MXPA00003599A - Synchronization techniques and systems for radiocommunication - Google Patents

Abstract

A method and apparatus for synchronizing and demodulating radio wave signals transmitted in frame format with an unique word, involving correcting the frequency offset of the received signal and differentially correlating the frequency corrected signal with the unique word. A first timing estimate is further refined by a 2-D search between frequency and time and a first frequency estimate is further refined by quadratic interpolation. The finely synchronized signal is demodulated using a Viterbi-based demodulator.

Description

SYNCHRONIZATION TECHNIQUES AND RADIOCOMMUNICATION SYSTEMS BACKGROUND OF THE INVENTION The present invention relates to digital radio systems and more specifically to the synchronization, as part of the processing, of a signal received in a radiocommunication system. Radiocommunication systems include the transmission of information in an air interface, for example, by modulating a carrier frequency with this information. When making a reception, a receiver tries to accurately extract the information of the received signal by performing an appropriate demodulation technique. However, in order to demodulate a received signal, it is necessary first to synchronize the timing between the transmitter and the receiver. For example, clock differences between the transmitter and receiver cause differences in bit timing. Furthermore, in some radiocommunication systems, information is transmitted in sudden increments, sometimes known as "frames". In these types of systems, it is also desirable to locate the beginning of a frame in such a way that the relevant information for a particular receiver can be isolated and demodulated. Unfortunately, there are numerous challenges associated with synchronizing a received signal. For example, even when the receiver is tuned to an assigned frequency at which its intended signal has been transmitted, a Doppler shift may result in a significant frequency shift between the frequency at which the receiver is synchronized and the actual frequency of the receiver. desired information signal when it reaches the receiver after it has traveled through the air interface. In addition, the crystal oscillator used in the receiver is only accurate within a certain number of parts per million, which may introduce an additional frequency shift. In addition to an unknown frequency offset, a receiver must also handle an unknown phase accuracy, that is, the receiver does not know the difference between the phase of the signal generated by its synthesizer at power-up and the phase of the received signal. Thus, the receiver faces at least three challenges to synchronize with the received signal: unknown timing, unknown frequency offset and unknown phase. Despite these challenges, the performance targets set for the recipients today are very high. For example, most receiver designs require that synchronization is almost always acquired "for example, 96% of the time" during the first frame in a sudden increase. This performance objective is even more striking in the field of satellite communication systems, where Doppler effects can be relatively important, energy limitations require the acceptance of a relatively low signal-to-noise ratio and the channels of often they can be relatively narrow. This last characteristic makes the frequency shift described above even more significant, since it is entirely possible that a desired information signal has a frequency shifted towards the center frequency of an adjacent channel. Due to the importance of synchronization and due to its impact on demodulation, the literature is full of several comments on these problems. For example, the impact of mobile cellular standards on the selection of modulator-demodulator (MODEM) and design is presented in an article by Camilo Feher entitled "MODEMS for Emerging Digital Cellular Mobile Radio Systems" (MODEMS for Emerging Cellular Radio Mobile Systems Digital) (IEEE Transactions on Vehicular technology, Volume 40, Number_2, May 1991). This article discusses several modulation techniques that are used in emerging radio systems of second generation. As indicated in this article, most system rules do not dictate the demodulation architecture. Manufacturing companies can employ coherent, differential or discriminating techniques for signal demodulation. The previous article focuses primarily on signals modulated according to II / 4-QPSK. Feher mentions that large frequency shifts make a shifted QPSK modulated signal inadequate for low bit rate communication systems. The present invention, however, overcomes this problem. A particular demodulation technique is described in a Gardner report entitled "Demodulator Reference Recovery Techniques Suited for Digital Implementation" (Demodulator Reference Retrieval Techniques Suitable for Digital Implementation) (Final Technical Report of ESA 1989, ESTEC Contract No. 6847 / 86NL / DG). Gardner suggests using a maximum likelihood approach to estimate the frequency error, and then a phase error detector to correct the phase. Another demodulation technique is described in an article by J. Ahmad et al entitled "DSP Implementation of a Preambleless All-Digital OQPSK Demodulator for Maritime and Mobile Data Commmunications" (IEEE, 1993, pages 4 / 1-5). According to Ahmad, phase-locked loops (PLLs), which are commonly used for