floor 3rd floor 4th floor 4th floor 4th floor 4th floor Bench Bench Bench Remote Bench Bench Bench Remote Bench Bench Bench Remote Bench Bench Bench Remote Bench Bench Bench Remote Bench Bench Bench Remote Bench Bench Bench Remote Bench Bench Bench Remote A' B' C* B* A' B' C* B' A* B' C B' A' B' C* B' A' B' C* B' A' B' C B' A' B' C* B* A' B' C B' 16.7dB 12.8dB 15.0dB 15.9dB 16.3dB 12.2dB 16.7dB 15.5dB 15.5dB 9.7dB 15.0dB 12.4dB 15.0dB 8.7dB 12.4dB 14.5dB 22.4dB 17.ldB 20.ldB 18.ldB 23.3dB 16.3dB 20.4dB 18.4dB 22.8dB 16.3dB 19.3dB 16.7dB 21.3dB 15.9dB 18.8dB 14.2dB 1.3x10"' 6.0x10"' 2.8x10"' 3.4x10"' 2.2x10"' 8.0x10-=* 4.Ox 10"' 3.5x10"' 7.5x10-' 1.6xlO"= 8.5x10-' 8.5x10-' 2.9x10-' 1.2x10-= 8.0x10-' 7.0x10-' 1.0x10-° 2.3x10-' 1.1x10-* 2.6x10-' 7.0x10-* 1.7x10-' 1.0x10-* 2. 1x10-' 5.0x10-* 2. 1x10-' 2.0x10-* 2.0x10-' 4.5x10-° 3.4x10-' 4.2x10-* 2.5x10-' - 63 -Table 7.1 Cont'd Trx Output Trx Locn Rcr Locn Rcr Eb/No BER (Rm 442) 3.0V Bsmt 3.0V Bsmt 3.0V Bsmt 3.0V Bsmt 3.0V 2nd floor 3.0V 2nd floor 3.0V 2nd floor 3.0V 2nd floor 3.0V 3rd floor 3.0V 3rd floor 3.0V 3rd floor 3.0V 3rd floor 3.0V 4th floor 3.0V 4th floor 3.0V 4th floor 3.0V 4th floor 4.0V Bsmt 4.0V Bsmt 4.0V Bsmt 4.0V Bsmt 4.0V 2nd floor 4.0V 2nd floor 4.0V 2nd floor 4.0V 2nd floor 4.0V 3rd floor 4.0V 3rd floor 4.0V 3rd floor 4.0V 3rd floor 4.0V 4th floor 4.0V 4th floor 4.0V 4th floor 4.0V 4th floor Bench Supply 'A' Bench Supply 'B' Bench Supply 'C* Remote 'B' Bench Supply 'A' Bench Supply 'B' Bench Supply *C Remote *B' Bench Supply 'A* Bench Supply 'B * Bench Supply 'C* Remote 'B' Bench Supply 'A' Bench Supply 'B' Bench Supply *C Remote 'B' Bench Supply 'A* Bench Supply *B' Bench Supply 'C Remote 'B' Bench Supply 'A' Bench Supply 'B' Bench Supply *C* Remote 'B' Bench Supply 'A* Bench Supply 'B* Bench Supply 'C Remote B* Bench Supply 'A' Bench Supply *B' Bench Supply *c Remote 'B' 25.2dB 1.0x10-° 23.7dB 2.0x10-° 24.6dB 3.0x10-° 22.2dB 6.0x10-* 24.8dB 2.0x10-* 22.6dB 3.0x10-* 22.9dB 8.0x10-° 20.4dB 1.2x10"' 23.3dB 5.0x10-° 20.9dB 5.0x10-* 22.4dB 8.0x10-" 19.9dB 1.0x10"' 23.3dB 3.0x10-* 21.1dB 9.0x10-* 22.2dB 4.0x10-° 17.5dB 1.0x10-' 29.5dB < 10-* 22.2dB 3.0x10-* 27.2dB 1.0x10-° 28.5dB 1.0x10-* 26.3dB < 10-* 20.5dB 1.5x10-' 27.6dB 1.5x10-° 33.2dB 1.5x10-° 25.7dB 1.5x10-° 21.8dB 2.2x10-' 24.6dB 1.7x10-* 30.4dB 5.0x10-° 24.3dB 1.0x10-* 20.ldB 3.5x10-' 23.7dB 3.0x10-* 33.5dB < 10-* - 64 -Table 7.2 Interfloor/Interphase 4.8kbps BER versus Eb/No at Transmitter Output of .IV, .2, .4V, and 2.5V in an Industrical Building Trx Output Trx Locn Rcr Locn Rcr Eb/No BER (Rm 442) lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV lOOmV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV 200mV Bsmt Bsmt Bsmt Bsmt 2nd floor 2nd floor 2nd floor 2nd floor 3rd floor 3rd floor 3rd floor 3rd f1oor 4th floor 4th floor 4th floor 4th floor Bsmt Bsmt Bsmt Bsmt 2nd floor 2nd floor 2nd floor 2nd f1oor 3rd floor 3rd floor 3rd floor 3rd floor 4th floor 4th floor 4th floor 4th floor Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote A* B' C B' A' B' C* B * A' B* C B' A' B' C* B' A' B* C* B' A* B' C B* A' B' C* B* A' B' C* B' 5.4dB OdB 4.0dB 15.3dB 2.7dB OdB 2.5dB 1.7dB 2.7dB OdB 2.5dB 3.2dB 2.7dB OdB OdB 4.3dB 10.9dB 1.3dB 10.8dB 15.7dB 10.5dB 5.4dB 7.3dB 17.7dB 9. OdB 7. OdB 7.3dB 17.2dB 8.5dB 4.4dB 5.4dB 16.6dB 8.0x10-= 2.5xl0~* 5.0x10-= 1.3x10-' 1.5x10-= 2.5x10-* 7.0x10-= 1.0x10-= 2.0x10-= 2.5x10-* 8.5x10-= 2.5x10-= 3.0x10-= 2.5x10-* 8.5x10-= 1.3x10-= 4.0x10—» 1.3x10-* 1.0x10-= 1.0x10-^ 2.0x10-= 1.1x10-* 4.5x10-= A.OxlO-*5 2.2x10-= 1.4x10-* 4.5x10-= 1.0xl0~» 3.0x10-= 1.5x10-* 5.5x10-= 2.5xlO-|B - 65 Table 7.2 Cont'd Trx Output Trx Locn Rcr Locn (Rm 442) Rcr Eb/No BER 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 400mV 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V 2.5V Bsmt Bsmt Bsmt Bsmt 2nd floor 2nd floor 2nd floor 2nd floor 3rd floor 3rd floor 3rd floor 3rd floor 4th floor 4th floor 4th floor 4th floor Bsmt Bsmt Bsmt Bsmt 2nd f1oor 2nd floor 2nd floor 2nd floor 3rd floor 3rd floor 3rd f1oor 3rd f1oor 4th floor 4th floor 4th floor 4th floor Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote Bench Supply Bench Supply Bench Supply Remote A* B' C* B* A' B' C B' A* B' C* B' A' B' C* B* A' B' C B' A' B' C B* A' B' C* B' A' B" C* B' 16.3dB 7.0dB 15.0dB 23.8dB 14.7dB 5.4dB 11.7dB 23.3dB 13.5dB 5.4dB 11-OdB 22.7dB 25.4dB 5.4dB ll.OdB 22.4dB 29.4dB 25.2dB 27.9dB 28.5dB 26.3dB 19.2dB 27.ldB 32.5dB 28.7dB 21.6dB 26.3dB 35.6dB 29.4dB 20.ldB 25.9dB 35.0dB 2x10-* 4.5x10-= 2.5x10-° < io-* 5x10-* 4.0x10-= 5.5x10-* 2x10-* 4x10-* 3.7x10-= 1.3x10-=* 8xl0~* 6x10-* 4.0x10-= 1.7x10-=* 7x10-* 3x10-* 1x10-* < 10-* < io-* < 10-* 1x10-* < 10-* < 10-* < 10-* 1x10-* < io-* < io-* 1x10-* 2x10-* < io-* < io-* - 66 -Table 7.3 Interfloor/Interphase 1.2kbps BER versus Eb/No at Transmitter Output of 150mV and 200mV in an Industrial Building Trx Output Trx Locn Rcr Locn Rcr Eb/No BER (Rm 442) 150mV 150mV 150mV 150mV Bsmt Bench Supply A' 16.8dB 5x10-° Bsmt Bench Supply 'B' 9.8dB 3.5xl0-' Bsmt Bench Supply 'C 14.2dB 2.0xl0~' Bsmt Remote 'B' 20.6dB 1x10-° 150mV 4th floor Bench Supply 'A' 12.6dB 1x10-° 150mV 4th floor Bench Supply 'B* 10.2dB 1.0x10-' 150mV 4th floor Bench Supply *C* 12.6dB 3.0x10-' 150mV 4th floor Remote 'B' 18.0dB 2xl0~" 200mV 200mV 200mV 200mV Bsmt Bench Supply 'A* 15.3dB lxl0~" Bsmt Bench Supply *B' 6.4dB 6.1x10-= Bsmt Bench Supply 'C 16.3dB 4x10-° Bsmt Remote 'B' 19.3dB < 10~" 200mV 4th floor Bench Supply 'A' 12.4dB 1x10-" 200mV 4th floor Bench Supply *B* 5dB 8.0x10-= 200mV 4th floor Bench Supply *C 8dB 1.0xl0~= 200mV 4th floor Remote 'B' 14.3dB 1x10-" - 67 -Table 7.4 Interfloor/Interphase 19.2kbps BER versus Eb/No at Various Transmitter Output in an Apartment Complex Trx Output Trx Locn Rcr Locn Rcr Eb/No BER lOOmV Rm317 Hallway Rm306 16.9dB 1.3xl0~3 150mV Rm317 Hallway Rm306 18.3dB 3.5xl0~* 300mV Rm317 Hallway Rm306 24.4dB 3.5x10-° 500mV Rm317 Hallway Rm306 29.ldB 1.0xl0~° lOOmV Rm321 Stairs 150mV Rm321 Stairs 300mV Rm321 Stairs 500mV Rm32i Stairs Rm306 Rm306 Rm306 Rm306 11.7dB 16.4dB 21.6dB 29.2dB 7.0x10-= 5.0x10— 3. 1x10-* 1.0xl0~e 50mV Rm217 Hallway Rm306 8.8dB 7.5x10-=* 64mV Rm217 Hallway Rm306 14.2dB 1.4X10-3* lOOmV Rm217 Hallway Rm306 15.6dB 7.0x10-° 200mV Rm217 Hallway Rm306 24.9dB 3.0x10-° 70mV Rm221 Stairs lOOmV Rm221 Stairs 200mV Rm221 Stairs 500mV Rm221 Stairs Rm306 Rm306 Rm306 Rm306 10.ldB 14.OdB 19.4dB 28.3dB 7.5xl0-3 1.8x10-=* 1.0x10—• 2.2x10-° lOOmV Rmll7 Hallway Rm306 17.4dB 1.0x10" 150mV Rmll7 Hallway Rm306 21.OdB 3.0x10" 300mV Rmll7 Hallway Rm306 24.2dB 6.5x10" 1.0V Rml17 Hallway Rm306 35.ldB 1.5x10" lOOmV Rml21 Stairs 150mV Rml21 Stairs 300mV Rml21 Stairs 500mV Rml21 Stairs Rm306 Rm306 Rm306 Rm306 13.ldB 16.7dB 24.ldB 29.7dB 2.0x10" 1.5x10" 2.0x10" 2.5x10" - 68 -Table 7.5 Interfloor/Interphase 19.2kbps BER versus Eb/No at Various Transmitter Output in a Residential House with a Hair Dryer 'ON' and 'OFF' HAIR DRYER Trx output Rcr Eb/No BER OFF* 35mV lOdB 7.5X10"3 'OFF' 60mV 16dB 1.2x10-=" OFF* HOmV 20dB 2.0x10-° OFF* 200mV 29dB 3.0x10"-* 'OFF' 500mV 35dB 3x10-" 'ON' 35mV 5dB 7.0x10-= 'ON' HOmV lOdB 3.0x10-'ON' 200mV 15dB 2.5x10-=* *0N' 300mV lBdB 5.Ox 10-* 'ON' 500mV 21dB 3.0x10-° Fig. 7.2 4.8kbps BER versus Eb/No Measurement in an Industrial Building Fig. 7.3 1.2kbps BER versus Eb/No Measurement in an Industrial Building Fig. 7.4 19.2kbps BER versus Eb/No Measurement in an Apartment Complex Fig. 7.5 19.2kbps BER versus Eb/No Measurement in an Residential House - 74 -over several days. BER measurements were carried out over a 5-day period, starting on a Thursday and ending on a Monday. The transmitter was -fixed at 4V output with the remote-receiver on the same phase as the transmitter (bench supply 'B' to remote 'B'). The results are tabulated in Table 7.6. The trend in cross sectional BER measurement indicates that power line loading has an adverse effect on power line communications, as indicated by the worst case BERs occurring in the afternoons of two weekdays. During the weekend period, the BER drops by a large amount of three order in magnitude or more, i.e. 10—3 to 10_*. This dynamic fluctuation in BER reflects the adverse effect of power line loading. 