Carrier Frequency Synchronization Device Using Two Different B/W Filters

This invention relates to a carrier synchronisation device for determining and compensating for differences in synchronisation between a local carder and a received modulated signal to be demodulated by the local carrier generated by said device, the device comprising:

means for determining an error signal .epsilon.(t) representative of a synchronisation error between the received modulated signal and the local carrier,

a local oscillator which generates the local carrier under control of a control signal u(t),

at least two low-pass filters connected in parallel between said means and the local oscillator, the two filters filtering the error signal and providing different loop bandwidths and

a lock mode detector, which supplies a selection signal for selecting one of the filters, the output of the selected filter supplying said control signal u(t).

The invention also relates to the use of this device for the synchronisation of signals transmitted in accordance with an OFDM (Orthogonal Frequency Division Multiplex) standard.

Such a synchronisation device is described, for example, in the JP KOKAI-60 1673 1. That document describes a circuit formed by a phase-locked loop which determines an error signal representative of a phase error, which error signal can be filtered either by a first filter or by a second filter, which filters have different pass bands, in a manner such that the error signal thus filtered produces a control signal which controls a local oscillator. The signal generated by this local oscillator is combined with the received signal in order to generate the error signal. A discriminator circuit detecting a lock-in mode or a capturing mode determines which of the two filters is to be put into operation.

How the discriminator circuit operates is not revealed. It is apparent only that the device performs a phase detection and that the discriminator circuit selects the mode of operation by a phase-related measurement.

However, there are also situations in which allowance is to be made for frequency differences. On the other hand, the discriminator circuit is fixed and cannot adapt itself automatically to operation in the capturing mode or in the lock-in mode.

A special use of the device is in the reception of signals transmitted in accordance with an orthogonal frequency division multiplex (OFDM) method, transmitted particularly by electromagnetic waves, by cable or by satellite, for example, for use in digital television. For uses such as the transmission of television pictures by OFDM modulation, phase detection is ineffective. Moreover, for large-scale uses the device should be not only robust and accurate but also inexpensive.

The technique of transmitting N orthogonal frequencies by frequency division multiplexing consists of splitting up the information to be transmitted by dividing it over a large number of elementary channels having a low bit rate 1/T, T being the bit period. Thus, a highly selective wide-band channel is converted into a large number of non-selective elementary channels. Since these elementary channels together form a wide-band channel, fading during transmission is unlikely to affect the entire channel simultaneously. This technique also enables intersymbol interference to be reduced.

Each elementary channel has a corresponding frequency, the frequencies as a whole being symmetrical about a carrier frequency. Since the use of selective filters during reception is difficult, it is preferred to allow the spectra to overlap but requirements are imposed on the orthogonality between the frequencies in order to eliminate intersymbol interference at the sampling instants. The complete spectrum of an OFDM signal will resemble a rectangular spectrum.

During reception the received signals should be demodulated and subsequently be decoded in order to recover the original information. For this demodulation a local oscillator is used whose frequency should be locked to the received carrier frequency.

It is an object of the invention to enable a locally generated carrier to be synchronised with a received modulated signal by means of the carder synchronisation device, synchronisation being effected with loop time constants which are adapted to the modes of operation in an automatically programmable manner.

This object is achieved with a synchronisation device in which, in order to supply the selection signal, the lock mode detector calculates differences between the control signal and at least one replica of said control signal delayed by a programmable delay means, which delay means produce specific delays for each filter, the differences being compared with at least one predetermined threshold by comparison means in order to supply the selection signal which determines the filter to be selected.

Thus, said device advantageously reduces frequency differences between the received modulated signal and the local carrier by the use of a control signal u(t) (filtered error signal) which is proportional to the frequency. The device switches from one filter to the other in an automatically programmable manner.

Preferably, said means for determining an error signal comprise either a frequency detector or a phase detector, which measures the synchronisation error by, respectively, calculating the frequency difference or the phase difference between the received signal and the local carder.