coherent detection, have an inappropriate range of performance for the initial carrier frequency. Ahmad suggests the use of a double filter discriminator to estimate the frequency shift, an AFC loop to correct the frequency error, and a second order split loop for the detection of phase error. However, among other drawbacks, the applicants or consider none of these schemes as sufficient to achieve a sufficiently accurate synchronization with a sufficiently high first frame success rate due to the low signal-to-noise ratios. Accordingly, it would be advantageous to provide new techniques for synchronization with a received information signal that overcomes these drawbacks. COMPENDIUM OF THE INVENTION The present invention relates to the synchronization of digital radio signals. Exemplary embodiments of the present invention include, among other steps, the approximate correction of the frequency offset of a received signal and then the differential correlation of the signal with corrected frequency to the unique word. The differential correlation offers a correlation peak that provides a rough estimate of the timing. Since the approach to synchronization is aided by data, it is a rapid approach that ensures a high probability of detection-of sudden increases in the first frame transmitted. Exemplary embodiments of the present invention include the application of a 2-D search between frequency and time to determine whether there is a better peak correlation than the peak correlation identified by the differential correlation described above. This second correlation floor offers a better sample estimate and sudden increase timing. Subsequently, a quadratic interpolation refines the frequency estimate that corresponds to this second correlation peak. These timing and frequency estimates are then used to synchronize the received sequence. The synchronized signal is then demodulated using, for example, a demodulator based on a Viterbi algorithm. BRIEF DESCRIPTION OF THE DRAWINGS _ The present invention will now be described with reference to the accompanying drawings in which: Figure 1 illustrates a block diagram of a receiver in accordance with an exemplary embodiment of the present invention; Figure 2 shows an exemplary embodiment of the frequency offset approximate corrector of Figure 1 / Figure 3 shows an exemplary embodiment of the fine frequency and timing corrector of Figure 1, Figure 4 shows a synchronizer and demodulator in accordance with a exemplary embodiment of the present invention useful for M-Inmarsat terminals; Figure 5 is a graph illustrating a probability and fast detection versus a signal-to-noise ratio for a simulation of an exemplary embodiment of the present invention; Figure 6 is a graph illustrating an estimated frequency error versus a signal to noise ratio for a simulation of an exemplary embodiment of the present invention; Figure 7 is a representation of a mobile cellular system based on ground; and, Figure 8 is a representation of a mobile cellular system based on satellite. DETAILED DESCRIPTION When the first frame of a desired information signal is received by a receiver, in most communication systems, the clock in the transmitter and the clock in the receiver are not "latched", that is, they are not synchronized . It is also likely that there is a frequency shift due to fading, shading, Doppler shift and other forms of random frequency modulation. In addition, the relative phase of the received signal will not be known in systems where no phase reference is available. Thus, said received signal indeed has an unknown timing, an unknown frequency and an unknown phase. As shown in Figure 1 which is a high level block diagram of a receiver in accordance with an exemplary embodiment of the present invention, the received signal is input to a receiver through an antenna 1. The receiver filters the signal received by employing either an analog filter or a digital filter 3 that passes the energy around the carrier frequency to which the receiver is tuned. This prefiltration is carried out, among other things, to remove the noise introduced by the radio channel. Since it is unknown in what direction and in what degree a frequency shift has occurred, this prefilter 3 must be relatively wide in relation to the designed bandwidth of the channel to take into account the expected larger frequency offset. For example, a 10 KHz filter should be used in the case of a system that uses 5 KHz wide channels. The pre-filtered signal is initially corrected for the frequency shift in an approximate frequency corrector. As will be described in more detail below, the approximate frequency corrector 5 permutes the prefiltered signal in accordance with an approximate frequency shift estimate in order to provide an approximately corrected signal. (frequency) Although only a single block for the approximate frequency corrector 5 is illustrated in FIG. 1, those skilled in the art will note that two or more approximate frequency correctors will be provided sequentially upstream of the synchronization and frequency fine corrector 7 to carry perform a rough iterative correction until the correction of the frequency shift to a desired level. As will be described in more detail below with respect to FIG. 2, the approximate correction of the signal for the frequency shift in an approximate frequency corrector 5 can be carried out without using the unique word (synchronization) included