7.5 CIassfication of Power Line Induced Errors Power line induced errors can be catergorized as follows: 1. Impulse noise induced errors, either periodic or aperiodic. 2. Momentary Eb/No reduction induced errors, either periodic or aperiodic. By far the most dominant source of errors are caused by periodic impulse noise. Fig. 7.6 records the periodic bit error occurance caused by periodic impulses in the received signal. The generation of periodic impulses is linked to operations of phase-controlled electrical equipement C28,29D. Aperiodic impulse noise is likely the result of high amplitude turn-on and turn—off transients (see Fig 2.8 in C4D). In addition, it was observed that power line impulses could appear 200us to 300us apart, leading to the impression of small error bursts 1263. - 75 -Table 7.6 Cross i Sectional 19.2kbps BER Measurement Time Thursday Fri day Saturday Sunday Mond. lam n/a 23 0 0 617 2am n/a 7 4 6 315 3am n/a 13 0 0 396 4am n/a 15 0 0 364 5am n/a 7 0 0 410 6am n/a 20 0 0 573 7am n/a 25 0 0 674 8am n/a 23 0 0 622 9am n/a 0 1 0 682 10am n/a 15 5 1 346 1 lam n/a 34 0 1 294 12am n/a 34 0 0 1999 1pm n/a 399 3 2 289 2pm n/a 624 0 0 599 3pm 109 657 O 0 329 4pm 219 1999 0 o n/a 5pm 28 865 0 2 n/a 6pm 188 1286 1 0 n/a 7pm 37 806 3 3 n/a 8pm 27 1426 0 3 n/a 9pm 53 1112 1 0 n/a lOpm 30 588 5 0 n/a 11pm 606 1 0 1005 n/a 12pm 32 0 4 570 n/a note: 1. All BERs in unit of 10~*. 2. 1999 indicates error counter overflow. 3. Continuous BER measurement achieved by leaving the HP1645A data error analyzer on automatic PRINTER mode at EXPONENT of '6'. - 76 -| Impulse. Fig. 7.6 Impulse Noise Induced Bit Errors - 77 -A momentary Eb/No reduction occurs when the signal undergoes a temporary amplitude dropout or when the noise undergoes a temporary increase. It is also possible for both the signal and noise to experience the same amplitude variations to preserve the same Eb/No both before and after a momentary amplitude fade. Origin of signal amplitude variations is linked to power line impedance modulation effects [17,203. Amplitude fading affects all frequencies but its depth is frequency dependent (see Fig. 3.5). Errors resulting from Eb/No reductions usually occur in small bursts. Errors caused by either periodic or aperiodic momentary Eb/No reductions are rare. This is because the CMFSK modem is usually operating in high Eb/No environments, and an error occurs mainly as a result of a noise impulse. Discussion on periodic and aperiodic signal amplitude fades also appears in Chan's thesis [263. 7.6 Discussion on Power Line Impulse Noise Suppression  Techni ques Most bit errors are a result of impulse noise. An impulse, upon entering a detection system, will sharply reduce Eb/No and cause a possible error. Some papers have been presented on optimum or near-optimum weak signal detections in non-Gaussian noise channels [43,443. Impulse noise suppression is essentially a non-linear process. One simple approach is to clip an impulse, thereby limiting its received energy. Another approach is to use redundancy and discrete Fourier transformation to perform impulse - 78 -noise cancellation C45]. A third method is to use smearing and de-smearing -filters in data transmissions to combat the effects of impulse noise [463. Here, a scheme is proposed to take advantage of the fact that a transmission becomes more white noise limited as the bit duration of the transmission increases. The idea is to simply increase the observation period between decisions. With a longer observation interval, impulse noise effects should be reduced. The widening of the observation interval necessitates the use of multilevel or M—ary modulation schemes to maintain the same data rate given a reduced transmission rate. A potential drawback is the requirement of accurate multilevel thresholds. This above concept was tried by designing and building a 4-level FSK modem and comparing its performance with that of the binary CMFSK modem under similar power line conditions. BER measurements are compiled in Table 7.7 and plotted in Fig. 7.7. The graph shows a trend of lower BER performance achieved by the 4-level FSK modem. The implementation of the 4—level FSK system is described in Fig. 7.8 and Fig. 7.9. It appears to be still suffering from impulse noise effects because its transmitted bit duration of lOOus is not wide enough to overcome an impulse noise whose width is typically around 30us. It is expected that a 16-1evel FSK system with 200us bit duration would yield a lower BER at the same received Eb/No. - 79 -Table 7.7 19.2kbps BER Measurement of a 4-Level FSK and the CMFSK Modem 4-Level FSK Modem Trx Voltage Rcr Eb/No BER 500mV 16. 8dB 5. 7x 1O"5* l.OV 21.8dB 1.3x10-' 1.5V 26.0dB 4.0x10—• 2.0V 27.8dB 7.0xl0-» 3.0V 31.7dB 9.0x10-* CMFSK Modem 500m V 16.8dB S.OxlO"38 l.OV 22.9dB 1.5X10-' 1.5V 27.4dB 4.5xl0—» 2.0V 29.8dB 1.5x10-* 3.0V 33.ldB l.SxlO-" note: Transmitter on Bench 'B'f receiver on Remote 'B*. Fig. 7.7 Comparison of 19.2kbps BER of a 4-level FSK and the CMFSK Modem - 81 -J±L T 5T 6 3 XR-220& 4-Level FSK 7 FH/R AMp *1 <fo>.</?3 5^ 1 / «—1 loo' 'ol' CDfoG6 1 <S D f-081 dY-0 Dat«. (X > Q \3.ZMz. 3. &KHz Q d(T2 *) J tele) Q J15 kHi A = I/KZC = 12. 1 . J KHZ V«f c = 1 ZS KHz Fig. 7.8 Block Diagram of 4-Level FSK Modulator - 82 -u / 9.2^2. 'Loa6"l— Lea. i ?7 Da-ta. Note: "Load" precedes " Ourr,j3 hi (A n a /c> q -r /I B c 1 1 i 'io' o 2 1 o o I •or o o o 'oo' Fig. 7.9 Block Diagram of 4-Level FSK Demodulator - 83 -8. SPECTRUM SPREADING 8.1 Discussion of Spectrum Spreading in Power Line Data  Commun i c at i on s The use o-f spread spectrum technology in power line modems has been suggested by several authors 14,17,18,193. Cited advantages include the ability to combat narrow band impairments and to allow simultaneous user access via Code Division Multiple Access ICDMA3. In addition, Ochsner suggested that relatively high transmission rates may be possible with lower transmission power than -for an equivalent narrow band system in the hostile environment o-f a power line channel 1173. These advantages assume that the power line channel has a much wider bandwidth than the baseband signal spectrum and that the channel attenuation is assumed to be reasonably flat within this bandwidth. Neither assumption is necessarily true in a power line. Transmission curves presented in Chapter 3 and in other papers 117,263 suggest that an usable power line channel extends -from 30kHz to 150kHz and that the variation in transmission difference within this 120kHz bandwidth can often exceed lOdB or even 20dB. Thus, this deviation from a wide and flat channel assumption should be incorporated in the analysis of any spread spectrum communications systems. The result is an expected degradation in performance in terms of a lower equivalent processing gain. Assuming a flat power line bandwidth of 120kHz, a direct sequence CDMA spread spectrum system with 20dB processing gain can, in theory, support 100 simultaneous - 84 -60bps users. In C183, the authors predicted a maximum number o-f 22 simultaneous 60bps users as their system's upper bound performance using their experimental single-user BER result at 60bps. The above hypothetical CDMA system does not include