Apparatus for Cancelling Periodic Carrier Phase Jitters

The present invention relates to a data transmission system for transmitting data over telephone lines or like analog lines and, more particularly, to a jitter cancelling apparatus for carrier phase control associated with a receiver of such a data transmission system.

In a data transmission system, signals usually undergo various kinds of deterioration as typified by amplitude distortions, delay distortions, carrier frequency offsets and carrier phase jitters while being transmitted over lines. Among them, amplitude distortions and delay distortions are almost time-invariant or, if time-variant, the variation is slow enough to allow such distortions to be compensated for by so-called automatic equalizers. Carrier phase jitters, on the other hand, result in time-variant distortions and this kind of distortion has hitherto been absorbed by a phase locked loop or like feedback control system. However, since the carrier phase jitters develop in, for example, a carrier supply device of an analog transmission link and include periodic components approximating the ac (alternate current) cycle of 50 or 60 Hz of a commercial power source, the phase locked loop will fail to absorb 60 Hz jitters under a baud rate of 2400 Hz unless its equivalent quality factor is limited to less than 40. This deteriorates the Gaussian noise suppression ability of the phase locked loop.

It is, therefore, an object of the present invention to provide a jitter cancelling apparatus which accomplishes considerable jitter suppression without sacrificing the noise suppression ability of a conventional phase locked loop.

According to one aspect of the present invention, there is provided a jitter canceller which comprises means for detecting a carrier phase of a complex baseband signal, storing means for storing detected carrier phases; phase predicting means responsive to said detected carrier phases for producing a predicted carrier phase; means for compensating for a carrier phase deviation of said complex baseband signal in response to said predicted phase; error detecting means for detecting an error associated with data descriminated from said complex baseband signal; and means for controlling said phase predicting means to minimize said error.

The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description taken with the accompanying drawings in which:

FIG. 1 is a block diagram showing a first embodiment of the present invention;

FIGS. 2-6 and 9 are block diagrams showing various parts of the present invention in detail;

FIG. 7 is a block diagram of a data receiver to which the jitter canceller of FIG. 1 is applied;

FIG. 8 is a block diagram showing a second embodiment of the present invention;

FIG. 9 is a block diagram of the adaptive digital filter of FIG. 8; and

FIG. 10 is a block diagram of a data receiver to which the jitter canceller of FIG. 8 is applied.

In the drawings, the same or similar structural elements are designated by like reference numerals; thick lines represent complex signals and thin lines, real signals or control signals.

Before entering into a detailed description of the present invention, the principle thereof will be described.

Briefly, the principle of the present invention resides in predicting a carrier phase jitter, which is a cause of signal deterioration, and removing it based on the result of the prediction.

Assume that signal processing in a receiver of an intended data transmission system is performed by a sample value system whose period is T.sub.s seconds, and that an instantaneous phase observed at a time m is y.sub.m. The instantaneous phase is represented by a sum of a true phase x.sub.m and noise n.sub.m introduced in the observation system:

Assuming that an estimated phase provided by linear weighted addition of observation sequences {ym} is x.sub.k, it is expressed as: ##EQU1## where a.sub.0, a.sub.1, . . . , a.sub.N-1 are prediction coefficients, and N is a prediction order. The prediction accuracy of an estimated phase increases with N as will be described later. The difference between the suffixes k and m reflects a difference in time between a point of phase observation and a point of verifying correctness of an estimated phase.

The prediction coefficients a.sub.0, a.sub.1, . . . , a.sub.N-1 have to be so adjusted as to accomplish as correct an estimated phase as possible and, for this purpose, correctness of the estimated phase x.sub.k has to be verified in one way or another. If the true phase x.sub.k is known, it will suffice to predetermine the prediction coefficients a.sub.0, a.sub.1, . . . , a.sub.N-1 such that a mean value of a square (x.sub.k -x.sub.k).sup.2 of an estimation error becomes minimum. However, if the true phase x.sub.k is known, estimating it is needless from the beginning; usually, x.sub.k is unknown. It is impossible, therefore, to employ the mean of (x.sub.k -x.sub.k).sup.2 as an evaluation function for determining the prediction coefficients. Fortunately, however, a discriminated error Z.sub.k included in discriminated data after the phase compensation using an estimated phase x.sub.k is substantially