in the received signal. In an exemplary embodiment of the present invention, the entire frame, that is, all the bits including the data bits and single word bits, is corrected for the frequency offset. The approximate frequency corrector 5 significantly reduces the overall time and power that are required to synchronize the received signal, especially in the case of cases in which large frequency shifts occur. Once the signal has been corrected approximately for frequency shift, the corrected signal may be filtered a second time using a narrower filter (not shown). Since the signal has been corrected for frequency shift, the second filter may be narrower than the filter 3 without losing part of the message signal, thus excluding additional noise. The fine frequency and timing corrector 7 then performs a differential correlation with the unique word in the totality of the approximately corrected signal to generate a correlation peak, as will be described in more detail below in relation to FIG. 3. The correlation peak offers a rough estimate of the timing. Then, a 2-D search between frequency and time (ie, the application of a time offset and in turn the application of a frequency rotation) can be used to determine if there is a correlation peak. The 2-D search results in a new set of correlation values. The correlation peak of this new set of correlation values is considered as providing the best estimate of sample timing and sudden increases. The frequency estimate that corresponds to the best estimate of timing is refined by interpolation (regularization). This interpolation finds a "best fit" frequency. Once the best estimates of timing and frequency are obtained, the signal is finely synchronized and ready to be demodulated by a demodulator 9. The demodulator 9 can be any demodulator, for example, those that use the Viterbi algorithm that tracks data received through a set of states and retrieves the message signal accordingly. 2-D search and quadratic interpolation can be repeated for better timing and frequency estimates. Figure 2 illustrates a more detailed exemplary embodiment of the approximate frequency corrector 5 illustrated in Figure 1. Here, an input signal that arrives on line 20 is sampled a predetermined number of times, for example, four times per bit, for provide an input sample stream. Since each bit is sampled four times, it is possible to take the first sample of each bit to form a first set of samples, the second sample of each bit to form a second set of samples, etc. The approximate frequency corrector 5 employs, in this exemplary embodiment, a differential detector 22 to differentially detect each set of samples in order to provide four sets of differential samples. Considering that a received sample is represented as; ? n. (1) where e3 = the frequency shift in time n; Sn = the energy associated with the desired signal at time n; and nn = the energy associated with the noise at time n. (2) later, the previously received sample, that is, at time n-l, would be represented by: the differential detector 2 carries out the operation rnr * n- ?, where the symbol "*" represents conjugation. This operation is expanded using equations (1) and (2) and you get:. ?? V «* - t W?« "'" .-? -.-? > + S "« V? + S -? < "+"? -i (3) The approximate frequency corrector 5 averages and determines the scale of the differential samples using an averaging device 24 and a scale counter 26 to provide an approximate estimate of frequency shift?. Taking a long-term average (ie, expected value) of the outputs of the differential detector 22, and considering a Gaussian noise distribution, the last three terms in equation (3) will approach 0. The scale counter 26 eliminates the bias introduced by the averaging device 24, as will be observed by experts in the field. In block 28, the conjugate of the approximate frequency shift estimate is formed to prepare the estimate for use to correct the frequency offset in the multiplier 30. That is, by multiplying the samples received with the conjugates of The approximate estimates of frequency offset, the input sample current is permuted to the correct frequency (tuned). An approximately corrected signal is then sent on line 32 to either the fine frequency and timing corrector 7 or to another approximate frequency corrector 5, if another iteration of the above described technique is desired to provide a higher frequency accuracy before synchronization with the received signal. Although this exemplary embodiment of the present invention describes the approximate frequency estimator 5 employing a differential demodulator (detector) 22 to provide phase difference information, those skilled in the art will note that other devices may be employed to provide this phase information. differential in unit 5. For example, two coherent correlators can be used to correlate the input symbol current with the unique word at time n and at time n-1 and take their phase differences. Using this alternative, however, it will be noted that the duration of the correlations must be sufficiently short that the frequency of the received signal does not change too much during the correlation process. The resulting phase differences are then averaged, placed on a scale and used to permute the received signal in accordance with what is described above. Figure 3 illustrates an exemplary embodiment of the fine frequency and timing corrector 7 illustrated in Figure 1.

