WO1999063721A1

WO1999063721A1 - Symbol synchronizing method

Info

Publication number: WO1999063721A1
Application number: PCT/FI1999/000456
Authority: WO
Inventors: Hang Zhang; Marko HEINILÄ
Original assignee: Nokia Networks Oy
Priority date: 1998-05-29
Filing date: 1999-05-27
Publication date: 1999-12-09
Also published as: FI106502B; FI981226A0; AU4517099A; FI981226A

Abstract

The invention is related to receiver structures, namely structures and methods for synchronization of a receiver to received data symbols. Symbols synchronization is obtained by using the length of a difference vector as the synchronization variable, and finding the averaged maximum of said length of a difference vector. The method according to the invention allows symbol synchronization even in presence of DC offset and other types of low frequency noise.

Description

Symbol synchronizing method

TECHNICAL FIELD OF THE INVENTION

The invention is related to receiver structures, namely structures and methods for synchronization of a receiver to received data symbols.

BACKGROUND OF THE INVENTION

Continuous phase modulation (CPM) is a class of modulation techniques, which are efficient in power and bandwith at the same time. In all continuous phase modulations, the RF signal envelope is constant and phase varies in a continuous manner. The constant envelope allows the use of nonlinear amplifiers, simplifying receiver and transmitter design. Tamed frequency modulation (TFM) is one continuous phase modulation scheme. One of the main benefits of TFM is the very narrow bandwith needed in comparison with most of the other modulation schemes. CPM signals can be described by

5(0 = (2E_b I T_bf⁵ cos(2πf₀t + φ(t,c ) + φ₀) (1)

where the excess phase function φ(t,a) is given by

= 2π∑a_ih_iq(t -iT_b), nT_b< t < (n+l)T_b (2)

(=0

where

{ α,. }are the M-ary data symbols, M even, taken from the alphabet

±1, ±3, ...., ±(M-1), /7, is the modulation index, which may vary from interval to interval, q(t) is the phase response function, g(t) is the frequency response, E. is the bit energy,

T_b is the bit interval, /o is the carrier frequency, and φ₀ is an arbitrary initial phase. For TFM, M equals 2, wherefore data symbols α, = ±1, and h_t = 0.5. Also for TFM, the bit period has the same length as a symbol period.

Characteristic for TFM modulation is, that the phase shift of the modulated carrier over one bit interval is determined not only by the current bit but by three consecutive input binary signals in accordance with the encoding rule:

φ{nT_b) = (/„_, +2I_n +I_n+1) (3)

where I_n represents the binary data at time t - nT_b and I_n =±1, i.e. for example -1 corresponds to bit 0 and +1 corresponds to bit 1. One can see from this coding rule, that phase changes of π/2 are obtained, if three successive bits have the same polarity, and the phase remains constant for three bits of alternating polarity. Phase changes of π/4 are connected with the bit configurations ++ -, H , -++ and K

The signal space diagram of TFM is shown in Figure 1. By introducing the correlation between the transmitted bits, TFM signal provides a narrower power spectrum than for example a MSK signal, since the changes of phase are smoother in TFM. TMF modulation is described further in the article entitled "Tamed Frequency Modulation, A Novel Method to Achieve Spectrum Economy in Digital Transmission", Fank de Jager and Cornelis B. Dekker, IEEE Trans. on Comm. Vol. COM-26, NO. 5, May 1978, pp. 534-542. CPM modulation is described further in the book "Digital Phase Modulation" by John B. Anderson, Tor Aulin and Carl-Erik Sundberg, Plenum Publishing Corporation, 233 Spring Street, New York, N.Y. 10013, on pages 15 to 53.