the effects of near/far field and implementation limitation in the CDMA analysis £47,483. The near/far problem usually results when a local receiver tries to pick out a remote and much weaker transmission in the presence of strong co-site transmissions. This problem is often so severe that a conventional DS CDMA system can not be used. A remedy is to raise the processing gain which results in either a reduction in the number of simultaneous users or decreased user throughput. Usually the number of users is retained and the throughput is forgone to preserve the spirit of a CDMA system. In a typical power line channel with average attenuations of 20dB to 30dB, the near/far problem translates into a lOO—fold to 1000—fold decrease in throughput and inclusion of implementation limitations will further reduce this rate. In the hypothetical DS CDMA system described above, the resulting throughput loss could be reduced considerably by using orthogonal codes which have similar spectral properties of long PN sequences 01000 bits long) and by using the power line ac voltage to align and synchronize all user codes. A frequency hopping CDMA system does not suffer as much from the near/far problem, but its use is limited by available power line bandwidth and increased modem complexity. However, modem complexity can be reduced if the ac voltage is used to provide initial hop synchronization. - 85 -Ochsner's suggestion that spread spectrum methods may allow a relatively high transmission rate system to operate with low transmission power is based on the notion that a narrow band system would use higher transmitter power to combat narrow band -fades. This does not mean that a spread spectrum system with a 300kHz spread bandwidth centered at 150kHz will be more robust than a narrow band system operating with a bandwidth o-f 1kHz at 40kHz. In -fact, much o-f the wide band transmitted energy would probably be lost to the channel be-fore reaching a remote user. Spread spectrum methods must be care-fully applied with regard to available channel bandwidth, high level o-f low frequency noise, and low transmissibi1ity of high frequency signal. A reasonably good spectral placement of the spreaded signal should coincide with that range of frequencies which possess good Eb/No figures. In Niederberger's paper C493, he pointed out that "What is important for a quality of a transmission channel is not the transmission characteristic or the disturbance environment on their own, but the relationship of these quantities to each other. The signal-to-noise ratio indicates those parts of the frequency spectrum where the supply network possesses good or bad transmission quality." (see Fig. II and Fig. Ill in his paper). This range of frequencies exists from 50kHz to 500kHz in a residential power line. In an industrial power line, the range is probably in the region between 50kHz and 100kHz (see Table 6.2). A 30kHz to 50kHz channel occupancy is often sufficient protection against narrow band impairments in either an industrial or a residential power line below 200kHz operation. - 86 -Wider channel occupancies are not recommended since the resultant signal spectrum may occupy more of the high noise and low transmission region of the power lines, resulting in reduced SNR and degraded performance. The experimental BER measurements at 1.2kbps, 4.8kbps, and 19.2kbps indicate that the major cause of power line impairments come from impulse noise whose duration approaches that of a data bit at high data rates. Therefore, as the transmission rate is increased, the system becomes more susceptible to impulse noise. This is analogous to pulse jamming in DS spread spectrum system. Consequently, more power is required to overcome impulse noise effects at high data rates (see BER versus Eb/No curves at 4.8kbps & 19.2kbps), which was not mentioned in Ochsner's paper. A power line spread spectrum system may be limited in its usefulness because of insufficient power line bandwidth and high path attenuation between users. But its other advantage of low spectral density is useful in lowering in-band and out-of-band interference to co-site users. For example, this low power flux density feature is important in satelite communication systems design [50,513. In our CMFSK modem, spread spectrum is applied to low rate transmissions to reduce otherwise narrow band interference to low level 'AWGN'-like noise. Coupled with the compact spectrum and rapid sidelobe roll off of the CMFSK modulation, the modem is prevented from generating narrow band tone interference, and it can therefore output more power to overcome line attenuation. - 87 -This will be discussed in more details in section 8.3. 8.2 Potential Spread Spectrum Applications in Q-ffice and  Residential Power Line Communications Given sufficiently wide bandwidth and low attenuation, a power line spread spectrum system can realize many of its stated benefits. For example, if the near/far problem is limited to lOdB and the power line bandwidth becomes 500kHz wide, a DS CDMA system could support 50 simultaneous users with pseudorandom PN codes at 500bps/user. In addition, any narrow band fades could be handled with slight system degradation. Such an ideal channel could be found in a residential house (see Fig. 3.9) with high frequency bypassing between the phases and signal isolation of the incoming power lines to confine most of the signal energy within the power lines of the house (see Fig. 8.1). Another possibility lies in an office floor where a section of the power lines is isolated and bypassed to serve as a local area network. The signal bypass and isolation could be performed at the service box to isolate the section of the power line network used in communications. In addition, a relay modem could be used to ensure reliable communications across phases (see Fig. 8.2). The relay could be installed in the service box along with the isolation and bypass hardware. In a custom installation, the number of simultaneous users could be traded off against the required throughput rate. - 88 -r 180° ~r> L Zsolat/'on. Bypass Service Box -O Neutral Fig. 8.1 Proposed High Frequency Bypassing and Isolation of a Power Line Network X.solext iorx 0° -H> /So0 H> Neutra.1 'Se.n/ice. Box Fig. 8.2 Proposed Across-Phase Modem Relay - 89 -8.3 Bene-fits o-f Spread Spectrum in Low Data Rate  Commun i c at i ons The application of spread spectrum to low speed transmission is described in this section. The purpose is to reduce narrow band interference generated by low speed transmissions, and to enable the transmitter to output more power to overcome path attenuation. The spectrum of a narrow band system, BSR X—10, is compared with that of the CMFSK modem at the same average output of IV. Both spectra are displayed in Fig. 8.3. The output of the BSR X-10 system exhibits a sharp spectral peak and a high percentage of energy concentrated in its narrow sidelobes. On the other hand, the CMFSK output shows a wider main lobe and only a small fraction of energy is contained in the sidelobes (main lobe contains >99% of the energy). As a result, the BSR X-10 system will generate tone interference and the CMFSK modem will generate 'AWSN'-Iike noise. A listening test was carried out to confirm the effect of narrow and wide band interference on AM radio stations which occupy the 540-1600kHz frequency band. A *30 AM/FM clock radio was placed side by side the modem in an residential outlet. When the BSR unit was turned on, strong tone interference was heard in several stations, with its effect strongest in weak stations. Operating under the same conditions, except with a larger transmitter output voltage of 4V as compared to IV in the BSR X-10 unit, the interference from the CMFSK modem was observed as a less annoying rise in background 'hissing' noise. This result indicates the benefit of spread spectrum in 8.3 Spectral Comparison of BSR X-10 and CMFSK Modem - 91 -reducing narrow band interference, thereby allowing more output power to be transmitted without infringing on FCC regulations or other applicable emission controls. 