As the input, the fine frequency and timing corrector receives a corrected input current approximately, for example, on line 32. This input current is differentially demodulated again in block 34. By differential demodulation of the symbol stream in block 34, the effects of any remaining frequency rotation, that is, not corrected by the approximate frequency corrector 5, are the differences between adjacent symbols. Each sudden increase in information includes a data field or word (also known as a synchronization word sometimes) that is unique to this sudden increase or time segment. This unique word is known from the receiver and can be compared to the signal received as part of the synchronization process using correlation techniques. Since the input symbol stream is differentially demodulated in block 34, the known single word is also differentially demodulated in block 36 to provide inputs analogous to correlator 38. Correlation can be found by using, for example, two fast transformations of Furier (FFTs) and circular convolution. Correlator 38 produces correlation values; One of these values is a correlation peak, this type of correlation is an approximate estimate of the timing. However, this peak is not the "true" correlation peak due for example to some remaining frequency offset. Since the "true" correlation peak will be found at least in the vicinity of the peak determined by the correlator 38, a certain number N of additional samples on both sides of the peak are also selected for further processing in block 40. Then, the fine frequency and timing corrector 7 employs a better timing estimator 42 to perform a 2-D search between frequency and time (i.e., by applying a time offset and in turn applying a frequency rotation) to determine whether a correlation peak better than the peak identified by the correlator 38. The time parameters for the search are set for the selection of samples in block 40. Simulations may be used to provide an estimate of the maximum remaining frequency offset associated with the input signal to the fine frequency and timing corrector 7. For example, simulations can show what e for a worst-case SNR, the remaining frequency offset must not exceed + 200 KHz. This information is used to set the frequency parameters for the search. The search can be conceptualized in the following way. Suppose that two axes are established - one for time and one for frequency. Numerous divisions associated with the N + l samples selected in block 40 are established along the time axis. Along the frequency axis, the maximum frequency shift error (for example, + 200 KHz) is divided into a reasonable number of multiples (for example, 40 multiples of 10 KHz each). More generally, this sets N + l times x M frequency shifts "bins". Note that by the approximate correction of the frequency offset upstream of the fine frequency and timing corrector 7, the number of "bins" used in the 2-D search engine 42 is reduced, which allows the present invention to synchronize the received signal more quickly and / or use lower MIPS processing power. Then, for each bin, that is, for each combination of N + l time shifts and M frequency shifts, the received input current and the single word are coherently correlated. A more detailed description of this type of 2-D search technique can be found in U.S. Patent No. 5,151,926, the disclosure of which is expressly incorporated herein by reference. The best timing estimator 42 therefore provides a new set of correlation values. The correlation peak of this new set of correlation values gives the best estimate of sample timing and sudden increase. The frequency estimate fo corresponding to the best estimate in timing is further refined (regularized) in a quadratic interpolator 44. First the quadratic interpolator 44 chooses two frequencies, fi and f2 in such a way that fi is slightly smaller than fo, and f2 be slightly larger than fo. The quadratic interpolator 44 performs a coherent correlation using fo, fi, and f2 and short segments of the single word, By using short segments of the single word a symbol is prevented from rotating in another symbol due to frequency changes during the correlation. The quadratic interpolator 44 adds magnitudes of the segments in a non-coherent manner to provide three new correlation values y0, yi, e y2 respectively. The best timing estimate is then found by using these values to evaluate equation (4) below. The quadratic interpolator then determines the maximum of the quadratic to determine the remaining frequency offset. o-where x ^ f ^. Having obtained a fine estimate of any remaining frequency shift 7, the fine frequency and timing corrector 7, swaps the input data stream using a multiplier 46 to provide the finely tuned signal. Finally, the fine frequency and timing corrector 7 performs a final correlation in the correlator 48 to locate the unique word and send the synchronized signal to, for example, the demodulator 9. The 2-D search and the quadratic interpolation can be iterated with smaller frequency bins to provide more accurate estimates. Figure 4 illustrates an exemplary embodiment of the present invention that is useful for Inmarsat-M terminals or the like. some of the details of the techniques described above are omitted here for clarity. The received data 91 can be buffered but is then filtered using a finite impulse response (FIR) prefilter 92 of sixty-four leads at a rate of four samples / bit. Filter 92 is Gaussian transition