Direct conversion receivers are receivers, which do not use intermediate frequencies for filtering and detection of received signals. In a direct conversion receiver, the received RF signal is mixed with a local oscillator signal, whose frequency corresponds to the carrier frequency of the RF signal. Direct conversion receivers have many advantages. For example, the bandwith filtering can be performed at low i.e. audio frequencies, allowing realisation of very narrow bandwiths with sharp edges. Also, intermediate frequency filters are not needed. However, direct conversion receivers have not been used for receiving TFM signals due to an inherent problem of direct conversion receiver structures, namely the existence of a DC offset at the mixer output as a result of imperfections of the mixer structure. Typically, the DC offset results from leaking of the local oscillator signal to the RF port of the mixer and subsequent mixing of the leaked signal with the local oscillator signal itself. The random fluctuations in phase of the leakage signal result in a relatively slowly and randomly changing DC offset signal.

In addition to DC offset problems inherent in direct downconverting applications, also other forms of low frequency noise are often present in a baseband signal. Low frequency phase noise resulting from transmitter phase noise or local oscillator phase noise can be presented as an example. The DC offset problem can be thought of as a very low frequency phase noise.

Figure 2 illustrates the problem created by the DC offset in the detection of TFM modulated data. Without any DC offset, the received signal has the constellation shown with white circles. The vectors Si and s₂, drawn with a thick dashed line, represent signals detected at two consecutive sampling times. Without any DC offset, the detection of vectors Si and s₂ is straightforward. However, DC offset changes the situation considerably. DC offset has the effect of moving the signal constellation in the IQ-diagram, as depicted by the black circles. The vectors s\ and s'₂ show the corresponding sampled signal vectors in the presence of DC offset. Clearly, any detector optimized for detecting the vectors Si and s₂, has difficulties in recognizing vectors s\ and s'₂. Therefore, the presence of DC offset results easily in a high detection error rate.

The only solution known so far for this problem is to not use a direct downconverting receiver, since this DC offset problem does not occur in heterodyne receivers.

Various coherent detectors or differential detectors have been studied for TFM modulated signals. However, these detectors based on signal phase or phase difference over one bit interval do not work well under DC offset, since the phase of a given constellation point will change due to DC offset, and also the phase difference of two consecutive signal vectors will change due to DC offset.

This problem has been solved for MSK type modulation. The patent application

W094/28662 describes a double differential detector structure, which is described in the following. If r_n is defined as the signal vector at time nTb, the difference vector between two received vectors over one bit interval is defined as

= r_n -r_n__x (4) and the phase difference δφ_n between Δ„ and Δ_n_. is defined as

<^„ = Z(Δ„) - Z(Δ„_.) (5)

and the decision could be made as follows: if 0 < δφ_n < 135° , then transmitted bit bn = 1; if 135° < δφ_n < -135° , then transmitted bit bn != bn-i; if -135° < δφ_n < 0, then transmitted bit bn= 0. This method works for MSK, since in MSK, the phase changes during every bit interval.

When we consider the idea of a double differential detector for TFM, the situation is a little different. For TFM, the quantity δφ_n is not a sufficient measure to make a detection, since the difference vector A_n can be zero. As described previously, in

TFM modulation, the phase does not change with certain information bit patterns. Therefore, the quantity δφ_n can also be zero for some bit patterns, which makes it impossible to make decisions about the bits based on the value of the quantity δφ„ .

Symbol synchronization refers to the adjustment of receiver sampling timing to those points in time, when the received signal parameters, e.g. phase and amplitude, are at a modulation constellation point. If the receiver timing is incorrect, the samples may be taken far from the constellation points, for example between the constellation points in Figure 1. In an article entitled "Simple baseband symbol synchronization scheme for tamed frequency modulation", S. Bellini, Electronics Letters, March , 1997, vol 30 No. 7, pp 548-549 a symbol synchronization scheme is presented in which the timing error information is extracted from the phase or the phase difference between two adjacent received vectors. Under DC offset, the origin of the received signal constellation will drift away from the expected location, which affects considerably the signal phase or phase difference over one bit interval. Therefore this method is sensitive to DC offset and can not be used for TFM symbol synchronization in a receiver having the DC offset problem. SUMMARY OF THE INVENTION

The object of the invention is to realize a symbol synchronization method for TFM signals, which performs well even under DC offset and other types of low frequency noise.