8.4 Application of Databand Spread Spectrum in the CMFSK  Power Line Modem In the CMFSK modem, spectrum spreading was performed at the databand level. The idea is to transmit a 19.2kbps PN sequence which has been Exclusive ORed with a lower rate data sequence. The receiver can recover the data sequence by Exclusive ORing the received bit stream with the same PN sequence and then making majority logic decisions. This baseband spread spectrum method is shown in Fig. 8.4. It is similar in some respects to message scrambling/descrambling in a digital communication system. Therefore, a transmission can be made secure by using long PN sequences generated by non—linear shift register feedback. The drawback of baseband spreading is the requirement of a minimum SNR to enable the recovery of each code bit before the data information can be retreived. In a DS or FH spread spectrum system, the data BER rather than the code BER requirement dictates the required SNR. However, our proposed scheme offers a faster acquisition speed, on the order 50 bits to 100 code bits, and is less complex in hardware realization than a conventional spread spectrum system. This is a classical tradeoff problem between acquisition speed and received SNR. Basically, a higher SNR conveys a more reliable information and hence allows a quicker acquisition and simpler implementation. Channel N°iSe. J(k)@ PA/CK) F5K Mc^uhtor Transmitter F5fs Demodulator RAW blTS PN ^aJe OK ftecoven Code Zynckronijdt-t'on OLK *5YNC" JL_ Code. Ge*e rato r-PNCK) O -o Fig. 8.4 Block Diagram of Databand Spread Spect rum - 93 -8.5 Code Synchronization using a Digital Matched Filter PN code acquisition is achieved via a digital version of the PN code matched filter discussed in C523. Basically, a delayed version of the incoming code chips are bit-by-bit multiplied by a predetermined PN sequence and summed. The summation output will exceed a threshold whenever the received code sequence is matched to the local sequence. Upon tripping the threshold, the receiver code is locked onto the incoming sequence and the retrieval of data bits begins. If a PN sequence is generated by a M-stage shift register, a M—stage DMF is sufficient for acquisition of the PN sequence. If the incoming bit stream has errors, more than M stages are required to give high acquisition and low false alarm probability because of the many high 'sidelobes' in the output of the DMF or sliding correlator C533. This is always the case in an acquisition system in which only a small portion of a long PN sequence or short codes are used in the correlation of the acquisition process. Our CMFSK modem has a 48-stage DMF. The taps used in the DMF multiply—and—sum network were obtained from a print-out of a 11—stage shift register (code length 2047) over 200 bits. A 48-bit sequence containing an equal number of 'l's and *0"s was chosen as the synchronization sequence and corresponding DMF tap values. A 11-bit sequence of fifty bits preceding this synchronization sequence was chosen as the initial code word state of the transmitter PN code generator. Therefore, a transmission begins by sending these fifty bits followed by the 48—bit 'sync* sequence. In the receiver, these - 94 --first -fifty code chips set the bit clock recovery in motion and are appropriately called the bit clock training sequence. Data transmission begins at the end of the 'sync' sequence. The block diagram of the sliding correlator or DMF is given in Fig. 8.5. The 48-bit synchronization sequence and the initial 11-bit code word state which initiates a transmission are omitted since other PN sequences can be used without loss of performance provided the PN code length exceeds 2000 bits. Reference C54D contains a circuit which provides PN code synchronization. 8.6 Tolerance to Power Line Small Error Bursts by Interleaving From Chan's thesis on error statistics [263, it seems that most errors are highly correlated, i.e. small bursts are common. Therefore, our proposed transmission scheme will suffer from small bursts of code chip errors. This is like a DS spread spectrum system being effectively jammed by a pulse jammer and a FH spread spectrum system being effectively jammed by a tone jammer. In both cases, forward error correction reduces the jamming effect. Observation of error patterns indicates that there is a large error—free and a small error—prone region. Obviously, the error-prone region is concentrated near the strong impulses in the received signal. Such a region may contain single-bit or multiple-bit errors. A small error burst made up of multiple errors will often reduce the capability of a random error correction code. Although a burst error code can handle small error bursts, it will not operate properly if a single-bit error then occurs outside the error correcting span of the code. '''glial Moitchea. Frl-ter CLK PNCL) & CLK CLK CD4-OS1-R o • • • ?NCi+K)S 7, / / v*\nen in "Sy»c" _/\_ Q3 Qx n n faralh) kn ^—J Qs 2V)T,4 coq-oai PS Pi Parallel Load 9 • "5jnc" pul: se +/5v Zo<ra/ PA/ Otnera±of-Ci' = 0 we aw Investor . If Ci = I , Siraiqht Connection. . + i , •- • j Ci+H-8 are the 4-8-brr Sync^on/'jatfort Sequence. Fig. 8.5 Block Diagram of Digital Matched Filter Threshold De-ttpc-iop <0 - 96 -There-fore, it is proposed to disperse the transmitted bits and then re-assemble them in the receiver. The benefit is to allow the use of random error correction codes on a small error burst now separated into mostly single-bit errors, and also on other random errors as well. The penalty of performing both burst and random error correction is a reduction in throughput. The de-assembly and re-assembly of transmission is accomplished by bit interleaving and de-interleaving in the transmitter and receiver respectively. Interleaving basicly randomizes error occurances and allows the use of random error correction codes in a burst error channel. The error randomization ability of bit interleaving is also useful in low data rate transmissions. The 19.2kbps received bi stream in the receiver undergoes majority logic decision before data bit is extracted. A small error burst concentrates all the 19.2kbps bit errors in a small time span and drasticly increases the probability of a majority decision error. Therefore, an interleaver can effectively spread out a small error burst and reduce its impact on majority logic decisions. 