up to -6 dB, that is, it has a width of approximately 6.84 kHz. the filtered data 93a is differentially detected by a differential detector 93b (1-bit spacing) for each sample point in order to provide four sets of differential samples. The differential samples 93c are averaged by averaging device 93d and scaled by a scale counter 93f to provide an approximate estimate of frequency shift 93g. The filtered data 93a is permuted by a multiplier 93h to provide a stream 93i of data corrected by frequency. This stream of data corrected by frequency 93i is filtered using a back filter 94 FIR of sixty-four derivations. The filter 94 has a width of 5 kHz and has a bandwidth of 3dB of 4.75 kHz. The current 95a of data corrected by filtered frequency is correlated differentially using a differential correlator 95b. The differential correlator 95b employs two FFTs of 2047 points and circular convolution. The differential correlator 95b provides a set of correlation values and a correlation peak 95c. A better timing estimator 95d conducts a 2-D search (a consistent correlation) in twenty one frequency bins and nine sampling points in relation to the correlation peak 95c. The best peak 95e of the new set of correlation values provides the sample timing and the sudden increase timing. The three best correlation values are taken from the 2-D search and interpolated quadratically by an interpolator 95f. The stream 95a of data corrected by filtered frequency is permuted by a multiplier 95h to provide the finely synchronized data stream 95i. The finely synchronized data stream 95i is demodulated by a Viterbi demodulator 96 using thirty-two states. The Viterbi demodulator calculates the branching metric values using the unique word to estimate the random phase error. When the approach illustrated in Figure 4 and described above was simulated with the QPSK modulation of displacement at a data rate of 5.7 kbps, with a signal-to-noise ratio (Eb / No) of 2 dB in a 5 kHz channel A GN, a square root cosine transmission filter with alpha = 0.6, and a continuous frame transmission with each frame comprising ninety-six single word bits followed by 576 data bits and twelve dummy bits, a probability was observed 98% of frame acquisition within 120 ms. In addition, the frequency error was estimated within a standard deviation of +/- 3 Hz. Figure 5 is a graph illustrating the simulation results showing the probability of detection of the frame with an accuracy of +/- 1/2 bit (1 bit precision) for the range of Eb / No values. The results indicate that in a ratio between signal and noise of 2 dB, the exemplary embodiment of Figure 4 provides a frame acquisition within a 1 bit precision with a probability of approximately 98%. That is, the technique of the present invention locates the start of the unique word within +/- 1 bit around the true start sample. Note that this detection probability depends on the presence of a valid transmitted frame, in other words, it was not considered a false alarm probability in this simulation. Figure 6 illustrates the standard deviation of residual frequency error for a range of values Eb / No. It is noted that the technique of the present invention offers a frequency estimate with a standard deviation of an accuracy of +/- 3 Hz at an Eb / No of 2 dB. This error corresponds to a phase rotation of approximately +/- 0.2 degrees / bit. Figure 7 illustrates a mobile cellular system based on general ground in which the techniques described above can be implemented where the receiver 112 is joined to another mobile or personal mobile user (not illustrated) by a radio transmitter based on ground 110. radio-transmitted signal 111 is synchronized and demodulated in accordance with the present invention. Figure 8 illustrates a general satellite-based mobile cellular system wherein a receiver 115 is linked to another personal or mobile telephony user (not illustrated) through a satellite-based radio transmitter 113. The signal transmitted by radio 114 is tuned and demodulated in accordance with the present invention. Progress in very large scale integration technology (VLSI) and digital signal processing (DSP) made digital modulation more economical than analog transmission systems. Programmable digital signal processors made it possible to implement digital modulators and demodulators entirely in programmatic. Accordingly, those skilled in the art will note that the equipment and / or programmatic implementations of the present invention are within the scope of the present invention. Multiple access schemes, for example, multiple frequency division access (FDMA), time division multiple access (TDMA) and code division multiple access (CDMA), are used to allow many subscribers to simultaneously share a width of finite band within the radio spectrum. Exemplary embodiments of the present invention depend on a particular multiple scheme and, therefore, may be employed in combination with these or other schemes. The present invention is not limited to existing schemes of digital modulation. It is useful for the demodulation of any radio-transmitted signal that requires frequency estimation and sample timing of sudden increments. The present invention has been described through exemplary embodiments to which the invention is not limited. Those skilled in the art will be able to devise modifications and changes without departing from the spirit and scope of the invention in accordance with that defined in the appended claims.