The objects are reached by using a parameter substantially dependent on the phase difference of sample vectors as the synchronization variable, such as the phase difference itself or the length of a difference vector between two sample vectors, and finding the averaged maximum of said parameter to obtain the optimum synchronization.

The symbol synchronization method according to the invention is characterized by that, which is specified in the characterizing part of the independent method claim. The symbol synchronization system according to the invention is characterized by that, which is specified in the characterizing part of the independent claim directed to a symbol synchronization system. The mobile communication means according to the invention is characterized by that, which is specified in the characterizing part of the independent claim directed to a mobile communication means. The base station according to the invention is characterized by that, which is specified in the characterizing part of the independent claim directed to a base station. The radio link system according to the invention is characterized by that, which is specified in the characterizing part of the independent claim directed to a radio link system. The dependent claims describe further advantageous embodiments of the invention.

Since in most cases the DC offset changes relatively slowly, one can assume that the DC offset is a constant during one bit interval. Therefore the effect of DC offset can be eliminated by taking the difference between two adjacent received vectors. In the following, r„(τ) = r(nTb + τ) denotes the received signal at time t = nT + τ where T_b is the lenght of the symbol in time, r_n.j(τ) = r((n-l)T_b + τ) denotes the received signal at time t = (n-l)T_b + τ. Then,

ι(τ) = r_n(τ) - r„._}(τ) (6)

denotes the difference signal between r_n(τ) and r„.ι(τ), where τ, 0 < τ < T_b, denotes the timing offset from the optimal time instant nT_b. When the DC offset signal r_DC changes very slowly in comparison to the symbol length T_b, r_DC (nT_b + τ) « r_DC ((« - \)T_b + τ) , 0 < τ < T_b, whereby we have

Δ„ (T) = (r_n (r) + r_DC,n (r)) - (r„_, (τ) + r_DW(τ)) r_n(τ) - r„_ι (τ) (7)

Therefore the difference signal _d_M(τ) is free of the effect of DC offset. Any further information extracted from the difference vector will also free of the effect of DC offset. Further, the inventor found that the length of the difference vector, i.e. the quantity |Δ„ (r)| is able to provide the required timing error information. Namely, the quantity |Δ„(r)|, i.e. the average of |Δ„ (T)| has the maximum at the correct sampling time instants nT_b, n = 0, 1, 2,... . Therefore, in an advantageous embodiment of the invention, a tracking loop is used to search the maximum of |Δ„ (r)| or an extremum of another quantity depending substantially on |Δ„(r)| for obtaining the correct liming.

In one advantageous embodiment of the invention, the average of square of the length of the difference vector, i.e. |Δ„(r)| , is used as the variable for obtaining correct timing. The invention is not limited to searching of a maximum of a function depending substantially on |Δ„(r)| . For example, in a further advantageous embodiment of the invention, ' the minimum of 1 - | IΔ n„ (Γ) '|I / | IΔ " Imax is searched for obtaining the correct timing. In the previous equation,

is the largest possible value of |Δ_n(r)|. Any function depending substantially on |Δ„(r)| and having an extremum substantially when |Δ„(r)| has a maximum, may be used in various embodiments of the invention for searching the correct symbol sychronization. Further, any averaging in various embodiments of the invention may advantageously be performed on the basis of a predefined number of quantities to be averaged, i.e. using an averaging window having a certain predefined length.

To facilitate the finding of the maximum of |Δ„(r)|, the received signal can be oversampled. The received signal vector can advantageously be sampled at the rate _s = r/T , where r is preferably an integer equal to or larger than 2. For each sample, a corresponding difference vector is calculated by computing the difference between the sampled signal vector and the signal vector sampled T_b ago. By averaging the lengths of difference vectors obtained with the same time offset from the sampling time nT_b, a set of values for |Δ„(r)| as a function of time offset from the sampling time nT_b is obtained. Figure 3 illustrates an example of such a set of values, marked by crosses. Figure 3 illustrates 11-fold oversampling, i.e. sampling with the value of r = 11. In one advantageous embodiment, the sampling times are adjusted according to the time offset of the highest obtained average.

