Advanced global navigation satellite system receiver design (797918), страница 15

Просмтор этого файла доступен только зарегистрированным пользователям. Но у нас супер быстрая регистрация: достаточно только электронной почты!

Текст из файла (страница 15)

Clearly the SSB technique requiresconsiderably more receiver architecture than the SCC technique. This is due thenecessity to isolate individual BOC sidebands. If analogue filters are implementedwith the SSB technique a multiplexed or additional ADC stage will also be required.Using digital filters imposes considerable demands on the receiver’s correlatorresources.Isolating individual sidebands provides no performance improvement and requiresadditional hardware requirements when compared to the SCC technique.

The SSBtechnique is also incompatible with BOC tracking schemes providing precise timinglocation (detailed discussion in the following section). Therefore, SSB is only afeasible technique for low performance receivers with small front-end bandwidths,receiving only a single BOC sideband.The SSC search technique provides correlator architecture compatible with preciseBOC tracking schemes including the double-estimation BOC tracking loop (seechapter 6) developed during this research.

Demonstrations of receivers using the SCCsearch technique are given in chapter 8 and chapter 10.98Receiver theoryTable 5-4, Hardware requirements of BOC search techniquesReceiver typeMultipliersIntegratorsLocal oscillatorsLow pass filtersPSK2×Carrier,21×Carrier,02×CodeSSC1×Code2×Carrier,41×Carrier,4×CodeSSB1×Code4×Carrier,42×Carrier,4×Code5.4041×CodeTracking BOC signalsThe standard BOC early minus late discriminator curve contains multiple zerocrossings only one of which corresponds to the correct timing location (Figure 5-20).A GNSS BOC receiver aims to achieve and maintain lock on the incoming signal atthe correct timing location.

In order to do so the receiver must action an algorithm ormitigation technique to achieve valid timing information, effectively removing theinfluence of the secondary peaks in the BOC correlation. Ideally, the receiver willachieve this with no loss of tracking sensitivity.2Discriminator error (chips)21ΛBF ( R , 0.0)10505101−22− 12R12Code error (sub-chips)Figure 5-20, BOC(6×fC, fC) discriminator curveHere we consider three standard approaches for solving the ambiguity in trackingBOC signals. Single sideband (SSB) tracking, multiple gate delay (MGD)99Receiver theorydiscriminators and the bump-jumping algorithm (BJ). Firstly we will compare therelative performance of each approach and then assess the impact of each to thereceiver hardware.5.4.1BOC tracking using a single sidebandThe first approach is to use the SSB technique described in the previous section.Treating each BOC sideband as a separate PSK signal creates an unambiguous PSKcorrelation peak ‘Λ’.

Then, correlating early and late replica signals (Figure 5-21) anunambiguous discriminator curve ‘ VΛ ’ can be formed using the standard PSKdiscriminators (Table 5-3).wIQa~ (t − τˆ )wIIuBOC(t)a(t − τˆ )cos(2π×(fC + fS)×t)sin(2π×(fC + fS)×t)wQIa~ (t − τˆ )wQQFigure 5-21, Single-sideband BOC trackingThis approach provides a robust solution. However, the receivers r.m.s. timing jitter isnow dependant on the underlying PSK chipping rate and not the subcarrier rate.

Thisdegrades the receivers timing sensitivity by a factor of 4 f S f C [Bello and Fante2005]. Therefore, this approach is only suitable for low precision receivers.100Receiver theory5.4.2BOC tracking with multiple gate discriminatorsThe second approach is to synthesise an unambiguous discriminator curve by using acombination of multiple correlator channels. This approach, first proposed in [Fante2004] and commonly termed the Multiple Gate Delay (MGD) discriminator.

Thistechnique proposes the use of K early and K late signals forming the followingcorrelations. w (IpE )W =  ( E ) w QpT()w (IpL )  1 TbtcosE=u BOC (t )ω 0 t + φˆ  dtw (QpL )  T ∫0sinb L (t )()5-31Where1b E (t ) = b  t − τˆ+  p −  × TDC21b L (t ) = b  t − τˆ−  p −  × TDC25-32TDC is the early to late spacing and p is a integer count from 1 to K.

The resultingcorrelations can be written as follows.()1τˆ − τ −  p −  × TDC  × d2()1τˆ − τ +  p −  × TDC  × d2()1τˆ − τ −  p −  × TDC  × d2()1τˆ − τ +  p −  × TDC  × d2w (IpE ) ≈ A × cos φˆ − φ ×w (IpL ) ≈ A × cos φˆ − φ ×(E )w Qp≈ A × sin φˆ − φ ×(L)w Qp≈ A × sin φˆ − φ ×5-33A vector of coefficients can then used to weight the influence of each of early and latecorrelations in an attempt to shape an unambiguous BOC discriminator. The101Receiver theorycomposite MGD discriminator is formed using of the non-coherent early-late powerdiscriminator whose error function can be written asK((E )( L)eτ (τ ) = ∑ c p × w (IpE ) + w Qp− w Qp− w (QpL )2222)5-34p =1eτ (τ ) ≈ ∑ c p × p =1K2 1τ −  p − 2  × TDC  −  1τ +  p − 2  × TDC  2where cp is the coefficient vector used to form the shape of the discriminator.Different combinations of coefficients are compared in [Fante 2004] to determine thebest possible discriminator synthesis.

