Hardware limitations of interference suppressing beamforming

Thesis (MEng)--Stellenbosch University, 2019.ENGLISH ABSTRACT: This thesis proposes two blind beamforming techniques for use in global positioning system (GPS) applications as well as a dual band multi-layered micro strip antenna for use at L1 and L2 GPS frequencies. Various direction of arrival (DOA) algorithms were tested to investigate their failure points. The antenna was built and measured. It was found that S11 is below -10 dB across the L1 and L2 bands from 1.22 GHz - 1.57 GHz. The gain is 1.7 dB below the simulation value in the L1 band however the L2 bands gain matches the simulation value of 7 dB. The results show that the antenna will function for GPS applications. The algorithms were to be tested, in practice, using the transient array radio telescope (TART) system which is a 24 element radio telescope designed for testing of imaging algorithms and surveying transient events. The system makes use of the MAX2769B universal GPS receiver which produces binary samples at a sample frequency of 16.368 MHz and a centre frequency of 4.092 MHz. Algorithm 1 uses an orthogonal subspace beamforming method to null interference and place antenna gain in the directions of the signals of interest (SOI). Algorithm 2 uses linearly constraint minimum variance (LCMV) beamforming and a beam search method in order to find available GPS satellites and null interference. Both algorithms were tested thoroughly by varying parameters such as signal to noise ratio (SNR), signal to interference ratio (SIR), integration time and the number of analogue to digital converter (ADC) bits to find their failure points and determine if they are viable techniques on the TART system. It was found that, when using a peak search method to determine success rates, and sweeping SNR, SIR, integration time (number of samples) and the number of ADC bits, the Bartlett DOA algorithm performs the best, possibly due to the number of signals present on the simulated array. It is shown that the success rates of algorithms 1 and 2 are similar, for the parameters that were swept. According to simulations, both methods are viable beamformers, for GPS applications, on the TART system and are not limited by its hardware. Algorithm 1 is more sensitive to variations in SIR, SNR and integration time when using 1 bit shifted data. The complex bit shifted data is created using a technique that allows one to create quadrature (Q) data by shifting the in-phase (I) data 1 bit, and is only possible due to the intermediate frequency (IF) being one quarter that of the sample frequency. Algorithm 1 also requires a priori information about the number of SOI's which is not required for algorithm 2. Algorithm 2 is more computationally expensive compared to algorithm 1 for GPS applications. Preliminary practical results were collected using the TART system, and analysed. The measurements taken show the potential for algorithm 1 to be implemented, as a beamformer, in the TART system, however thorough practical testing was not possible due to the TART system becoming unusable.Hierdie tesis stel twee blinde bundelvormingstegnieke voor vir die gebruik in GPS toepassings, sowel as 'n dubbelband mikrostrook antenna vir gebruik by L1 en L2 GPS frekwensies. Verskeie invalshoek afskattings algoritmes (DOA) is getoets om hulle valingspunte te ondersoek. Die antenna was gebou en gemeet. S11 is onder -10 dB oor beide die L1 en L2 frekwensiebande van 1.22 GHz - 1.57 GHz. Die aanwins is 1.7 dB onder die gesimuleerde waarde in die L1 band, maar die L2 band aanwins stem ooreen met die gesimuleerde waarde van 7 dB. Die resultate wys dat die antenna sal funksioneer vir GPS toepassings. Die algoritmes moes prakties getoets word deur van die TART teleskoop stelsel gebruik te maak. Dit is 'n 24 element radio teleskoop ontwerp vir die toets van beeldingsalgoritmes en soektogte na kort tyd gebeurtenisse. Die stelsel maak gebruik van die MAX2769B universele GPS ontvanger wat binêre monsters teen 'n monsterfrekwensie van 16.368 MHz by 'n senterfrekwensie van 4.092 MHz lewer. Algoritme 1 gebruik 'n ortogonale ruimte bundelvormingstegniek om 'n nul te stuur in die rigting van 'n sterk steursein en ter selfde tyd antenna aanwins te plaas in die rigtings van belangrike seine. Algoritme 2 gebruik lineêr beperkte minimum variansie (LCMV) bundelvorming en 'n bundel soektog metode om beskikbare GPS seine te vind en steurseine te kanselleer. Beide algoritmes is sorgvuldig getoets deur parameters soos sein tot reis verhouding (SNR), sein tot steursein verhouding (SIR), integrasietyd, en die aantal analoog na digitale (ADC) bisse te verstel om te valingspunte te vind en sodoende vas te stel of dit 'n bruikbare algoritme is om te gebruik op die TART stelsel. Dit is gevind dat, deur 'n piek soektog algoritme te gebruik om sukseskoers te bepaal, en SNR, SIR, integrasietyd en aantal ADC bisse te verstel, dat die Bartlett DOA algoritme die beste vaar, moontlik as gevolg van die aantal seine teenwoordig in die gesimuleerde samestelling. Dit is getoon dat die sukseskoers van algoritmes 1 en 2 soortgelyk is, vir die parameters wat verstel is. Volgens die simulasies is albei metodes werkbare bundelvormers, vir GPS toepassings, op die TART stelsel en word nie beperk deur die stelsel hardeware nie. Algoritme 1 is meer sensitief vir variasies in SIR, SNR en integrasie tyd wanneer 1 bis geskuifde data gebruik word. Die kompleks bis geskuifde data is geskep deur van 'n tegniek gebruik te maak wat mens toelaat om 'n kwadratuur sein (Q) op te wek deur die in fase sein (I) met 1 bis te skuif. Dit is slegs moontlik as die monsterfrekwensie presies 4 maal die tussenfrekwensie (IF) is. Algoritme 1 benodig ook vooraf kennis van die aantal SOI's, wat nie nodig is vir algoritme 2 nie. Algoritme 2 is meer bewerkingsintensief in vergelyking met algoritme 1 vir GPS toepassings. Voorlopige praktiese resultate was met die TART stelsel bekom en geanaliseer. Die metings toon potensiaal vir algoritme 1 om geïmplementeer te word, as bundelvormer, vir die TART stelsel, maar volledige toetse kon nie voltooi word nie omdat die stelsel onklaar geraak het