In another advantageous embodiment of the invention samples s(nT_b - At) and s(nT_b + At) are taken a time offset At before and after the sampling time instant nT_b. The difference f(τ) of the averages of lengths of the corresponding difference vectors A„(nT_b - At) and A„(nT_b + At)

f(τ) =

- \A_n (nT_b +At + r)| (8)

is used to control the adjusting of the sampling time instant. In equation (8) the variable τ denotes the adjustment of the sampling time. The value of τ is changed, until the difference /(r) is zero. Searching for the zero of the difference f(τ) may be performed in many ways and with many different search strategies. This method effectively corresponds to finding a zero of the derivative

different search strategies for finding a zero of a derivative of a function are known by the man skilled in the art, and therefore such strategies are not described here in further detail. The time offset At may advantageously be At = T_bIA.

It may be noted here, that instead of averaging of the lengths of the difference vectors as described previously, the averaging may as well be performed on the results of the difference function. In other words, equation 8' is calculated first for a number of samples,

f(τ) = |Δ„ (nT_b - At + _r)| - \A„(nT_b + At + r)| (8')

and the result is averaged for obtaining /(r) .

In a further advantageous embodiment of the invention, no oversampling is used. Instead, the sampling time nT_b is changed a small amount, and the resulting change in the value of the quantity |Δ„| or another quantity substantially depending on |Δ_n| is monitored. If the resulting change is negative, then the sampling time is changed in the opposite direction. If the resulting change is positive, the sampling time is changed further in the same direction. If the resulting change is approximately zero, the sampling time is deteπnined to be close enough to the optimum time. Naturally, the limits defining how large a positive change is determined to be a positive change, how large a negative change is determined to be a negative change, and which values of the resulting change correspond to being approximately zero, may be set in any desired way to optimize the performance of the adjusting method. The optimum values for the limits may be found, for example, by simulation or experimenting.

In a further advantageous embodiment of the invention, instead of oversampling, interpolation is used to produce calculated signal samples representing samples taken before and after the actual sampling time. For example, according to such an embodiment, the difference f(τ) can be calculated by interpolation using three consecutive difference vectors:

/(r)

where the coefficient c may be for example 0.75 and the coefficient d, 0.25. These values of c and d correspond to At = T_b/4. The example of equation (9) shows linear interpolation as an example, but any other type of interpolation and other values for the coefficients c and d for obtaining other values of At may be used as well in various embodiments of the invention. For example, parabolic or cubic interpolation may be used. Use of interpolation has the additional benefit, that no oversampling is needed, which simplifies the construction of the system. Such a low sampling rate can be used, since the bandwith of TFM modulated signals is very narrow, being roughly .5/T_b. Therefore, a sampling rate of \IT_b is high enough to fulfill the Nyquist criterion.

The length of a difference vector can be used in many ways for finding the optimum symbol synchronization. Generally these methods correspond to finding the maximum of the average length of the difference vector by varying a timing offset parameter. Difference vectors may be sampled at different values of the timing offset parameter by oversampling and/or interpolating sampled values. These methods for finding the optimum synchronization, some examples of which have been given in the previous paragraphs, can be classified into two main groups: feedforward type methods and feedback type methods. In the feedforward type methods, a number of difference vector lengths are obtained at various timing offset parameter values. The optimum timing may then be found by observing, which timing offset gives the longest averages, or at which timing offset values the derivative of the resulting distribution i.e. differences between neighboring averages is nearest to zero.

