Search

In this doctoral dissertation, the diversity techniques are considered in order to reduce the influence of fading on the system performances in different propagation environments. The second chapter considers the selective combining (SC) and the maximal ratio combining (MRC) with two and three inputs. SC and MRC receivers with two inputs are considered when Nakagami-m fading is at the first input while α-μ fading is at the second input. Then, an SC receiver with three inputs is considered when Nakagamim, Rice and Rayleigh fading are presented. As a special case, SC and MRC receivers with three inputs with Rice, Nakagami-m and α-μ fading are considered. The SC performance of the two-input receivers is considered in the case of the Nakagami-m fading and Nakagami-m interference at one input, while at the second input are the Rice fading and Rice interference, as well as the case when a desired Rice signal and Nakagami-m interference are presented. For these models the probability density function (PDF), cumulative density function (CDF), moments, level crossing rate (LCR) and the information capacity of the channel at the output from the SC and MRC receivers are calculated, the numerical results are graphically presented and confirmed by simulations. In the third chapter, the significant contribution of the doctoral dissertation is the formation of the broken Nakagami-m distribution and the analysis of the broken Nakagami-m fading. An SC receiver with two and three inputs with broken Nakagami-m fading is examined. The dissertation also examines the case of the SC two-way receivers where there is a broken Rice/Nakagami-m fading. For such cases, expressions for PDF, CDF and LCR are derived. In the fourth chapter, the macrodiversity system is analyzed. The macrodiversity SC receiver with two micro-diversity SC receivers, when α-μ fading is present at the inputs and the correlated Gamma shadowing, as well as the case when Rayleigh, Nakagami-m fading and the correlated Gamma shadowing are at the inputs are examined. The case of macrodiversity SC receivers with three microdiversity SC receivers is further considered when Rice, Nakagami-m, κ-μ fading and correlated Gamma shadowing are presented at inputs. For these models, the PDF, CDF, LCR and the information capacity of the channels at the macro-diversity SC receiver output are calculated and numerical results are graphically presented.

Za telekomunikacione sisteme unutar zgrada, koncept “bežičnog je uveo “novi” prenosni medijum, odnosno kanal za prostiranje radio talasa u zatvorenim prostorima, koji se koristi za povezivanje terminala i ostale opreme u lokalnu računarsku mrežu (LAN). Eksperimentalne studije, urađene u cilju opisivanja kanala za prostiranje talasa u zatvorenim prostorima, su pokazale da su karakteristike prostiranja talasa u zatvorenim prostorima: velika slabljenja pri prostiranju, prostiranje po više puteva između predaje i prijema, spore vremenske promene kanala I feding. Neke primene bežičnih kanala u zatvorenim prostorima zahtevaju veoma brz prenos podataka koji je ograničen prostiranjem po više puteva I fedingom. Kad se poveća brzina prenosa podataka, prostiranje po više puteva prouzrokuje interferenciju među simbolima u toku procesa detekcije simbola. Uobičajeni način borbe protib fedinga koji nastaje pri prostiranju po više puteva, a kao posledica toga i interferencije među simbolima, je upotreba adaptivne ekvalizacije na prijemu. Problem osetljivosti digitalnog radio telekomunikacionog sistema u zatvorenim prostorima na prostiranje po više puteva pri promenjivim parametrima kanala i sitema, je obrađen u ovoj tezi, na primeru 4-QAM telekomunikacionog sistema, sa LMS linearnim ekvalizatorom sa kašnjenjem od jednog simbola između susednih koeficijenata kao i sa LMS nelienarnim (DF) ekvalizatorom sa kašnjenjem manjim od jednog simbola između susednih koeficijenata na prijemu. Rešavanju problema je pristupljeno iz 3 pravca. Prvo je data teorijska analiza problema. Metodom simulacija na računaru je izvršena detaljna procena performansi sistema za veoma brzi prenos podataka po disperzivnim kanalima sa sporim fedingom, koriteći statistički eksponencijalni model kanala sa Relejevim fedingom koji je predložen u jednoj studiji, kao i Rajsov model kanala sa fedingom koji se bazira na merenjima izvršenim u radnim prostorijama i fabričkom okruženju. Konačno. Impulsni odzivi kanala izmereni u naučno-istraživačkoj laboratoriji pri različitim uslovima prostiranja, topografijama (konfiguracijama) i frekvencijama, su takođe korišćeni za procenu performansi sistema. Prosečna verovatnoća greške po bitu je korišćena kao performansni kriterijum. Krive prosečne verovatnoće greške po bitu ukazuju da, bilo kod kanala generisanih uz pomoć računara ili kod izmerenih kanala, linearni ekvalizator smanjuje grešku približno 10 puta (odnos signal-šum od 8dB) u poređenju sa neekvalizovanim kanalom, a da DF ekvalizator nadmašuje linearni ekvalizator dalje smanjujući grešku 2 puta (SNR=8dB). Zapravo, počevši od slučaja frekvencijski-ravnog fedinga, ustanovljeno je da ekvalizator može da smanji grešku za kanale generisane uz pomoć računara skoro do normalizovane vrednosti brzine prenosa podataka od 4 (Relejev