where C is a constant, and .alpha..sub.k is noise which has no correlation with x.sub.k or x.sub.k. In this instance, a mean square value J of the error Z.sub.k is produced by

where A is a power of .alpha..sub.k. It follows that if the prediction coefficients a.sub.0, a.sub.1, . . . , a.sub.N-1 are predetermined to minimize the mean square value of the discrimination error, the resulting prediction coefficients will be those which make the mean value of (x.sub.k -x.sub.k).sup.2 minimum.

Hence, if prediction coefficients a.sub.0 (k+1), a.sub.1 (k+1), . . . a.sub.N-1 (k+1) at a time (k+1) are adjusted one after another in relation to prediction coefficients a.sub.0 (k), a.sub.1 (k), . . . a.sub.N-1 (k) at a time k using an equation shown below, prediction coefficients which provide a minimum (x.sub.k -x.sub.k).sup.2 are obtainable: ##EQU2## where .epsilon. is an adjustment coefficient and 1/.epsilon. denotes a period of time for averaging Z.sup.2.sub.k. The above equation may be rewritten as: ##EQU3## This kind of adjustment algorithm is generally known by the name of a method of steepest-descent. Convergence of the algorithm is insured so long as the correlation matrix .phi. as defined by the following equation is a positive constant: ##EQU4## where R.sub.yy (1)=y.sub.m .multidot.y.sub.m+1.

Fundamentally based on the above-described principle, the present invention contemplates to removal of carrier phase jitters by rotating the phase of a received complex baseband signal by a predicted phase.

Assuming that the k-th sample value of a complex baseband signal is .gamma..sub.k, and that a complex signal provided by rotating the phase thereof by x.sub.k is .eta..sub.k, there holds a relation:

If .eta..sub.k is a complex signal resulting from demodulation of a VSB (vestigial-sideband modulation) signal, its real part carries desired data and, hence, the discrimination error Z.sub.k is produced by

where R.sub.e {.multidot.} represents an operation for obtaining the real part while d.sub.k represents discriminated data. Therefore, ##EQU5## where Im {.eta.} represents a signal associated with an imaginary part of .eta..sub.k. The adjustment algorithm for prediction coefficients may thus be written as:

What the equation (1) implies is as follows. The i-th prediction coefficient at the time (k+1) is provided by subtracting from the i-th prediction coefficient at the time k a product of a quantity of adjustment which is multiplied by a suitable adjustment coefficient. That is, the i-th quantity of adjustment at the time k is provided by multiplying a common quantity, which is a product of a discrimination error Z.sub.k at the time k and an imaginary part at the time k, by y.sub.m-i which is an output of the i-th stage of a register.

Meanwhile, if .eta..sub.k is a complex signal resulting from demodulation of an ordinary QAM (quadrature-amplitude modulation) signal, then a prediction coefficient adjustment algorithm corresponding to the equation (1) may be expressed as:

where Z.sub.R,k and Z.sub.i,k represent respectively a real part discrimination error and an imaginary part discrimination error, respectively. The equations (1) and (2) may be generalized as an adjustment algorithm:

k m

Here, .xi..sub.k is regarded as a signal associated with a phase compensation error at the time k.