In the feedback type methods, a derivative-based i.e. a difference-based approach is most advantageous. In this type of methods, the difference of the average lengths of difference vectors at two observation offset time instants is used to control the synchronization timing in order to bring the difference of the average lengths close to zero, i.e. to bring the maximum of the average lengths of the difference vectors between the two observation offset time instants. The difference vectors at the two observation offset time instants may be obtained by oversampling and/or by interpolation. Advantageously, double oversampling is used, although other types of oversampling may be used as well.

When the DC offset or any other type of low frequency noise does not disturb the signal too much, the phase difference of signal sample vectors can be used to finding the optimum synchronization timing in a similar manner as the length of a difference vector. The correct symbol timing τ maximizes the average of expression (10):

i.e. the average of the absolute value of the signal phase change over T_b. The previous embodiments, which teached the use of difference vector length, can be used with difference of phases as well.

Generally, in the inventive method, the value of a parameter substantially dependent on the phase difference of sample vectors is maximised or rninimised in order to find the optimum synchronization. The parameter is advantageously the phase difference itself. In the presence of DC offset or other types of low frequency noise, it is advantageous to use the length of a difference vector as said parameter.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention is described in more detail in the following with reference to the accompanying drawings, of which Figure 1 illustrates the symbol constellation of tamed frequency modulation,

Figure 2 illustrates the effect of DC offset,

Figure 3 illustrates the finding of the synchronization by oversampling,

Figure 4 shows a block diagram of an advantageous embodiment of the invention,

Figure 5 illustrates an advantageous embodiment of the invention applied in a receiver structure using difference vector length for detection of transmitted data,

Figure 6 illustrates a block diagram of an advantageous embodiment of the invention, in which the timing of sampling of the received signal is adjusted,

Figure 7 shows a block diagram of a mobile communication means according to the invention,

Figure 8 illustrates further advantageous embodiments of the invention, and

Figure 9 illustrates the block diagram of an advantageous embodiment of the invention, in which the angle of the IQ vector is used as the control variable.

Same reference numerals are used for similar entities in the figures. Figures 1, 2, and 3 were explained previously.

DETAILED DESCRIPTION

Figure 4 illustrates a block diagram of a receiver structure according to an advantageous embodiment of the invention. Such a receiver structure can be used, for example, in mobile communication means and in base stations of cellular mobile communication networks. The receiver structure comprises a sampling part 200 and a synchronizer part 400. The sampling part receives the RF signal, and converts the received signal to baseband I and Q signals with mixers 210a, 210b. A local oscillator 205 provides the signal to be mixed with the received RF signal, and a phase shifter 206 produces a 90° phase shift to the local oscillator signal taken to the mixer 210b. The output signals of the mixers 210a and 210b are taken to matched filters 215a, 215b, whose filtering properties are optimized for filtering TFM signals. The output signals from the filters 215a, 215b is taken to switch elements 225, which produce samples of the quadrature I and Q signals. Each corresponding pair of I and Q samples define a signal sample vector. The switch elements 225 are controlled by a sampling oscillator 220. The sampling part 200 is an example of a structure, which can be used in a receiver structure according to the invention. As a man skilled in the art knows, many other structures can be used to produce a sampled, downconverted signal. The invention is not limited to using the sampling part structure 200 shown in Figure 4. The switch elements 225 performing the sampling comprise typically A/D converters or other types of sampling means.

The resulting I and Q samples are taken to synchronizer part 400, where the samples are taken to interpolation blocks 430a, 430b. The interpolation blocks 430a, 430b produce calculated samples on the basis of the I and Q signal samples, and perform the calculation according to the value of the control signal from the loop filter block

455. The interpolated I and Q signal samples are taken to a delay block 435a, 435b and an adding block 440a, 440b, which delay and adding blocks calculate the difference of consecutive I and Q samples. The delay blocks 435a, 435b delay the I and Q signals for the period of one T_b, so that the differences are calculated of samples taken at corresponding times of symbol periods, if oversampling is used.