Two classes of synthesised discriminators aredefined, smooth and bumpy. smooth discriminators synthesise a discriminatorapproaching a monotonic error function, which provides a single shallow zerocrossing. bumpy discriminators synthesise a single steep zero crossing with manyundulations across the discriminator characteristic.

Assuming a BOC(2×fC, fC) signaland using four early-late discriminator combinations (K =4) the use of coefficients [11.25 1.5 1.75] and TDC = 0.525 × TS creates a smooth discriminator, the use ofcoefficients [1 1.125 1.25 1.375] and TDC = 0.2 × TS creates a bumpy discriminator.The composite smooth and bumpy discriminator curves are shown in Figure 5-22.Clearly the smooth discriminator more closely reflects the shape of a PSK ‘VΛ’discriminator providing robust acquisition of the correct timing location. The bumpydiscriminator has a number of nulls in the envelope which will slow down theacquisition process and potentially cause false-lock states. It is argued that thermalnoise will prevent false-lock occurring as the code error polarity is true either side ofeach null. Also, it is feasible to envisage a scheme which may acquire with a smoothenvelope and transition to the bumpy discriminator for precise timing location.102Receiver theory1K = 4 SmoothK = 4 BumpyDiscriminator error (chips)0.9490.5ΛSM ( R , 0)ΛBMP ( R , 0)64202460.5− 0.9611−6R6Code error (sub-chips)Figure 5-22, Composite MGD discriminators for a BOC(2×fC, fC) signalAssuming lock at the correct location the bumpy discriminator can deliverconsiderably less timing jitter than the smooth discriminator due to its steep zerocrossing.

In [Bello and Fante 2005] the r.m.s. timing jitter of these compositediscriminators are compared to that given by the conventional BOC discriminator(Equation 4–20) and the SSB technique. Considering a BOC(2×fC, fC) signal, thepaper concludes that the SSB technique degrades the receivers timing sensitivity by afactor of 8 compared to the conventional BOC discriminator.

This is because thetiming jitter is now proportional to the discriminator formed from the underlying PSKmodulation which is less sensitive by a factor of 4 f S f C . The bumpy (K = 4)discriminator shown in Figure 5-22 is considered to be the best choice of compositediscriminator for precise timing, as the timing jitter is only slightly (0.14 dB) worsethan the conventional BOC discriminator. We have verified this result by inspectionof the slope of the resulting bumpy discriminator curve whose zero crossing isshallower by a factor of 1.017, which equates to 0.146 dB worse timing jitter.The bumpy discriminator suffers a slow response time for acquisition and large errorsteps due to the nulls present in its discriminator envelope.

An example acquisition ofthe bumpy and smooth discriminator is shown in Figure 5-23 (derived from Mathcad103Receiver theorysimulation). The loop bandwidth and noise conditions are set equal for both cases, BL= 1 Hz, C/N0 = 30 dB-Hz. The response of the bumpy discriminator flattens at subchip intervals as it passes a null in the discriminator significantly lengthening the loopsettling time. Also, the jitter of the bumpy discriminator can be seen to beconsiderably less than that of the smooth discriminator.3.5K = 4 SmoothK = 4 Bumpy3.1913Timing error (sub-chips)2.52tT k−tRC ktT1 k− tRC1 k1.510.5050100150200250300350400450500− 0.20.50k500Loop iterationsFigure 5-23, Acquisition example of MGD discriminators (BL = 1 Hz , C/N0 = 30 dB-Hz)Running multiple acquisitions across different initial time offsets provides acomparison of the acquisition time of the MGD discriminators with the equivalentacquisition times using the SSB technique.Table 5-5 shows the acquisition times of the MGD discriminators and the SSBtechnique, assuming a BOC(2, 1) signal, a carrier to noise density of C/N0 = 24 dBHz, a loop bandwidth of BL = 1 Hz and averaging across 20 acquisitions at each timestep.

We find agreement with the results presented in [Bello and Fante 2005] whichcompare the average loop settling time of the MGD discriminators. The smoothdiscriminator closely follows that of the equivalent SSB discriminator characteristic,with an acquisition performance only 1.23 times worse than the SSB technique. Thebumpy MGD discriminator is worse by at least factor of 2 across the dataset.Therefore, composite MGD discriminators can potentially remove the ambiguity ofBOC tracking.

However, the receiver designer must make a trade-off between thereceiver’s response time and timing jitter.104Receiver theoryTable 5-5, Acquisition times of the MGD discriminators and the SSB technique for BOC(2, 1),BL = 1 Hz, C/N0 = 24 dB-Hz5.4.3Initial chip offset1/41/23/4Smooth MGD acquisition time (ms)73010511623Bumpy MGD acquisition time (ms)90619193065SSB acquisition time (ms)6118891216SSB / Smooth1.191.181.33SSB / Bumpy1.482.162.52The bump-jumping algorithmThe final state of the art approach we consider in solving the BOC ambiguity problemis the ‘bump-jumping’ algorithm (BJ), proposed by in [Fine and Wilson 1999]. Thisalgorithm determines whether or not the correct correlation peak is being tracked bycomparing the amplitude of the peak currently being tracked to the amplitude of theadjacent peaks.

Характеристики

Список файлов учебной работы