Claims (1)

1. CLAIMS __ A receiver comprising: a frequency estimator for determining a first frequency shift estimate associated with a received signal; a device for correcting said received signal using said first estimate to generate a corrected signal; a frequency and timing estimator receiving said corrected signal and determining a second frequency shift estimate using a single known word of said receiver and to produce a timing estimate; a device for further correcting said corrected signal employing said second estimate to generate a signal corrected by frequency; and a device for synchronizing with said frequency corrected signal employing said timing estimate. The receiver according to claim 1, wherein said first frequency estimator further comprises: a device for providing differential phase information between samples of said received signal; an average determination unit for averaging said differential phase information in order to generate an average value; and a scale determining unit for determining the scale of said average value to adjust for bias and to generate said first estimate. The receiver according to claim 1, wherein said means for correcting further comprises: a multiplier for multiplying a value related to said first estimate with said received signal. The receiver according to claim 1, wherein said frequency and said timing estimator further comprises: a first differential demodulator to differentially demodulate said corrected signal; a second differential demodulator to differentially demodulate said unique word; and a correlator to provide a first set of correlation values between said differentially demodulated corrected signal and said differentially demodulated single word, wherein a correlation peak associated with said correlation values provides an initial value of said timing estimate. The receiver according to claim 4, wherein said fine frequency and timing estimator further comprises: a device for selecting said correlation peak a predetermined number of samples surrounding said correlation peak; a two-dimensional coherent correlator employing said selected number of samples and said correlation peak to correlate said unique word with a plurality of versions with frequency offset and time offset of said received signal to generate a second set of correlation values; and an interpolator to regularize a selected subset of said second set of correlation values in order to generate said second estimate. The receiver according to claim 1, wherein said further correction device further comprises: a multiplier for multiplying a value related to said second estimate with said received signal. The receiver according to claim 1, wherein said synchronizing device further comprises: a correlator to correlate said unique word with said frequency corrected signal to identify a location of said unique word within said received signal. The receiver according to claim 2, wherein said device is a differential demodulator. The receiver according to claim 2, wherein said device includes at least one coherent correlator. The receiver according to claim 1, wherein said received signal is received by employing one of the following access methodologies: FDMA, TDMA and CDMA. 11. A receiver for receiving a signal including an information table, said information box includes a single word, said receiver comprises: an approximate estimator to estimate approximately a frequency shift of said signal and to correct said signal by employing said estimated displacement; a fine estimator for receiving said corrected signal, estimating a fine frequency offset, correcting said corrected signal using said estimated fine frequency offset, and identifying a location of said unique word within said frame; and a demodulator for demodulating said signal using said location. 12. A device for estimating a frequency offset associated with a received signal, comprising: a unit for providing differential phase information between samples of said received signal; a unit to establish averaging to average said differential phase information in order to generate an average value; and a scale determining unit for determining the scale of said average value in order to adjust to take into account the bias and to generate said frequency shift estimate. . The device according to claim 12, wherein said unit is a differential demodulator. . The device according to claim 12, wherein said unit includes at least one coherent correlator. . A device for estimating a frequency offset associated with a received signal and for estimating a location of a single word within said received signal, comprising: a first differential demodulator for differentially demodulating samples associated with said received signal; a second differential demodulator to differentially demodulate said unique word; a correlator to provide a first set of correlation values between said differentially demodulated received signal and said differentially demodulated single word, wherein a correlation peak associated with said correlation values provides initial values of said frequency shift estimate and said timing estimate; and a device for determining whether said initial values of said frequency shift estimate and said timing estimate can be improved. . The device according to claim 15, wherein said means for determining further comprises: means for selecting said correlation peak and a predetermined number of samples surrounding said correlation peak; a two-dimensional coherent correlator employing said selected number of samples and said correlation peak to correlate said unique word with a plurality of frequency shift and time offset versions of said received signal to generate a second set of correlation values; and an interpolator to regularize a selected subset of said second set of correlation values to generate said frequency shift estimate and timing estimate. The device according to claim 16, wherein said interpolator is a quadratic interpolator that operates in accordance with the following equation: > . "% -a (x + x2) XeTX2 where y0, yi and y2 are said selected subset of correlation values; and x0, xi and x2 are the frequencies that correspond, to i, yi and y2, respectively. 18. A method for synchronizing a data stream transmitted in a frame format and having a unique word, comprising the steps of: adjusting the data stream to provide a data stream corrected by frequency; correlating said corrected data stream by frequency with a copy of said unique word to provide a plurality of correlation values; calculating a frequency estimate using at least one of said various correlation values; correct said * data stream using said frequency estimation; Y. synchronize said corrected data stream. 9. A method according to claim 18, wherein said step of adjusting the data stream provides a corrected data stream per frequency further comprising the steps of: differentially demodulating samples of said data stream; and averaging and determining the scale of said differential samples to generate a rough estimate of the frequency shift; and permute said data stream by said approximate frequency shift estimate. 20. A method according to claim 18, wherein said correlating step further comprises the steps of: differentially demodulating said stream of corrected data by frequency and said unique word before performing said correlation. 21. A method according to claim 18, wherein said calculating step further comprises: carrying out a 2-D search on several frequency bins and several selected sample points based on at least one of said various correlation values in order to identify a set of correlation values, and to carry out a quadratic interpolation of said set of correlation values and a set of corresponding frequency estimates to identify said frequency estimate. The receiver according to claim 11, further comprising: a first filter upstream of said approximate estimator for filtering said signal; and a second filter for filtering said corrected signal before the introduction of said corrected signal to said fine estimator, said second filter is narrower than said first filter.