8.7. Experimental Results for Interleaved Spread Spectrum Data  Transmi ssi ons A 31-bit interleaver and de-interleaver were built to test the above concept (Fig. 8.6). These were chosen because of hardware simplicity. When interleaving is used with hard decisions, we use the term Interleaved Majority Decision. With no interleaving and hard decisions we use the term Simple D/\rA CLK DATA CLK PATA Q± CLK MTA Qi QQ CLK CP+ 09 f DATA Ql CLK DATA Ql a7 Pi Pa 'Q' Pi PS CLK • Loacj CDf-oZj DATA Q<3 I*-ber/easing or De-/iter/ea^ir^ Csee. Tqb/e 8-1 ) m P9 ' ' ' P/6 fara-Uel PL Pe CLK DATA 1 \P2f rar«,llel Ct-K Loud DATA >4 aKf-31 Inter leais-eA on. De-Interleaved b" CLK CJH-SB& Qif. CFPPi DPz DP? t>fr ft I ft I PE MR "0< CLK C.DH-52.G) cF DPL V b' o -o' Fig. 8.6 Block Diagram of Inter1eaver/De-inter1eaver - 98 -Table 8.1 31-Bit Interleaver/De-interleaver Connections Interleaver De-interleaving Ql -> P8 Ql -> P4 Q2 -> P16 Q2 -> P8 Q3 -> P24 Q3 -> P12 Q4 -> PI Q4 -> P16 Q5 -> P9 Q5 -> P20 06 -> P17 Q6 -> P24 Q7 -> P25 Q7 -> P28 Q8 -> P2 Q8 -> PI Q9 -> P10 Q9 -> P5 Q10 -> P18 Q10 -> P9 Qll -> P26 Qll -> P13 Q12 -> P3 Q12 -> P17 Q13 -> Pll Q13 -> P21 Q14 -> P19 Q14 -> P25 Q15 -> P27 Q15 -> P29 Q16 -> P4 Q16 -> P2 Q17 -> P12 Q17 -> P6 Q18 -> P20 Q18 -> P10 Q19 -> P28 Q19 -> P14 Q20 -> P5 Q20 -> P18 Q21 -> P13 Q21 -> P22 Q22 -> P21 Q22 -> P26 Q23 -> P29 Q23 -> P30 Q24 -> P6 Q24 -> P3 Q25 -> P14 Q25 -> P7 Q26 -> P22 Q26 -> Pll Q27 -> P30 Q27 -> P15 Q28 -> P7 Q28 -> P19 Q29 -> P15 Q29 -> P23 Q30 -> P23 Q30 -> P27 Q31 -> P31 Q31 -> P31 - 99 -Majority Decision. A third scheme uses so-ft decisions to retrieve a data bit by multiplying the incoming analog 19.2kbps bit stream with a local PN sequence. This process we call Correlation Decision. The three schemes are outlined in block diagrams in Fig. 8.7. The performance of these schemes is compared to the narrow band results obtained in the 1.2kbps and 4.8kbps data rate transmissions in Chapter 7. The corresponding BERs, assuming random error statistics, are given by: 1. Interleaved Majority Decision, BER = p r"'= 2. Simple Majority Decision, BER = p 3. Correlation Decision, BER = QC (nEc/No)1/2 3 4. Narrow band, BER = QC (Eb/No) 3 where p = bit error rate of a code chip = QC (Ec/No)1/2 3, n = processing gain (divisible by 2), Ec = average energy in a code chip, Eb = average energy in a data bit = nEc. In AWGN, both the correlation and narrow band methods are equivalent and have better BER performance over either the simple or interleaved majority logic decision method. Experimental results at 1.2kbps and 4.8kbps data rate at 19.2kbps code rate are plotted in Fig. 8.8 and Fig. 8.9. From these graphs, it is observed that: 1. Narrow band results are best at both data rates. 2. Interleaved majority decision is next best, with the BER decreasing rapidly at high Eb/No values. 3. Both simple majority and correlation decision lead to almost identical BER performance as the two other schemes at 1.2kbps data rate. - 100 -Z^teyrn hi/bumf Code clock feco^e ry CLY-cdc Generator Jtt6 MrA C/-K 1 O r*l-•&1~ejrate/ Dump I %\SCLK Code Clock Re Co 31-hii- Pe -inferlea v-er CLK j]JSr/vc" Code Syr,tkremjaiion & f>A) Cc^e Generator T7 xt6 lArA clk -£>MTA C/'i) Tr\-terleaved Majon'-ty T>ec]sior% Integrate / j>ur»p 7T tl&CLK Fecev-try CLK. Switching Multiplier £ode {fuehrer i^nf/cn tN Code (senera-kor (iii) Corre/a--kien, T> eci's t'otx. R 16-fag* Vtytnf Fig. 8.7 Block Diagram of Simple Majority Logic Decision, Interleaved Majority Logic Decision, and Correlation Decision Methods -{> DATA Fig. 8.8 Comparison of 1.2kbps BER Performance of the Interleaved Majority Decision, Simple Majority Decision, Correlation Decision, and Narrow Band Methods - 102 -Fig. 8.9 Comparison of 4.8kbps BER Performance of the Interleaved Majority Decision, Simple Majority Decision, Correlation Decision, and Narrow Band Methods - 103 -4. Both simple majority and correlation decision gives noticeably worsar BER performance when compared to narrow band results in in 4.8kbps data transmission. The interleaved majority decision, however, is able to -follow closely the narrow band result. The benefit of a simple 31—bit interleaver is obvious in (4). The interleaver separates every adjacent bit in the original bit stream by at least 3 bits in the interleaved bit stream. A 63—bit and 127-bit interleaver would provide 4—bit and 5-bit minimum separation, respectively. Higher degree of inter— leaving is not necessarily beneficial because of the periodicity in bit error occurance, and also, because of the long delay between reception and the output of the first data bit. The interleaver used in this report is a block periodic interleaver and its performance may be degraded in a periodic burst error environment. In this situation, a puesdorandom interleaver may prove to be more robust C553. This PN code interleaved spread spectrum technique combined with CMFSK modulation has several potential advantages: 1. Compact transmission spectrum independent of data rate. 2. Fast sidelobe roll off independent of data rate. 3. 'AWGN'-like wideband interference independent of data rate. 4. Tolerance to small power line induced error bursts. The main disadvantage with this scheme is that a minimum SNR must be present at the receiver in order to operate the code clock recovery circuit. This minimum SNR corresponds to a code BER of .01 to .05. When this BER is exceeded, the recovered bit - 104 -clock shows high jitter, thus increasing the possibility o-f lost tracking and missed acquisition. An improvement o-f this analog clock recovery circuit may be possible by using a digital clock recovery circuit. Although a digital PLL controlled bit clock recovery suffers from higher residual clock jitter because of the discrete nature of the circuit, it offers robustness, ease of implementation, elimination of an ultra stable VCO or VCXO (voltage controlled oscillator or voltage controlled crystal oscillator) used in an analog PLL, high reliability, crystal timing leading to drift-free and adjustment-free operation, and easy VLSI adaptability. These are important considerations in the design of low cost power line modems. Although the narrow band results consistently yield the best BER results, adjustments of the two offset frequencies become harder as the data rate is decreased. Such adjustments are necessary because of the loose tolerance, low cost components used in the prototype CMFSK modem. Therefore, an added bonus in using spread spectrum is the ability to use loose tolerance parts and components because of the large frequency difference between the two offset frequencies. However, if the offset frequencies can be accurately set by a crystal controlled FSK modulator in the transmitter and if a digital FSK demodulator can be used in the receiver, narrow band methods may prove to be useful. - 105 -9. CONCLUSIONS 9.1 Summary In this thesis, the conceptualization and design of a Coherent Minimum Frequency Shift Keying modulation technique suitable for communications over an intrabuilding polyphase power line network is described. The technique includes optional databand spectrum spreading. An actual modem is then implemented and tested. This modem is subsequently used to obtain new results detailing the performance of CMFSK signalling on intra building electric power distribution networks. The average Bit Error Rate measurement at 1.2kbps, 4.8kbps, and 19.2kbps transmission speeds indicate CMFSK transmissions are largely white noise limited at low data rates but become impulse noise limited at high data rates. Typical Eb/No values to achieve IO-3 BER performance are 11.5dB at 1.2kbps, 13dB at 4.8kbps, and 20dB at 19.2kbps data rate, respectively. Multilevel and M-ary modulation schemes are proposed to reduce impulse noise effects by making a transmitted bit longer relative to an impulse. A 4-level FSK system, which was designed and built to verify this concept, showed some BER improvement. Further suppression of impulse noise effects may be possible with more levels, since BER degradation due to multilevel threshold sensitivity problems is not as serious as the jamming effect of impulse noise. The CMFSK modulation technique was found to emit less - 106 -harmful interference to AM stations when compared to an existing commercial product, BSR X-10. This interference reduction is attributed to the compact spectrum and rapid sideband rolloff of the CMFSK modulation scheme. The application of CMFSK technology to low data rate transmissions was made more robust by using a