Referring now to FIG. 1 of the drawings, a jitter canceller embodying the present invention and based on the principle described above is shown in a block diagram. The jitter canceller, generally 100, comprises aninput terminal 101 to which a complex baseband signal is applied, a phase detector 704 for detecting an instantaneous phase of the complex baseband signal and a register 103 to which instantaneous phases are sequentially stored. The first stage output of the register 103 is weighted by a prediction coefficient at, for example, a first weighting section 111, the second register stage output at a second weighting section 112, and the final register stage output at the final weighting section 113. The respective weighted results are added up by an adder 108 to provide a predicted value x.sub.k of an instantaneous phase. A complex trigonometric function generator 109, in response to the predicted value x.sub.k, produces a complex signal exp(-jx.sub.k) which comprises a real part cos x.sub.k and an imaginary part -sin x.sub.k. A phase rotator 104 multiplies a complex signal applied thereto from the input terminal 101 by the exp (-jx.sub. k) supplied from the complex trigonometric function generator 109, applying the product to an error detector 106. The error detector 106 functions to discriminate received data in conformity to a specific modulation system such as VSB or QAM while providing a discrimination error signal associated therewith. The error signal is immediately fed back to correlators 121, 122 and 123. Each of the correlators 121-123 computes a correlation between the error signal and the output of the register stages of register 103 associated therewith, thereby adjusting the prediction coefficient in the associated weighting section 111, 112 or 113 to reduce the correlation.

Various schemes may be contemplated for constructing the correlators 121, 122 and 123 shown in FIG. 1 and one example is shown in FIG. 2. In FIG. 2, the register output is applied to a terminal 201 of the correlator and the error signal to a terminal 202, and they are led to a multiplier 203 to provide a product. The resulting product is supplied to an integrator, which is made up of an adder 204, a first weighting section 205 and a one-sample delay 206, to be thereby averaged with respect to a given sampling time. At this instant, assuming that the weighting coefficient of the first weighting section 205 is .beta.(.beta.<1l), the integrating time is determined by 1/(1-.beta.). The integrated signal is multiplied by a suitable coefficient by a second weighting section 207 and, then, applied to an output terminal 208 as an adjusting signal for the previously mentioned prediction coefficient.

Another specific structure of the correlators shown in FIG. 1 is shown in FIG. 3. The correlator of FIG. 3 comprises signal detectors 303 and 304, an Exclusive-OR gate (EXOR) 305 and a selector 306. The signal detectors 303 and 304 detect respectively a polarity of the output of the register 103 given from an input terminal 301 and that of the error signal supplied from an input terminal 302, and the detected polarities are applied to the EXOR 305. The EXOR 305 supplies a control terminal of the selector 306 with a logical "1" if the detectors 303 and 304 share the same polarity and a logical "0" if otherwise. In response to the logical level of the control signal, the selector 306 selects a positive value +.DELTA. or a negative value -.DELTA. and applies its output to an integrator, which comprises an adder 307, a weighting section 308 and a one-sample delay 309. The integrator smooths the output of the selector 306 to produce an adjusting signal for the prediction coefficient, the adjusting signal being sent out from an output terminal 310.

The structure of the error detector 106 shown in FIG. 1 depends upon the modulation system used. In the case where a VSB modulation system is used, the error detector 106 may be constructed as shown in FIG. 4 by way of example. In FIG. 4, a real part and an imaginary part of a complex baseband signal are applied to input terminal 401 and 402, respectively. A discriminator 403 discriminates data carried by the real part, while an adder 404 detects discrimination error. The error signal is multiplied by the imaginary part in a multiplier 405 to produce a phase compensation signal to an output terminal 406.

Where the modulation system is QAM, on the other hand, the error detector 106 may be constructed shown in FIG. 5. In FIG. 5, real and imaginary parts of a complex baseband signal are entered through input terminals 501 and 502, respectively. Discriminators 503 and 504 respectively discriminate data which are carried by the real and imaginary parts of the complex baseband signal. Adders 505 and 506 serve to detect discrimination errors in the real and imaginary parts of the baseband signal, respectively. A multiplier 507 multiplies the error in the real part of the baseband signal by the imaginary part of the baseband signal, while a multiplier 508 multiplies the error in the imaginary part of the baseband signal by the real part of the baseband signal. The two different products are added by an adder 509 to produce a phase compensation signal, which is applied to an output terminal 510.

A specific construction of the error detector 106 usable with stagger QAM is shown in FIG. 6. In stagger QAM, real and imaginary parts of a complex baseband signal, which arrive at input terminals 601 and 602, alternately carry data each by one sample. The error detector 106 includes a first selector 603 for selecting one of the two signals which carries data, while applying the selected signal to a discriminator 605. Meanwhile, a second selector 604 selects the other of the two signals which does not carry data, while applying the selected signal to a multiplier 607. A subtractor 606 detects discrimination error which is then led to the multiplier 607. The multiplier 607 multiplies the two signals by each other to deliver a phase compensation signal to an output terminal 608.