The difference results are taken to a difference vector length calculating block 445, which calculates the length of the difference vector corresponding to each pair of I and Q sample differences. The calculated lengths are taken to difference function calculation block 450, which may calculate the difference function f(τ) according to equation (8'), for example. The result is taken to loop filter block 455, which performs the averaging of the differences. The value of f(τ) or a corresponding value is then used to control the interpolation blocks 430a, 430b.

The embodiment of Figure 4 presents an example of obtaining the signal sample vectors corresponding to correctly symbol synchronized sample vectors by calculation, namely by interpolation. In some embodiments of the invention, the correctly synchronized sample vectors are obtained not by calculation, but by adjusting the sampling times. One example of such an embodiment is described later in this specification, in conjunction with Figure 6. Figure 4 illustrates further one example of combining the symbol synchronization with demodulation for obtaining the actual transmitted data. Figure 4 shows a demodulator block 410, which receives as its input the interpolated I and Q sample vectors, and performs the demodulation based on the interpolated sample vectors. The input for the demodulator block 410 can be taken, for example, from the outputs of the adding blocks 440a, 440b as well, in which case the demodulation needs to be done on the basis of the differences of the received sample vectors.

A further advantageous way for demodulating the received data is to use the output of the difference vector length calculation block 445. An example of such an embodiment is described further in conjunction with figure 5.

Figure 5 illustrates a further advantageous embodiment of the invention. In this embodiment, the output of the difference vector length calculation block 445 is used to determine the transmitted data. The difference vector lengths are taken to a viterbi detector block 480, which detects the transmitted data on the basis of the difference vector lengths. Determination of the transmitted data on the basis of the difference vector lengths is described in further detail in a patent application entitled "Demodulation method", which application has the same date of filing as the present application by the same applicant. The ftmctioning of the synchronizer part 400 and the rest of the components in Figure 5 have been explained in connection with Figure 4, and need not be explained here in further detail.

In a further advantageous embodiment, the calculation of differences is performed before the interpolation, i.e. the interpolation is performed on the difference samples. In such an embodiment, the delay and adding blocks 435a,440a; 435b,440b would be located between the I and Q sample stream inputs of synchronization part 400 and the interpolating blocks 430a;430b.

Figure 6 illustrates a further advantageous embodiment of the invention, in which the timing of the sampling oscillator 220 is adjusted for symbol synchronization. The sampling oscillator 220 receives a control signal from the synchronizer part 400, which control signal adjusts the timing of the sampling oscillator 220. Figure 6 also illustrates such an embodiment of the invention, which implements interpolation according to equation (9) instead of oversampling.

In figure 6, the I and Q signal samples from the sampling part 200 are taken to a delay block 435a, 435b and an adding block 440a, 440b, which delay and adding blocks calculate the difference of consecutive I and Q samples. The delay blocks 435a, 435b delay the I and Q signals for the period of one T_b. The difference samples are taken to calculation block 460, which calculates early and late difference vectors by interpolation from three consecutive difference samples, and calculates the lengths of the early and late difference vectors. In other words, the calculation block calculates the two terms of equation (9), i.e. the length of the early difference vector

+ τ) + dA((k - l)T_b + r))| and the length of the late difference vector \(cA(kT_b + τ) + dA((k + l)T_b +

. The length of the early difference vector is outputted via output E and the length of the late difference vector via output L to an adding block 465, which subtracts the length of the early difference vector from the length of the late difference vector. The result is taken to a loop filter 455, which preferably averages the result. The averaged result is taken to sampling oscillator 220 for controlling the timing of the sampling oscillator. When the length of the early difference vector is greater than that of the late difference vector, the resulting control signal from loop filter 455 causes the sampling oscillator 220 to shift the sampling times to a later instant. Correspondingly, when the length of the early difference vector is smaller than that of the late difference vector, the resulting control signal from loop filter 455 causes the sampling oscillator 220 to shift the sampling times to an earlier instant. Naturally, some kind of signal processing such as scaling of the averaged result given by the loop filter 455 may be needed in order to obtain a control signal for controlling the oscillator 220 in the desired way. In the example of Figure 6, such signal processing is performed in the sampling oscillator block 220.