databand spread spectrum technique. On power line channels, narrow band transmissions which accompany low data rates are subject to potential narrow band impairments. A commercial—use power line modem is restricted by law from transmitting high power to overcome narrow band impairments. As well, modem costs may be reduced due to relaxed power output requirements. Properly designed spread spectrum signalling enables a power line modem to survive narrow band impairments at an acceptable transmitter power level. The degree of spreading is limited by the power line bandwidth, data bandwidth, noise, and transmission characteristics. The effect of impulse noise jamming is observed when the bit error patterns at 19.2kbps transmission contain small error bursts. Therefore, a periodic 31—bit block interleaver was built to disperse small error bursts. This interleaved spread spectrum scheme proved its robustness when BER of an interleaved and non-interleaved 4.8kbps data rate, 19.2kbps code rate spread spectrum system was compared. The interleaver scheme gives close to ideal BER performance. In higher speed transmission, the randomization effect of an interleaver enables the use of good random error - 107 -correction codes at a slight reduction in throughput. Besides its ability to combat narrow band -fades, the spread spectrum CMFSK modulation is able to reduce its narrow band emissions and thereby allows adequate power to be transmitted to overcome path attenuations. This has important implications because it is now possible to satisfy emission regulations even at enhanced output levels. 9.2 Cost Estimate of the CMFSK Modem The cost of our prototype receiver lies heavily with its front end of powr line coupling network, BPF, and FSK demodulator ($25). The rest of the receiver circuitry can be implemented from CMOS ICs ($10) and a crystal clock (#5). The modem power supply is also relatively simple ($10). The receiver section of the modem would typically be under #60 while the transmitter section would typically cost an additional #10 for the FSK modulator and linear power amplifier ICs. 9.3 Recommendations For Further Research There are several items which may be worth investigating: 1. Means and practicality of high frequency bypassing in a residential house and an office floor via the service box, to improve across—phase transmission capability. 2. Use of high carrier frequency, above 250kHz, to realize transmission rates above 50kbps. These high data rates are desirable in some office automation environments. These high data rates are of course limited to short distance coverage or power line conditions defined in Section 8.2. 3. FEC by random error correction codes using either the shortened BCH(16,8) of Dunbar et al. C56] or other codes - 108 -described by Batson et al. C573. A suitable code should have double or even triple errors correction capability. 4. FEC by burst error correction codes using Fire codes with a 10-bit burst correcting capability. 5. Power line communication protocols which provide reliable and robust networking with reasonable throughput. A likely candidtate is Carrier Sense Multiple Access with positive acknowledgement. 6. Data pulse shaping, modulator output filtering, and choice of spectral efficient modulation methods (see Feher C30,583 and Klose et al. C593) to control out-of-band emissions at enhanced output levels. 7. Randomization ability of a 63-bit or 127-bit interleaver on power line induced errors as compared with that of a puesdorandom interleaver (see Clark et al. C55D). 8. Impulse noise suppression by using smearing and de—smearing filters (see Beenker et al. C46D). 9. Efficient power line spectrum ultilization by having one band assigned to 'ORIGNATE' and another band to 'REPLY*. 10. Multilevel or M-ary schemes to combate impulse noise errors. Such a possible scheme is multilevel FSK, with FSK detection using FFT to reduce decision threshold sensitivity problem (see Klose et al. C593 and Feher C583). 11. Use of digital bit synchronization methods to improve the reliability and robustness of code clock recovery (see Dunbar et al. C561 and Payzin C603). 12. Design and implement an interleaved soft decision correlation data detector to check if it yields any improvements over interleaved hard decision majority logic decision discussed in Section 8.7. Another area worth looking into is the voice communications capability of this CMFSK modem. In Section 5.5, it was mentioned that the raw data output of the XR-2211 FSK demodulator could be ultilized as an analog output. This is possible because the CMFSK modem is essentially a commercial power line FM intercom converted to carry digital bit pulses. Therefore, analog voice transmissions from this modem is expected to be reasonably high - 109 -quality and low distortion. This is by no means an exhaustive list of ideas for future work. Some of the above involve A/D conversion at the front end and digital signal processing by the receiver. This seems to be the trend in future communications systems because a digital receiver is inherently more reliable and robust than an analog equivalent, and the cost can be lowered through large scale producti on. - 110 -REFERENCES 1. D.R. Beuerle, A.G. Hudson, and H.J. Fiedler, S5B Carrier  for Utility Control and Communication. IEEE National Telecommunication Conference, pp. 2.1.1-2.1.7, 1976. 2. Arthur CM. Chen, Automated Power Distribution. IEEE Spectrum, pp. 55-60, April 1982. 3. Glen Lokken, Neil Jagoda, and Raymond J. D'Auteuil, The  Proposed Wisconsin Electric Power Company Load Management  System Using Power Line Carrier Over Distribution Lines, IEEE National Telecommunication Conference, pp. 2.2.1-2.2.3, 1976. 4. Peter K. van der Gracht, A Spread Spectrum Modem for Use  on Electrical Power Lines. M.A.Sc Thesis, Electrical Eng. Dept., University of British Columbia, Canada, January 1982. 5. J.B. O'Neal, Jr., Overview of Power Distribution Line  Carrier Communication Systems. IEEE Globecom'83, pp. 464-467. 6. J. Carlos Dangelo and Sarosh N. Talukdar, A Stochastic  Model for PLC systems. IEEE Transactions on Power Apparatus and Systems, vol. PAS-100, No. 11, pp. 4464-4472, November 1981. 7. M.P. Perry and M.R. Stambach, A System Analysis of Power  Line Carrier Communications with Gas Insulated Conductors  and Overhead Lines. IEEE 1980 Canadian Communications & Power Conference, pp. 240—243. 8. Ken C. Shuey, Distribution Transformers at Power—Line  Carrier Frequencies. IEEE Globecom'83, pp. 483-486. 9. W.R. Vincent, J.M. Clemmemsen, and R.L. Bollen, The  Measurement of Radio Noise Associated with Distribution  Lines. IEEE Globecom'83, pp. 473-477. 10. K.W. Whang, G.C. Cagle, and J.W. Smart, The power  Distribution System as a Communication Medium for Load  Management and Distribution Automation. IEEE Globecom'83, pp. 478-482. 11. Kay N. Clinard, Utility Experience and Desires for  Distribution Power Line Carrier Communications. IEEE Globecom'83, pp. 487-491. 12. F.W. Gutzwiller, J.E. Francis, Jr., E.K. Howell, and W.R. Kruesi, Homenet: A Control Network for Consumer Applications. IEEE Transactions on Consumer Electronics, vol. CE-29, No. 3, pp. 297-304, August 1983. - ill -13. Masahiro Inoue, Kazuho Uemura, Yoshiji Minagawa, Mitsunobu Esaki, and Yoshiyuki Honda, A Home Automation System. IEEE Transactions on Consumer Electronics, vol. CE—31, No. 3, pp. 516-527, August 19S5. 14. William R. Kruesi and Phillip R. Rogers, Residential  Control Considerations. IEEE Transactions on Consumer Electronics, vol. CE-28, no. 4, pp. 563-570, November 1982. 