This device is particularly interesting in order to ensure that the reception frequency of a receiver is locked to the transmission frequency. For this purpose it is necessary to switch from a capturing mode to a lock-in mode and vice versa depending on the transmitting conditions obtaining in the transmission channel. To this end the delay means apply a first delay and select a first filter providing a large loop bandwidth, in order to cause said device to operate in the capturing mode, and the delay means apply a second delay smaller than the first delay and select a second filter providing a narrower loop bandwidth, in order to cause said device to operate in the lock-in mode.

The device can be used for the synchronisation of block-format OFDM signals of which at least one block is a repetitive synchronisation block. In a special embodiment of the invention the frequency detector determines the error signal by determining a synchronisation error of the received synchronisation blocks. The first delay and the second delay may then be equal to an integral multiple of the repetition rate of the synchronisation blocks.

This has the advantage that an improved performance of the device in the capturing mode can be obtained.

These and other aspects of the invention will be apparent from and elucidated with the aid of the embodiments described hereinafter.

The invention will be more fully understood with the aid of the accompanying drawings, which is given by way of non-limitative examples and of which:

FIG. 1 shows diagrammatically a receiver which comprises a carder synchronisation device in accordance with the invention.

FIG. 2 is a diagram of a detector for detection of the capturing/lock-in mode.

FIG. 3 is a diagram showing an example of the programmable delay means.

FIGS. 4 and 5 are diagrams of analog and digital low-pass filters, respectively.

FIG. 1 shows diagrammatically a receiver 10 comprising a demodulator 16 and a carrier frequency synchronisation device 12. A mixer 13 combines the modulated signal r(t) received by the demodulator 16 with a carrier generated locally by a local oscillator, for example, a voltage-controlled oscillator VCO 20. The output of the mixer 13 is a baseband signal, which is applied to a sampler 17 (sampling rate T/N) and a demultiplexer 18, which supplies a demodulated sampled signal s(kT/N). To lock the oscillator frequency to the frequency of the received modulated signal, use is made of the synchronisation device 12, which operates in a loop. For this purpose it receives data from the demultiplexer 18, from which data synchronisation error information is extracted for controlling the carrier frequency generated by the local oscillator 20.

The synchronisation error is determined, for example, by means of a frequency detector FD 22, which determines the frequency difference between r(t) and the local-oscillator frequency and generates an error signal .epsilon.(t). A frequency detector 22 suitable for OFDM signals is described in, for example, Patent Application WO 92/16063. The error signal is applied in parallel to two low-pass filters Hf.sub.1 and Hf.sub.2 whose outputs can be selected to supply a control signal u(t). A lock mode detector LID 25 is used for determining the instant at which it is necessary to switch over from the filter Hf.sub.1 (capturing mode) to the filter Hf.sub.2 (lock-in mode) and vice versa. The lock mode detector LID 25 effects this change-over by means of a selector 26. The detector 25 employs the control signal u(t) to supply a selection signal SEL 27, which selects one of the filters.

FIG. 2 shows a diagram of the detector 25. It comprises programmable delay means 30, which produce a delay D, and a subtracter 32, which each receive the control signal u(t). The subtracter calculates differences between the control signal u(t) and a replica of the same signal delayed by the means 30. A comparator 34 compares the differences with a threshold Th. The comparator supplies a selection signal SEL, which controls the selector 26 connected to the outputs of the filters. Alternatively, the selector 26 may precede the filters. The selection signal SEL is also applied to the delay means 30 to select a specific delay for each filter and to make the device automatically programmable.

In the lock-in mode the delay D is set to a small value to insure that the synchronisation device responds rapidly to frequency changes. Likewise, the delay D is increased at the beginning of each capturing period in order to mitigate the problem of erroneous locking.

The filter characteristics should be selected so as to obtain an acceptable capture time and a minimal error variance .sigma..sup.2 and zero average error under stable frequency conditions. An acceptable capture time is a compromise between a capture time of minimal length and minimal jitter.