Referring to FIG. 7, a basic construction of a data receiving system to which the specific jitter canceller described above is applied is shown in a block diagram. A complex baseband signal supplied to an input terminal 701 is applied to a phase control loop made up of a phase rotator 702, a phase detector 704 and a voltage controlled oscillator 703, whereby phase variations sufficiently slower than the sampling rate are absorbed. The output of the phase control loop is fed to an ordinary automatic equalizer 705 so as to equalize amplitude distortions and delay distortions caused by a transmission line. Therefore, the output of the equalizer 705 contains only jitters in that the phase variation is fast with respect to time. The jitters are absorbed by the jitter canceller 100 in accordance with the present invention. A discrimination error signal necessary for the adjustment of tap coefficients of the equalizer 705 is provided by the jitter canceller 100. It follows that the tap coefficients of the equalizer 705 are free from fluctuation due to jitters, insuring the expected equalizing function. An instantaneous phase to be stored in the register of the jitter canceller is provided by the phase detector 704. A certain delay exists between the complex baseband signal applied to the jitter canceller 100 and the instantaneous phase contained in the baseband signal for predicting a jitter. However, such a delay does not bring about noticeable deterioration to jitter cancelation characteristics of the jitter canceller because, usually, jitters are mostly a periodic component.

Referring to FIG. 8, a second embodiment of the present invention is shown in a block diagram. This embodiment is suitable for use with parallel data transmission which employs a plurality of carriers. In FIG. 8, the jitter cancalling system comprises an input terminal 801 to which a pilot signal is applied, and input terminal 802-804 to which first to third complex baseband signals are applied, respectively. An adaptive filter 805 extracts high-speed periodic phase jitters having a large signal-to-noise (S/N) ratio out of the pilot signal, applying the jitters as a reference jitter to jitter cancellers 806-808. The jitter cancellers 806-808 remove the high-speed periodic phase jitters from the respective complex baseband signals. The order of the adaptive filter 805 is selected such that even with the worst S/N ratio predicteable in the pilot signal there can be supplied a reference jitter whose S/N ratio is larger enough for the jitter cancellers 806-808 to accomplish their predetermined function. Each of the jitter cancellers 806-808 may be implemented by the structure shown in FIG. 1.

A specific construction of the adaptive filter 805 for reference extraction is shown in FIG. 9. In FIG. 9, a phase detector 902 detects the instantaneous phase of the pilot signal and applies it to the shift register 103 via a delay 903. Outputs of the respective stages of the shift register 103 are weighted by their associated weighting sections 111-113 by prediction coefficients applied from the correlators 121-123. The respective weighted results are added up by the adder 108 an output of which is led to the complex trigonometric function generator 109 as a predicted instantaneous phase x.sub.k and to the jitter cancellers 806-808 as the reference jitter. In the structure of FIG. 9, the delay 903 serves to prevent the correlation of noise components from being reflected by the weighting sections 111-113.

Referring to FIG. 10, there is shown in a block diagram a data receiving system to which the jitter cancelling system of FIG. 8 is applied. A separating circuit 1001 separates received data into a pilot signal and respective complex baseband signals associated with the individual data channels. The pilot signal is supplied to the adaptive filter 805 assigned to a reference jitter, while the respective baseband signals are coupled to pre-processors 706 which are provided in one-to-one correspondence with the data channels. Each of the pre-processors 706 comprises an equalizer 705 for equalizing amplitude distortions and delay distortions, and a phase locked loop for absorbing slow phase variations which is made up of a voltage controlled oscillator 703, a phase detector 704 and a phase rotator 702. The circuitry shown in FIG. 10, like the circuitry shown in FIG. 7, is capable of removing slow phase variations and fast periodic jitters in the respective channels.

As described above, it will be seen that the present invention provides a jitter canceller which suppresses faster phase variations after sufficiently slow phase variations are suppressed by a phase control loop.