15. Federal Communicatios Commission Rules and Regulations, Volume 2, Part 15, Subpart A. 16. Carole C. Harris, Impact o-f FCC Requlartory Proceedings  Power Utility Telecommunication Systems. IEEE 1978 National Telecommunications, pp. 9.2.1—9.2.4. 17. Heinz Ochsner, Data Transmission on Low Voltage Power  Distribution Lines using Spread Spectrum Techniques, IEEE 1980 Canadian Communications & Power Conference, pp. 236-239. 18. P.K. van der Gracht and R.W. Donaldson, Communication  Using Pseudonoise Modulation on Electric Power  Distribution Circuits. IEEE Trans, on Communications, vol. COM-33, No. 9, pp. 964-974, September 1985. 19. Botaro Hirosaki, Satoshi Hasegawa, and Kaoru Endo, A Power  Line Home Bus System Using Spread-Spectrum Communication  Technologies. 1985 International Conference on Consumer Electronics, conference paper. 20. Dan Hariton and Paul Patterson, AC Power Line Modem For  Consumer and Industrial Environments, 1985 International Conference on Consumer Electronics, conference paper. 21. BSR X-10 Owners Manual, BSR (Canada) Ltd. 22. Line upon line. IEEE Spectrum, pp. 25, September 1985. 23. Nonwire Multidrop Network Users Guide, CYPLEX Division, Controlonics Corporation, MA, USA. 24. PowerLAN™, ExpertNets, Acton, MA, USA. 25. Steven E. Sarns, Home Control Computer: Part II. Radio-Electronics, pp. 64-70, May 1984. 26. Morgan H.L. Chan, Communication Channel Characteristics  and Behaviour of Intrabuildino Power Distribution Circuits M.A.Sc Thesis, Electrical Eng. Dept., University of British Columbia, Canada, September 1985. - 112 -27. Albert A. Smith, Jr., Power Line Noise Survey, IEEE Transactions on Electromagnetic Compatibility, vol. EMC-14, no. 1, pp. 31-32, February 1972. 28. Roger M. Vines, H. Joel Trussel1, Louis J. Gale, and J. Ben O'Neal, Jr., Noise on Residential Power Distribution  Circuits. IEEE Transactions on Electromagnetic Compatibility, vol. EMC-26, No. 4, pp. 161-168, Nov. 1984. 29. R.E. Owen, W.R. Vincent, and W.E. Blair, Measurement o-f  Impulsive noise on Electric Distribution Systems. IEEE Transactions on Power Apparatus and Systems, vol. PAS-99, no. 6, pp. 2433-2438, Nov./Dec. 1980. 30. Kamilo Feher, Digital Communications. Satellite/Earth  Station Engineering. Prentice Hall Inc., Englewood Cli-f-fs, N.J., 1983. 31. Subbarayan Pasupathy, Minimum Shi-ft Keying: A Spectrally  Efficient Modulation. IEEE Communications Magazine, vol. 17, no. 7, pp. 14-22, July 1979. 32. W.C. Lindsey, Phase-Shifted-Keyed Signal Detection with  Noisy Reference Signals. IEEE Transactions on Aerospace and Electronic Systems, vol. AES-2, no. 4, pp. 393-401, July 1966. 33. V.K. Prabhu, PSK Performance with Imperfect Carrier Phase  Recovery. IEEE Transactions on Aerospace and Electronic Systems, vol. AES-12, no. 2, pp. 275-285, March 1976. 34. V.K. Prabhu, Imperfect Carrier Recovery Effect on Filtered  PSK Signals. IEEE Transactions on Aerospace and Electronic systems, vol. AES-14, no. 4, pp. 608-615, July 1978. 35. Shinjiro Oshita and Kamilo Feher, Performance of Coherent  PSK and DPSK Systems in an Impulsive and Gaussian Noise  Environment, IEEE 1981 International Communications Conference, pp. 56.4.1—56.4.5. 36. T.A. Schonhoff, Linear Demodulation of CPFSK in Impulsive  Atmospheric Noise. IEEE 1978 Canadian Communications & Power Conference, pp. 46. 37. John A. Malack and John R. Engstrom, RF Impedance of  United States and European Power Lines. IEEE Transactions on Electromagnetic Compatibility, vol. EC—IB, no. 1, pp. 36-38, February 1976. 38. Roger M. Vines, H., Joel Trussel1, Kenneth C. Shuey, and J.B. O'Neal, Jr., Impedance of the Residential Power—  Distribution Circuit. IEEE Transactions on Electromagnetic Compatibility, vol. EMC-27, no. 1, pp. 6-12, February 1985. - 113 -39. EXAR Integrated Systems, Inc. , Stable FSK Modems Featuring  the XR-2207. XR-2206. and XR-2211. AN-01. 40. John G. Proakis, Digital Communications. McGraw-Hill Book Company, 1983, ch. 4, pp. 200-203. 41. R.E. Ziemar and W.H. Tranter, Principles o-f Communications  : Systems. Modulation, an Noise. Houghton Mifflin Company, Boston, 1976, ch.7, pp. 340-341. 42. EXAR Integrated Systems, Inc. , Clock Recovery System. AN-19. 43. A.D. Spaulding and D. Middleton, Optimum Reception in an  Impulsive Interference Environment — Part Is Coherent  Detection; Part II — Incoherent Detection. IEEE Trans, on Communications, vol. COM-25, no. 9, pp. 910-934, Sept. 1977. 44. H.H. Lu and B.A. Eisenstein, Detection of Weak Signals in  Non-Gaussian Noise. IEEE Trans, on Information Theory, vol. IT-27, no. 6, pp. 755-771, November 1981. 45. Jack Kei1 Wolf, Redundancy, the Discrete Fourier Transform  and Impulse Noise Cancellaion. IEEE Transactions on Communications, vol. COM-31, no. 3, pp. 458-451, March 1983. 46. G.F.M. Beenker, T.A.C.M. Claasen, and P.J. van Gervwn, Design of Smearing Filters for Data Transmission Systems. IEEE Transactions on Communications, vol. COM-33, no. 9, pp. 955-963, September 1985. 47. Raymond L. Pickholtz, Donald L. Schilling, and Laurence B. Milstein, Theory of Spread—Spectrum Communications — A  Tutorial. IEEE Transactions on Communications, vol. COM-30, no. 5, pp. 855-884, May 1982. 48. Irving R. Smith, Tradeoff Between Processing Gain and  Interference Immunity in Co-Site Multichannel Spread  Spectrum Communications. IEEE Transactions on Communications, vol. COM-30, no. 5, pp. 959-966, May 1982. 49. F.P. Niederberger, Return Signalling Systems Used as a  Compliment to Low Frequency-Power Line Carrier (LF—PLC). IEEE 1980 Canadian Communications & Power Conference, pp. 181-184. 50. J.W. Seyl and M.H. Kapel1, Application of Spread Spectrum  to the Shuttle Orbiter Communication System. IEEE National Telecommunication Conference, pp. 32.1.1-32.1.5, 1976. 51. Edward A. Torerro, A Macromarket for microstations. IEEE Spectrum, pp. 59, November 1985. - 114 -52. Stephen S. Rapport and Donald M. Grieco, Spread-Spectrum  Signal Acquisition: Methods and Technology. IEEE Communications Magazine, vol. 22, no. 6, pp. 6—21, June 19B4. 53. Noman Mahmood, Spread Spectrum Communication System -for  Civil Use. Wireless World, pp. 76-80, March 1983. 54. Hewlett-Packard Company, Model 1645A Data Error Analyzer, service manual, 1976. 55. George C. Clark, Jr. and J. Bibb Cain, Error-Correction  Coding for Digital Communications. Plenum Press, New York, 1981, ch.B, pp. 345-351. 56. B.J. Dunbar, D.V. Gupta, M.P. Horvath, R.E. Sheehey, and S.P. Verma, Dataport-Channel Units for Digital Data System  56—kbps Rate. The Bell System Technical Journal, vol. 61, no. 9, pp. 2741-2756, November 1982. 57. B.H. Batson and G.K. Hutch, Modulation and Coding for  Spread Spectrum Communications Systems. IEEE National Telecommunication Conference, pp. 32.5.1—32.5.5, 1976. 58. Kamilo Feher, Digital Communications. Microwave Applications. Prentice Hall Inc., Englewood Cliffs, N.J., 1981, ch. 3, ch. 6, & ch. 7. 59. Dirk R. Kloss and Gwendolyn L. Freeman, Adaptive Digital  Detection of FSK Signals. IEEE 1978 Canadian Communications & Power Conference, pp. 9—12. 60. A. Erbil Payzin, Analysis of a Digital Bit Synchronizer. IEEE Transactions on Communications, vol. COM-31, no. 4, pp. 554-560, April 1983. - 115 -BIBIOBRAPHY 1. Phillip M. Hopkins, Pseudonoise Synchronization at Low  Siqnal-to-Noise Ratio. 1976 National Telecommunications Conference, pp. 32.2.1-32.2.5. 2. Laurence B. Milstein and Donald L. Schilling, The Effect  of Frequency-Selective Fading on a Noncoherent FH-FSK  System Qperting with Partial-Band Tone Interference. IEEE Transactions on Communications, vol. COM—30, no. 5, pp. 904-912, May 1982. 3. Mitsuhiko Mizuno, Randomization Effect of Errors by Means  of Freguency-Hopping Technigues in a Fading Channel. IEEE Transactions on Communications, vol. COM—30, no. 5, PP. 1052-1056, May 1982. 