Predmet ove doktorske disertacije je kalibracija distribuiranog
vixekanalnog prijemnog sistema beiqnim putem, tako xto se vrxi obrada primljenih radio signala. Distribuirani prijemnici imaju nezavisne analognodigitalne konvertore i lokalne oscilatore, te su vremenski, frekvencijski i
fazno nesinhronizovani. Pod kalibracijom se podrazumijeva procjena i kompenzacija vremenskih, frekvencijskih i faznih pomaka izmeu signala u prijemnim
kanalima u cilju postizanja vremenske, frekvencijske i fazne sinhronizacije, koja
je preduslov za pravilno funkcionisanje obrade signala sa antenskih nizova.
Pretpostavlja se da postoji prostorna koherencija signala, xto znaqi da se
faza nosioca predvidivo mijenja po prostoru tako xto je linearna funkcija vremena propagacije talasa i frekvencije nosioca. Pored toga, smatra se da se snaga
reflektovanih komponenti signala (Non-Line-of-Sight { NLoS) moe zanemariti u
odnosu na direktnu komponentu (Line-of-Sight { LoS). Ove pretpostvake vae pri
prenosu u mmWav...e opsegu u malim elijama, xto je karakteristiqno za nadolazeu petu generaciju mobilne telefonije, gdje se oqekuje primjena rezultata
istraivanja u tezi.
Pretpostavlja se da se predajnici i prijemnici ne kreu. Prijemnici su povezani sa fuzionim centrom putem digitalnih linkova, posredstvom kojih nije
mogu prenos analognih referentnih signala za sinhronizaciju. Zbog jednostavnosti, teza se najveim dijelom bavi dvokanalnim prijemnim sistemima, a proxirenje na vixekanalne sisteme moe se postii posmatranjem parova kanala.
Analizirana su dva scenarija.
U prvom scenariju, izmeu prijemnih kanala postoje konstantni vremenski pomak, poqetni fazni pomak i promjenljivi frekvencijski pomak. Cilj je razvoj i
ispitivanje performansi metoda koje, procjenom i kompenzacijom pomaka izmeu
primljenih signala, omoguuju distribuirani digitalni beamforming (BF) i lokalizaciju izvora radio signala korixenjem datog sistema. Sa tom namjerom je
u tezi predloena procedura za kalibraciju. U sluqaju lokalizacije, procedura
obuhvata zdruenu procjenu i kompenzaciju vremenskog i trenutnog faznog pomaka
izmeu prijemnih kanala. U tu svrhu se koristi predajnik za kalibraciju, beacon,
koji xalje uskopojasni i xirokopojasni pilot. U sluqaju beamforming-a, procedura
obuhvata zdruenu procjenu i kompenzaciju vremenskog i trenutnog faznog pomaka
izmeu primljenih korisnih signala i, opciono, ekvalizaciju korisnog signala.
Pored uskopojasnog i xirokopojasnog pilota koje xalje beacon, koriste se i uskopojasni pilot (ako se vrxi ekvalizacija) i xirokopojasna preambula, poslati
od strane korisniqkog predajnika. Poxto je frekvencijski pomak promjenljiv,
ne procjenjuje se eksplicitno, ve je formulisan novi adaptivni algoritam za
procjenu trenutnog faznog pomaka. Nekoherentni algoritam tipa maksimalne vjerodostojnosti (Maximum Likelihood { ML), predstavljen u drugom dijelu teze, vrxi
procjenu vremenskih pomaka. Numeriqki prost algoritam baziran na korelaciji
signala koristi se za procjenu konstantnog faznog pomaka. U svrhu evaluacije
predloene procedure i algoritama vrxene su Monte-Karlo simulacije i eksperimenti sa softverski definisanim radio-ureajima. Rezultati eksperimenata
su pokazali odliqno slaganje sa rezultatima simulacija, xto potvruje ispravnost matematiqkog modela i usvojenih pretpostavki...
The dissertation deals with over-the-air calibration of a distributed multichannel receiving system, by processing the received radio signals. Distributed receivers have
independent analog-digital converters and local oscillators, so they are time, frequency, and
phase unsynchronized. The calibration includes estimation and compensation of time, frequency, and phase offsets between the signals in the receiving channels with the aim of achieving
time, frequency, and phase synchronization, which is a prerequisite for proper operation of array
processing.
Spatial coherence of signals is assumed, meaning that the carrier phase changes predictably
over the space, being the linear function of the propagation delay and the carrier frequency.
Additionally, we assume that the Non-Line-of-Sight components (NLoS) are negligible compared to the Line-of-Sight (LoS) components. These assumptions are valid for communication in
mmWave range in small cells, which are typical of 5G, where the ap...plication of the investigation
in the thesis is expected.
All of the transmitters and receivers are assumed to be stationary. The receivers are connected to a fusion center via digital links, which are incapable of conveying analog reference
signals for the synchronization. For the sake of simplicity, the thesis mostly addresses twochannel receiving systems, but the generalization to multichannel systems can be achieved by
dealing with pairs of the channels. Two scenarios are analyzed.
In the first scenario there is a constant time offset, an initial phase offset and a variable
frequency offset between the receiving channels. The goal is to develop and investigate the
performance of the methods that, by estimating and compensating for the offsets between the
received signals, enable distributed digital beamforming (BF) and the radio source localization
using that system. To this end a procedure for calibration is proposed in the dissertation. In
the localization case, the procedure includes joint estimation and compensation of the time
offset and instantaneous phase offset between the receiving channels. For that purpose a
calibration transmitter, the beacon, sends a narrowband and a wideband pilot. In the BF case,
the procedure includes joint estimation and compensation of the time offset and instantaneous
phase offset between the received signals, and, optionally, equalization of a user signal. For
that purpose a narrowband pilot (if the equalization takes place) and a wideband preamble sent
by the user transmitter are used, in addition to the beacon signals. Since the frequency offset
is time-variable, it is not estimated explicitly, but a new adaptive algorithm for instantaneous
phase offset estimation is formulated. A non-coherent maximum likelihood (ML) algorithm,
presented in the second part of the dissertation, is used for the time shifts estimation. A
numerically cheap algorithm based on signal correlation is used for the constant phase offset
estimation. Monte-Carlo simulations and experiments using software defined radio devices have
been carried out to evaluate the procedure and algorithms. The results of the experiments
have shown excellent matching with the simulation results, which confirms the correctness of
the signal model and adopted assumptions. The adaptive algorithm for instantaneous phase
offset estimation has shown capability of following abrupt changes in frequency offsets. The
obtained accuracies show that the proposed procedure and algorithms are especially suited to
receive BF and non-coherent/semi-coherent localization. If the frequency offset changes slowly
enough, application in transmit BF is also possible. The proposed procedure is modular, i.e.
every algorithm can be replaced by another algorithm of the same type...