Rest of the components in the sampling part 200 and the functioning of the sampling part 200 as well as the demodulation block 410 in Figure 6 have been explained in connection with other figures, and need not be explained here in further detail.

Fig. 7 shows a block diagram of a digital mobile communication means according to an advantageous embodiment of the invention. The mobile communication means comprises a microphone 301, keyboard 307, display 306, earpiece 314, antenna duplexer or switch 308, antenna 309 and a control unit 305, which all are typical components of conventional mobile communication means. Further, the mobile communication means contains typical transmission and receiver blocks 304, 311. Transmission block 304 comprises functionality necessary for speech and channel coding, encryption, and modulation, and the necessary RF circuitry for amplification of the signal for transmission. Receiver block 311 comprises the necessary amplifier circuits and functionality necessary for demodulating and decryption of the signal, and removing channel and speech coding. The signal produced by the microphone 301 is amplified in the amplifier stage 302 and converted to digital form in the A/D converter 303, whereafter the the signal is taken to the transmitter block 304. The transmitter block encodes the digital signal and produces the modulated and amplified RF-signal, whereafter the RF signal is taken to the antenna 309 via the duplexer or switch 308. The receiver block 311 demodulates the received signal and removes the encryption and channel coding. The resulting speech signal is converted to analog form in the D/A converter 312, the output signal of which is amplified in the amplifier stage 313, whereafter the amplified signal is taken to the earpiece 314. The control unit 305 controls the functions of the mobile communication means, reads the commands given by the user via the keypad 307 and displays messages to the user via the display 307. In a mobile communication means according to the invention, the mobile communication means comprises a synchronizing part 400 performing symbol synchronization according to the invention. In some embodiments of the invention, the synchronizing part 400 advantageously has the structure shown in Figure 4. However, other structures for symbol synchronizing according to the invention can also be used.

Figure 8 shows an example of an embodiment of the invention. In the example of Figure 8, a synchronizing part 400 according to the invention is used in at least some base stations 360 of a mobile communication network for symbol synchronization of TFM signals received from the mobile communication means 350. Figure 8 further illustrates a base station controller 370 controlling the base stations 360 and two radio link units 371 for connecting the base station controller 370 to the rest of the mobile communication network 380. Figure 8 also illustrates a ftirther advantageous embodiment of the invention, namely the use of synchronizing parts 400 according to the invention in radio links. The demodulation method according to the invention is very suitable to be used in continuous high speed communications, where a continuous symbol synchronization, symbol synchronization tracking and data demodulation is needed. High speed radio links are one example of such an advantageous application of the invention.

Figure 9 shows a block diagram of an advantageous embodiment using phase difference instead of difference vector length as the decision variable. The structure of Figure 9 is based on the use of a early-late gate. The sensitivity of the timing error detector is best, i.e. the error signal is largest, if the time difference between the delayed and the advanced branches is the same as the distance between a minimum and a maximum in the average of expression (10). Since the distance of the extrema is T_bl2, one of the branches should be delayed by T_b/4 and the other advanced by T_bIA. In the case of TFM one sample per bit is enough for signal reconstruction: the delay operation by T_bIA and the advance operation by T IA can be done with linear interpolation even when the sampling rate is only one sample per bit.

Since the inventive method uses only phase differences or difference vector lenghts over T_b, the method can be used in the presence of a frequency error. One further benefit is, that the method can be used without the knowledge of bit decisions.

