Research On Carrier Recovery Methed In Digital Communication System

IntroductionDigital communication technique plays an important role in moderncommunication systems. With the development of micro-electronics, computer anddigital signal processing technology, modem modulation/demodulation techniquescome true from analogue to all digital and even software implementation, As aresult all-digital receiver emerge as the times require. It is necessary in digitalcommunication system. It also makes carrier recovery, which is one of the keytechnologies of the all-digital receiver develop quickly. Carrier recovery may becalled carrier synchronization. It can compensate frequency offset and phase jitterbrought by transmission and modulation. This paper is mainly focused on carrierrecovery Method in the digital communication systems.1. The development of all-digital receiverCarrier recovery technique develops with the development of the receiver.From analog to all digital the principle and implement of carrier recovery in thereceiver have changed much. In contrast to conventional receiver, all-digitalreceiver does not require the feedback of signals to the analog part, so all-digitalreceiver overcomes difficulties in phase lock loop design. It needs not voltage butaccurate estimated value of the carrier phase and frequency offset .So we mustderive the parameter estimation algorithms. Because of the complexity ofalgorithms, choosing algorithms which can estimate efficiently and implement byDSP easily is a main topic question of all-digital receiver.2. The object discussedIn this paper quadrature amplitude modulation (QAM) is selected as the objectto discuss its carrier recovery technique. Because it is a highly bandwidth efficienttransmission technique and is used in many of these applications which need forever-increasing throughputs through fixed bandwidths such as digital TV. Thisrequires accurate and robust carrier synchronization in the receiver.3. Signal modelIn this paper the signal model is assumed that the system with additiveGaussian white noise(AWGN) channel is already equalized and that both symboltiming and relative gain control have been established. Moreover the receivedsignals with an unknown fixed phase and a frequency offset have been ideallyfiltered and sampled at the optimum sampling instant. The frequency offset isrelatively small, so there isn't intersymbol interference.4. The standard of estimator performance In order to analyze estimation performance and design appropriate systemequipment, it is important to understand the theoretical estimator performancelimits The Cramér–Rao lower bound (CRLB) is frequently employed as a standardthat is compared with the mean-square estimation error(MSE) or variance. Themodified CRLB (MCRLB) is a good approximation for the true bound for QAMsignals at higher signal-to-noise ratio (SNR). In this paper, the true CRLB ispresented for the estimation of phase and frequency offset of the common QAMsignals in AWGN channels.5. The main estimation algorithms of the carrier phase offset(1) Maximum likelihood estimation(ML) Maximum likelihood estimation is used in many of the parameter estimationfields. The complexity of the algorithm is due to its computation complication evenfor small constellations. Moreover it needs to solve a nonlinear maximizationproblem in order to find the results. which makes the ML estimation impractical.Clearly, we must resort to suboptimal implementations. So some simple algorithmssuch as, decision-directed estimation, power-law estimation, histogram algorithm,etc are derived from ML estimation.(2) Decision-directed estimation(DD) The phase synchronization problem is often divided into an acquisition and atracking part. Decision-directed estimation (DD) often works in tracking mode orthe data communication having preambles used to assist estimation, which is alsocalled data-aided estimation (DA). The preamble is a known data sequence thatmay be sent during the acquisition interval in order to establish synchronization.Once the initial phase error has been reduced to a suitably small amount by carrieracquisition, a tracking synchronizer (perhaps DD) can work efficiently . But if DDmode works without the first acquisition or preambles, it may take a long time toconverge or fail to converge because of the possible large initial phase error. TheDD requires the decision value of the transmitted signals. The performance isdirectly affected by the quality of its tentative symbol decisions. During actual datatransmission, the receiver decisions on the transmitted sequence may be regardedas accurate enough to consider them known. which will have performance virtuallyidentical to the known data estimator when errors are rare. Carrier acquisitionincluding power-law estimation, higher order statistics, histogram algorithm, etcwill be introduced in the following paragraph.(3) Non-decision-directed estimation The non-decision-directed estimation needn't decision value of the transmittedsignals. It can result from the data being modeled as a random process. Power-lawestimation (PL) is one of the classical algorithms. The algorithm only requires thestatistical mean estimation of the transmitted signals. The Monte Carlo simulationresults illustrate that at high SNR the effects of self-noise of the algorithm tend todominate the performance of the estimator, especially for cross constellations andthat increasing SNR to a high enough level has only a minor effect on performance. The estimation problem is even more complicated for cross constellationswhich do not have the corner points on which some simple carrier phaseacquisition algorithms are based, and which contain significant phase information.(4) Non-data-aided estimation (NDA) Because of efficiency reasons sometimes, the carrier acquisition must often beperformed without the use of training sequences or a preamble. It is often callednon-data-aided estimation. It is used especially in applications where no preambleis allowed or it is desired to estimate the parameter directly from detections on thereceived data signal. High-order statistics (HOS) algorithm in this paper is a representative. ForQAM it uses four-order statistics and doesn't require the knowledge of the order ofthe system. The method can be expected to work for any square and nonsquareQAM constellations. A literature analyzes that the HOS estimation and PLestimation are equivalent virtually.(5) Histogram algorithm (HA) What the histogram algorithm does is to produce in time as data arrive anestimate of the probability density of the phase angle averaged over all datasymbols and additive noise. The HA then finds the maximum of this densityfunction as the maximum likelihood estimation. To verify this, I derive anexpression for the density function of the phase angle of the received data.Histogram data obtained through simulation clearly converges to the deriveddensity function as the number of data observed increases. The histogramalgorithm can be used for square, cross and high order constellations. Thesimulation results illustrate the performance of HA at high SNR is better than PL,but at low and moderate SNR it is worse than PL.(6) Many-stage estimation algorithm Some papers developed combined algorithms such as many-stage estimationalgorithm.. The PL-DD algorithm is derived from it. The base idea of PL-DD isthat before the DD working we should give it a coarse phase estimation value bysome other algorithm which can do fast acquisition. PL estimation is selectedbecause of the better performance at low SNR cases. so that DD can converge in ashort time; then use the self-adjusting of DD to compensate the poor performanceof PL in moderate and high SNR. In order to acquire a better estimationperformance DD algorithm can work many times. The simulations illustrate theperformance of this algorithm is much better than one-stage algorithm such as HAand PL. The algorithm adapts to both small constellations and large constellations,But it has low estimation efficiency and that the computational complexity is toohigh to use it in practice.6. The main estimation algorithms of the carrier frequency offset The algorithms of carrier frequency offset can be classified like the carrierphase offset estimation. Application of maximum likelihood (ML) estimationindicates that the optimal solution to this problem amounts to locating the peak of aperiodogram. In many instances, however, this task is too time-consuming andsimpler methods are preferable such as data-aided estimation. Non-data-aidedestimation of the carrier frequency offset can be viewed as approximation to the...