The capture time Tacq may be defined as the time necessary to reduce the normalised frequency error to an average value smaller than 0.02. For an OFDM signal it is possible to choose, for example, a filter Hf.sub.1 which provides a second-order loop having a normalised pass band B.sub.l.T.sub.f =2 (T.sub.f being the interval between two synchronisation blocks) and an attenuation factor of 20. In the case of additional Gaussian noise this yields a capture time of the order of 15 synchronisation blocks for an initial frequency error of .DELTA.f.T.sub.s =5 (where .DELTA.f=31.25 kHz and a symbol length T.sub.s =160 microseconds). At the end of the locking period the frequency stability is determined by the variance of .DELTA.f.T.sub.s. For an S/N ratio=30 dB this yields .sigma..sup.2 =9.multidot.10.sup.-5 and for S/N=13 dB this yields .sigma..sup.2 =1.5.multidot.10.sup.-4. However, during the locking period the filter Hf.sub.1 alone cannot correctly cope with the problems resulting from jitter and the presence of a residual frequency error which persists for a long time. To reduce the jitter in the lock-in mode the filter is replaced by a second filter Hf.sub.2 which provides a loop whose pass band is narrower than that provided by the filter Hf.sub.1. The filter Hf.sub.2 may provide, for example, a first-order loop such that the normalised bandwidth is B.sub.l.T.sub.f =0.25. At the end of the locking period the filter Hf.sub.2 yields a frequency stabilisation having a variance .sigma..sup.2 =2.multidot.10.sup.-8 for S/N=30 dB and a variance .sigma..sup.2 =3.multidot.10.sup.-6 for S/N=13 dB.

This shows that the frequency stabilisation in the locking period has improved when the filter Hf.sub.2 is used.

The filters Hf.sub.1 and Hf.sub.2 may be of an analog or of a digital type. It is possible to provide more than two filters.

FIG. 3 shows a diagram of an example of the delay means 30. The means comprise a delay element 40 such as a shift register (in a digital version) or a delay line (in an analog version). Such delay means are programmed to provide a delay with the aid of a selector 42, which selects either a delay D1 or a delay D2. The selector is controlled by the selection signal SEL.

The delay values D1 and D2 depend on the application for which the synchronisation device is used. In the present example of OFDM, the parameters of the filters Hf.sub.1 and Hf.sub.2 being as specified above, the delay D has been programmed, for example, to change from 3 Tf to Tf, corresponding to the capturing mode and the lock-in mode, respectively (Tf=one OFDM frame period). A value of 0.8 has been selected for the threshold value Th.

FIG. 4 shows examples of analog filters. In FIG. 4A an amplifier having a gain -A receives the input signal .epsilon.(t) via a resistor R. The input of the amplifier is coupled to its output by a feedback capacitance C. The filtered signal u(t) is available relative to ground G on the output of the amplifier. When the synchronisation device 12 (FIG. 1) is used in conjunction with the filter shown in FIG. 4A, the loop will be of the first order. The loop will be of the second order with a filter as shown in FIG. 4B. The difference with respect to FIG. 4A is that a resistor R2 is arranged in series with the capacitance C and that the values of the elements R1 and C are adjusted.

The filters Hf.sub.1, Hf.sub.2 can be implemented as digital filters. The filters shown in FIGS. 5A and 5B are the digital counterparts of the filters shown in FIGS. 4A and 4B, respectively. In the present case the signals are digitized signals .epsilon..sub.k and u.sub.k. A summing element adds the signal .epsilon..sub.k to a delayed replica of the output signal of the summing element, which replica is supplied by the delay means 52. The output of the summing element is also coupled to the input of a multiplier 54, which multiplies the output signal of the summing element by a weighting coefficient c so as to form the filtered signal u.sub.k. The filter in FIG. 5A serves to construct a first-order loop.

FIG. 5B shows a filter by means of which a second-order can be obtained. It also comprises a summing element 50, a delay means 52 and a multiplier 54, the elements being arranged in the same way as in FIG. 5A. The multiplier 54 multiplies the output signal of the summing element by a weighting coefficient b. The input of the summing element receives the signal .epsilon..sub.k, which is also applied to a multiplier 56 (weighting coefficient a), whose output is connected to a summing element 58, which also receives the output signal of the multiplier 54. The filtered signal u.sub.k is available on the output of the summing element 58.

A filter Hf.sub.1 which provides a second-order loop is used in the capturing mode and a filter Hf.sub.2 which provides a first-order loop is used in the lock-in mode.