Although the present invention allows symbol synchronization in the presence of DC offset and other types of low frequency noise, the inventive symbol synchronization method can also be used in conventional TFM receiver structures, in which the amount of DC offset and low frequency noise is not disturbingly large. Further, although in the previous examples, the length of the difference of consecutive received sample vectors is used for obtaining synchronization, also the length of the difference over two consecutive bit intervals may be used for obtaining synchronization. Also, although the previous description describes synchronization to a TFM modulated signal, the inventive synchronization method can be used for obtaining synchronization to MSK, GMSK and GTFM modulated signals as well. In particular, the invention can be used with Viterbi type TFM receivers in digital microwave frequency radio links.

In the following claims, the term sample vector refers to a pair of corresponding I and Q signal samples.

In view of the foregoing description it will be evident to a person skilled in the art that various modifications may be made within the scope of the invention. While a preferred embodiment of the invention has been described in detail, it should be apparent that many modifications and variations thereto are possible, all of which fall within the true spirit and scope of the invention.

Claims

1. Symbol synchronizing method for synchronizing to the symbol timing of a tamed frequency modulated signal, characterized in that the method comprises steps, in which

- sample vectors are obtained from the signal according to a controlling parameter,

- parameters substantially dependent on phase differences of said sample vectors are calculated from said sample vectors,

- a function depending substantially on said parameters is calculated, and - said controlling parameter is adjusted for obtaining an extremum value of said function.

2. A method according to claim 1, characterized in that said parameter is the phase difference of consecutive sample vectors.

3. A method according to claim 1, characterized in that said parameter is the length of a difference vector calculated as the difference of two sample vectors.

4. A method according to claim 3, characterized in that said difference vectors are calculated from consecutive sample vectors.

5. A method according to claim 1, characterized in that said function comprises calculation of the average of said parameters.

6. Symbol synchronization system for synchronization to symbols of a tamed frequency modulated signal, characterized in that the system comprises

- means for producing signal sample vectors from the tamed frequency modulated signal according to a control signal,

- means for calculating parameters substantially dependent on phase differences of said sample vectors,

- means for calculating the value of a function depending substantially on said parameters, and - means for producing said control signal for adjusting the production of signal sample vectors in order to find an extremum value of said function.

7. A system according to claim 6, characterized in that the system further comprises means for calculating difference vectors from said sample vectors, and that said means for calculating parameters are arranged to calculate the length of said difference vectors.

8. A system according to claim 7, characterized in that said means for calculating the value of a function calculate the average of the length of said difference vectors.

9. Mobile communication means, characterized in that it comprises a symbol synchronization system, which system comprises

- means for producing signal sample vectors from the tamed frequency modulated signal according to a control signal, - means for calculating parameters substantially dependent on phase differences of said sample vectors,

- means for calculating the value of a function depending substantially on said parameters, and

- means for producing said control signal for adjusting the production of signal sample vectors in order to find an extremum value of said function.

10. Mobile communication means according to claim 9, characterized in that said symbol synchronization system further comprises means for calculating difference vectors from said sample vectors, and that said means for calculating parameters are arranged to calculate the length of said difference vectors.

11. Mobile communication means according to claim 10, characterized in that said means for calculating the value of a function calculate the average of the length of said difference vectors.

12. Base station of a mobile communication network, characterized in that it comprises a symbol synchronization system, which system comprises

- means for calculating parameters substantially dependent on phase differences of said sample vectors, - means for calculating the value of a function depending substantially on said parameters, and

13. A base station according to claim 12, characterized in that said symbol synchronization system further comprises means for calculating difference vectors from said sample vectors, and that said means for calculating parameters are arranged to calculate the length of said difference vectors.

14. A base station according to claim 13, characterized in that said means for calculating the value of a function calculate the average of the length of said difference vectors.

15. Radio link system, characterized in that it comprises a symbol synchronization system, which system comprises

16. A radio link system according to claim 15, characterized in that said symbol synchronization system further comprises means for calculating difference vectors from said sample vectors, and that said means for calculating parameters are arranged to calculate the length of said difference vectors.

17. A base station according to claim 16, characterized in that said means for calculating the value of a function calculate the average of the length of said difference vectors .