slučaj), I, 2.2, 1 I 0.5 (Rajsov slučaj) I vrednost Rajsovog faktora od 2, 6.8 I 11dB, respektivno, dok su u slučaju izmerenih kanala vrednosti minimuma krivih verovatnoće greške između 2 i 3, a pozicije minimuma zavise od veličine prostorije i topografije. Posle minimuma, performanse sistema degradiraju kad se bilo brzina prenosa ili disperzija kanala povećaju, zbog negativnog efekta interferencije među sibolima koji potire poboljšanje usled implicitne (vremenske) raznovrsnosti. Kad je odnos signal-šum nizak (do 12dB zakjlučno), primećeno je da se javlja efekat prostiranja greške. Kod DF prijemnika. Ipak, pokazano je da umetanjem nove, “modifikovane” automatske funkcije čišćenja u filtar povratne sprege kod DF ekvalizatora, se značajno može smanjiti neželjen uticaj efekta prostiranja greške, a samim tim i prosečna verovatnoća greške po bitu i do 2 reda veličine. Na kraju, performanse ekvalizatora su upoređene i sa teorijskom performansom prilagođenog filtra za kanale sa Relejevim fedingom, koji predstavljaju performansnu granicu za sve tehnike eliminacije interferencije među simbolima. Ustanovljeno je da DF ekvalizator sa tačnim povratnim simbolima, koji ima bolje performanse od bilo linearnog ili DF ekvalizatora sa detektovanim povratnim simbolima, daje prosečnu verovatnoću greške po bitu samo oko dva puta veću od “optimalnog” prijemnika čije performanse predstavljaju teorijsku granicu najvećeg mogućeg poboljšanja koje se može dobiti korišćenjem mnogo komplikovanijih prijemnika sa detektorima koji odlučuju po principu funkcije najvećeg izgleda. The concept of wireless in-building systems has introduced a “new” medium the indoor radio propagation channel, which can be used for interconnecting terminals and computers in LAN-s. Experimental studies have shown have shown that indoor radiowave propagation can be characterized by large propagation losses, multipath, slow time variations and fading. Some computer applications require very high signaling rates. When the data rate is increased, themultipath propagation in the indoor channel causes intersymbol interference in the detection process, thus adversely affecting the system performance. A common approach to combat multipath fading and hence intersymbol intereference, is to use an adaptive equalizer. This thesis considers a 4-QAM communication system with LMS symbol-spaced linear and LMS fractionally-spaced decifion feedback adaptive equalizer at the receiver. The problem is approached in three ways. The system performance evaluation for high-speed signaling over slowly fading dispersive diversity channels is considered first, using the statistical exponential Rayleigh fading channel model, and the Rician fading channel model. These models are based on propagation measurements performed in office and factory environments. Next, the channel impulse responses measured in the research laboratory building for different propagation environments, topographies and frequencies are used for system performance predictions based on computer simulation. Finally, a theoretical performance analysis of the problem is given. The average bit error rate (BER) is used as the system performance quality measure. The average BER curves indicate, for both computer generated and measured channels, that a linear equalizer reduces the BER by approximately a factor of ten at SNR=8dB compared with unequalized channel case. The decision-feedback equalizer surpasses the linear equalizer by further reducing the BER by approximately a factor of two at SNR=8dB. Using the frequency-flat fading case as a reference the decision-feedback equalizer is shown to be able to reduce the BER almost up to the normalized data rate of 4 for the case of Rayleigh fading, and up to 2.2, 1 and 0.5 for Rician fading with Rice factor values of 2, 6.8 and 11dB, respectively. The error performance curves for the measured channels have their best values between the normalized data rates of 2 and 3, with the actual minimum positions depending on the room size and lay-out. Beyond the BER minima, system performance degrades when either the data rate or the multipath spread is increased. This is ue to the negative effect of intersymbol interference, which offsets the time divesity gains. Low signal-to-noise ratio (SNR ≤ 12dB) are observed to cause error propagation effects in the decision-feedback receiver. Insertion of a novel modified automatic “clear” function in the backward filter of the decision-feedback equalizer significantly reduces the error propagation effect, decreasing the BER by up to two orders of magnitude. Finally, equalizer performance is compared with the theoretical matched filter performance for Rayleign fading diversity channels; this represents a lower bound for all intersymbol interference elimination techniques. The decision-feedback equalizer with correct symbols fed back, being better in performance than either a linear equalizer or the decision-feedback equalizer with detected symbols fed back, is shown to perform worse than the ideal diversity combining receiver by a factor of approximately two, which represents the theoretical limit for possible performance improvement, using much more complex receivers with maximum likelihood detectors.