Given the widespread adoption of cellular-phones, it can be assumed that the presence of a
phone can predict with good certainty the presence of a human being. Therefore the location
of phones in restricted areas can aid in anti-poaching, anti-smuggling, illegal immigration,
and search-and rescue operations. There are numerous obstacles associated with regulations
and policies which restrict the direct use of the cellular network, therefore the acquisition of
a non-network cooperative (NNC) direction finding (DF) receiver system is required. This
dissertation addresses the development of such a NNC system.
The system requirements for a NNC-DF is analysed to illustrate the design challenges, such
as DF accuracy, dynamic range, inter-channel interference, processing requirements and cost.
Theoretical analysis of different receiver designs, DF estimation algorithms, processing methods,
and sensory input configurations, are done and investigated through simulation. The
simulation results are used to optimise the system parameters in terms of processing time
versus DF accuracy. The optimised results are then used to discuss the design process for an
operational system.
Several Multiple Signal Classification (MuSiC) based algorithms are used for the directionof-
arrival (DOA) estimation, as these algorithms are super-resolution phase interferometry
algorithms. Linear and circular sensor arrays of four to six elements are considered for
the investigation. A selection of receivers which use different levels of signal isolation and
integration methods are used and compared.
The simulation results illustrate that receiver designs with high signal separation have superior
results, but the associated processing requirements make these receivers impractical.
Many of the simpler receiver architectures achieved competitive DF accuracy, and required
only a fraction of the processing resources. Exploiting the resource block (RB) structure of
Long-Term Evolution (LTE), the 12-carriers per RB can be combined to improve DF estimation.
It was found that integration of the autocovariance matrix (ACM) of 12 carriers in a
RB (MuSiC based algorithms require the ACM for estimation) yields the best results. The
Root-MuSiC algorithm resulted in the optimal performance versus processing time for linear
arrays, and the frequency-domain Root-MuSiC algorithm for circular arrays. Advanced forms
of the MuSiC algorithm, which use the weighted least squares (WLS) algorithm, required additional
processing, but results did not improve significantly. It was also found that the design
of the receiver had a greater influence on the performance than the DF algorithms.
Optimisation was done so as to find the best combination of the following:
receiver design,
integration method,
windowing method,
DF algorithm,
antenna configuration and
antenna size.
The optimisation compared the processing time to DF accuracy of the different DF systems.
It was shown that for uniform circular arrays (UCAs), simple receiver architectures with
ACM integration over a RB, using a rectangular window and the frequency-domain (FD)-
Root-MuSiC algorithm, yielded the best processing time versus DF accuracy. Similar results
were found with uniform linear arrays (ULAs), with the exception that the Root-MuSiC
algorithm performed better. Optimisation proved efficient DF receiver design.
It was concluded that the best possible DF accuracy often requires an impractical system.
Similarly, arbitrary large arrays yield excellent results, but are expensive and impractical for
mobile applications. Through optimisation of the simulation results the development of a
realisable system with the best possible performance is possible.
Gegewe die wydverspreide gebruik van sellulêre fone, kan dit aanvaar word dat die teenwoordigheid
van n selfoon met sterk sekerheid die teenwoordigheid van n mens kan voorspel.
Gevolglik kan die bepaalde ligging van n selfoon in n verdagte gebied help om bv. stropery,
smokkel, en onwettige immigrasie teen te werk, asook help met soek- en reddingsoperasies.
Wette en regulasies bied struikelblokke wat die direkte gebruik van n sellulêre netwerk beperk
om die ligging van n sellulêre foon te bepaal. Daarom is dit noodsaaklik om n nie-netwerk
koherente (NNK) rigtingpeiling (RP) - stelsel te gebruik om die ligging van sellulêre fone te
bepaal. Hierdie verhandeling spreek die ontwerp van so n NNK stelsel aan.
Die stelselvereistes vir n NNK RP-stelsel word ontleed, terwyl die geassosieerde ontwerpsuitdagings,
soos akkuraatheid, dinamiese bereik, inter-kanaal steuring, verwerkingsvereistes en
koste in ag geneem word. Die teoretiese analise van verskillende ontvangerontwerpe, rigtingpeiling
algoritmes, en verwerking metodes is gedoen en bevestig deur simulasie. Die simulasieresultate
word gebruik om die stelselparameters te optimiseer in terme van verwerkings-tyd
teenoor rigtingpeilings akkuraatheid. Die resultate wat as optimaal beskou word, word bespreek
in terme van die ontwerp van n operasionele stelsel.
Verskeie MuSiC-gebaseerde algoritmes word gebruik vir die rigting-van-aankoms estimasie,
omdat hierdie algoritmes super-resolusie fase-interferometriese tegnieke is. Lineêre en
sirkulêre sensorsamestellings met vier tot ses elemente word ondersoek in hierdie studie.
n Verskeidenheid van ontvangers met verskillende seinisolasietegnieke en integrasiemetodes
word gebruik en vergelyk.
Die simulasieresultate wys daarop dat die ontvangerontwerpe, met hoë seinisolasietegnieke,
bostaande resultate toon, maar dat die verwerkingsvereistes van dié tegnieke onrealisties
hoog is. Eenvoudiger ontvangerontwerpe bereik kompeterende rigtingpeilings akkuraathede,
maar vereis slegs n fraksie van die verwerkingshulpbronne. Deur die ontginning van die
verwerkingsblok (VB) van Langtermyn Evolusie (LTE), kan die 12 draers per VB gekombineer
word om die rigtingpeiling te verbeter. Daar is bevind dat die integrasie van die
kovariansiematriks van elk van die 12 draers per VB (MuSiC gebaseerde algoritmes benodig
n kovariansiematriks vir estimasie) die beste resultate lewer. Die Wortel-MuSiC algoritme
het die beste akkuraatheid teenoor verwerkingstyd vir lineêre samestellings gelewer, en die
frekwensie-domein (FD)Wortel-MuSiC algoritme die beste resulate vir sirkulêre samestellings
gebeur. Gevorderde weergawes van die MuSiC-algoritme, wat die geweegde kleinste kwadrate
algoritme gebruik, vereis addisionele verwerking, maar lewer resultate wat nie beduidend
beter is nie. Dit is ook bevind dat die ontwerp van die ontvanger n groter invloed op die
akkuraatheid van die estimasie het as wat die rigtingpeiling algoritmes bied.
Optimisering is gedoen om die beste kombinasie van die volgende faktore te vind:
ontvanger ontwerp,
integrasie metode,
venstermetode,
rigtingpeilings algoritme,
antennasamestelling-opstelling en
antennasamestelling-grootte.
Die optimiseringsproses vergelyk die verwerkingstyd en rigtingpeilings akkuraatheid van
die verskillende stelselparameters. Daar is getoon dat eenvormige sirkulêre samestellings
met eenvoudige ontvanger-argitektuur en kovariansiematriks integrasie oor n VB, met die
FD-Wortel-MuSiC-algoritme, die beste verwerkingstyd teenoor rigtingpeilings akkuraatheid
lewer. Soortgelyke resultate is gevind met eenvormige lineêre samestellings, met die uitsondering
dat die Wortel-MuSiC algoritme. Die gebruik van optimiseringstegnieke lei tot
doeltreffende rigtingpeilingontvanger ontwerp.
Die mees gewensde rigtingpeilings akkuraatheid vereis dikwels n onwerkbare stelsel. Verder
lewer arbitrêre groot samestellings uitstekende resultate, maar is duur en oneffektief vir mobiele
radio toepassings. Deur optimisering van die simulasieresultate kan werkbare stelsels
ontwikkel word met die beste moontlike werkverigting.