4. J.R. Nicholson and J.A. Malack, RF Impedance of Power  Lines and Line Impedance Stabilization Networks in  Conducted Interference Measurements. IEEE Transactions on Electromagnetic Compatibility, vol. EC—15, no. 2, pp. 84-86, May 1973. 5. Steve Ciarcia, Build a Power Line Carrier Current Modem. Byte, pp. 36-42, August 1983. 6. R.C. Dixon, Spread Spectrum Systems. 2nd ed., New York: Wiley, 1984. 7. Don Lancaster, Active-Filter Cookbook. Howard W. Sams & Co., Inc., 1975. - 116 -APPENDIX A. POWER LINE NOISE SPECTRAL DENSITY DETERMINATION A simple spectrum analyzer was constructed out o-f three 6th order BPFs (Fig. A.l). The theorectical center frequencies and bandwidths of the filters are 30+5kHz, 60+10kHz, and 120+20kHz, with each BPF's Q equal to 3. A higher Q is limited by hardware difficulties. Their measured parameters are listed in Table A.l. This spectrum analyzer was then used to collect noise voltage, Vnal.., and converted to noise spectral density, No: No = Power Line Noise Spectral Density = V2 -—notmm ( Gain x Attenuation )= x BPF Bandwidth This No value is across the power line impedance in parallel with the coupling's resistance. C37,383 gives some typical power line impedance. Coupling's resistance is the sum of the internal resistance, Ri„t of 3.0+.3 ohm, and the input resistance Rm of 12 ohm. The values of 'Gain' and 'Attenuation' are those found in Table A.l to Table A.5. The noise spectral density data shown in Table 4.1 and Table 4.2 is derived from Table A.6 and Table A.7. A Briiel & Kaejr analog voltage meter, Model #2426, is used to provide the voltage readings. Two 4th order BPFs were incorporated in the receiver to perform the carrier selection in Section 6.3. Subsequently, both filters were used to collect the cross sectional noise density data shown in Table 4.3. The parameters of these two 4th order BPFs are listed in Table A.2. i Spray ue ~fcav6for*t\r * * Z2-tOQ •O MID O-O HIGH Fig. A.l A Simple Spectrum Analyzer used in Power Line Noise Spectral Density Measurement - 118 -Table A.1 Parameters of 6th Order Band Pass Filters fcp Gain BW—f t-iiat->—3CJB- f IOM-S^B LOW 24.6kHz 12.05 30.6kHz - 19.6kHz = 11.0kHz MID 58.7kHz 11.05 71.4kHz - 48.2kHz = 23.2kHz HIGH 143.8kHz 11.05 179.5kHz - 119.0kHz = 60.5kHz Table A.2 Parameters of 4th Order Band Pass Filters fe> Gain BW-f h i 9h-3dB—f I ow-3dB MID 57.3kHz 6.0 71.2kHz - 42.0kHz = 24.2kHz HIGH 137.8kHz 5.5 166.0kHz - 114.0kHz = 52.0kHz Table A.3 Frequency Response of Power Line Coupling Terminated at 100k ohm Frequency V±„ Vout Attenuation 10kHz 4.90 V 4.87V O.lldB 20kHz 4.95 V 4.92V 0.lldB 30kHz 4.97V 4.95V 0.07dB 40kHz 4.99V 4.97 V 0.07dB 60 kHz 5.00 V 4.98V 0.07dB 80kHz 5.00 V 4.99 V 0.03dB 100kHz 5.01V 5.00 V 0.03dB 120kHz 5.05 V 5.02 V 0.03dB 150kHz 5. 03V 5.02V O.OldB 200kHz 5.01V 4.99V 0.07dB 300kHz 4.90 V 4.88 V 0.07dB - 119 -Table A.4 Frequency Response of Power Line Coupling Terminated at 12 ohm Frequency Vlri Vout Attenuation 10kHz 200mV 130mV 3. 74dB 15kHz 200mV 147mV 2. 67dB 20kHz 200mV 150mV 2. 50dB 30kHz 200mV 155mV 2. 21dB 40kHz 200mV 157mV 2. lOdB 50kHz 200mV 158mV 2. 05dB 60 kHz 200mV 158mV 2. 05dB 70kHz 200mV 160mV 1. 94dB 80kHz 200mV 160mV 1. 94dB 90kHz 200mV 160mV 1. 94dB 100kHz 200mV 160mV 1. 94dB 110kHz 200mV 160mV 1. 94dB 120kHz 200mV 160mV 1. 94dB 130kHz 200mV 160mV 1. 94dB 140kHz 200mV 159mV 1. 99dB 150kHz 200mV 159mV 1. 99dB 160kHz 200mV 159mV 1. 99dB 170kHz 200mV 158mV 2. 05dB 180kHz 200mV 158mV 2. 05dB 190kHz 200mV 158mV 2. 05dB 200kHz 200mV 156mV 2. 16dB - 120 -Table A.5 Frequency Response of Power Line Coupling Terminated at 3.3 ohm Frequency V4r, Vovlt Attenuation 10kHz 200mV 36.5mV 14.7dB 15kHz 200mV 55.0mV 11.2dB 20kHz 200mV 69.OmV 9.24dB 30kHz 200mV 89.5mV 6.98dB 40 kHz 200mV 100.5mV 5.98dB 50kHz 200mV 108.0mV 5.35dB 60kHz 200mV 110.5mV 5.15dB 70kHz 200mV 113.OmV 4.96dB 80 kHz 200mV 113.0mV 4.96dB 90kHz 200mV 113.0mV 4.96dB 100kHz 200mV 113.0mV 4.96dB 110kHz 200mV 111.OmV 5.lldB 120kHz 200mV HO.OmV 5.19dB 130kHz 200mV 107.5mV 5.39dB 140kHz 200mV 105.5mV 5.56dB 150kHz 200mV 103.5mV 5.72dB 160kHz 200mV 101.OmV 5.93dB 170kHz 200mV 99.OmV 6. 1ldB 180kHz 200mV 97.5mV 6.24dB 190kHz 200mV 95.OmV 6.47dB 200kHz 200mV 92.5mV 6.70dB - 121 -Table A.6 Measured Noise Voltage in LOW, MID, and HIGH Frequency Band in an Industrial Building Phase A Phase B Phase C Remote B~ Input (across 12 ohm, +5%) Rms +2mV Peak LOW BPF output Rms +20mV Peak MID BPF output Rms + 10mV Peak HIGH BPF output Rms +2mV Peak 60mV 1.0V 280mV 3.2V 160mV 2.4V 42mV 0.58 V 58mV 0.7V 200mV 2.0V 200mV 1.6V 37mV 0.60 V 62mV 0.7V 200mV 2.0V 165mV 1.7V 43mV 0.80 V 57mV 0.9V 290mV 3.0V 175mV 2.6V 32mV 1.0V Remote B is an outlet used transmission as defined in for remote reception and section Ill.ii. - 122 -Table A.7 Measured Noise Voltage in LOW, MID, and HIGH Frequency Band in a Residential House Phase 0e Phase 180« Input (across 12 ohm, ±57.) Rms +0.5mV Peak LOW BPF output Rms +2mV Peak MID BPF output Rms +2mV Peak HIGH BPF output 11.5mV 160mV 39mV 0.6V 47mV 0.35 V 13.5mV 220mV 55mV 0.6V 38mV 0.45V Rms +0.5mV Peak 7.5mV 0.9V 12.OmV 0.75 V - 123 -APPENDIX B. SCHEMATICS OF CMFSK TRANSMITTER AND RECEIVER The detailed block diagrams o-f the transmitter and receiver showing all the sub-systems are shown in Fig. B.l and Fig. B.2. The schematics o-f the transmitter is shown in Fig. B.3. The schematics o-f the receiver is broken up into three parts and are shown in Fig. B.4, Fig. B.5 and Fig. B.6 respectively. Fig. B.4 contains essentially the analog circuits of the CMFSK receiver while Fig. B.5 and Fig. B.6 contain mostly digital circuits. In Fig. B.6, the internal DATA and CLK signals are bufferred Cby CD4050] to provide level—compatible signals for the HP1645A data error rate analyzer in BER measurements. In both the transmitter and receiver circuits the power supply section is omitted because of its simple implementation with a 16V/0.5A power supply transformer, filter capacitors, and a voltage regulator IC. CLoCK W77A I-24-32 MHz 19.2 KHz /W Genera-boy* C-D4-Q94-FSK Madu la-tor ^ Pwr Amp 5» Ouiput Cou^lin^ pPC2.oo2 1> I IT VAC o tyd^iie fake Tra.psfor}»er> 11Z2100 Fig- B.1 Detailed Transmitter Block Diagram A A 117 VAC Input Cou_j>l!ry^ ^ ftU Order BPF FSK Demodulator PLL Sprayue false Hans former 11Z2.LO0 XR-5532. raw data "3 £d$e. Detection. CD 4-olT DATA Threshold D&cfsion O DATA CLK CD4-OI3 CD Sit- Clack Recovery PLL 3O2-7KWZ CD4S2Q Pulse Stteicker CD+538 I9.2KHZ. Fig. B.2 Detailed Receiver Block Diagram 20 21 22. 23 +I5v ~ M32MHz tHF_ •280 ecu CD409+ DATA Ql Qz <33 (4- 13 9 a io /a •xR-2BOe> 5 I 3 1 r 7 8 50KK2. I OK Fig. B.3 Schematics of Transmitter \AM-2.42 2.42K ltzztoo + 7.5v-j- -1>»F IC1 XR-SS3 2. A5K fad ~p -1/JF BITS wv- -vW 470 fF IC2 XR-5532 XC3 XR-OQl Fig. B.4 Schematics of Receiver - Input Coupling, 4th Order BPF, and FSK Demodulator RavJ Bits vW—-VV\A4 16 13 14-10 XR-2212 12. s n ICl XR-OQ2. £2nF-~T~ Fig. B.5 Schematics of Receiver -Pulse Stretcher, and Bit Hard Limiter, Edge Detector, Clock Recovery —<3— DATA O- /S.£fc4p* PAX A 2Za Ext O-19^ KHz CLK 22 o -A/W BER Analyser CD4-094-Qs Ql _t>AT/4_ 56/1:. 56 K 8 C7>^05O 4 2 3 5 Q8 Qi 5a; 56* 3 5 f £ '3 CD4-0/3 9 IO 11 +/Sv OIW NTS —pWv-KE5ET JL -3— 38. f KHz. o ^Fig. B.6 Schematics of Receiver - Digital Integrate/Dump, Threshold Decision, and Data/Data Clock Output 

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
United States 14 0
China 7 3
Germany 2 9
Japan 2 0
Canada 1 0
Brazil 1 0
City Views Downloads
Unknown 12 10
Beijing 6 0
Mountain View 4 0
Ashburn 2 0
Tokyo 2 0
Shenzhen 1 3

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}
Download Stats

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0064794/manifest

Comment

Related Items