MXPA96004225A - Phase detector in a carrier recovery network for a vestigial sideband signal - Google Patents

Abstract

A television signal receiver for processing an HDTV signal transmitted in a vestigial sideband (VSB) format with a one-dimensional data constellation includes a first carrier recovery network (18), an equalizer (20), and a second carrier recovery network (22, 30, 62). A multiple stage quantizer network (50, 66) exhibiting progressively finer resolution is associated with the operation of the equalizer for providing blind equalization without need of a"training"signal. The second carrier recovery network includes a phase detector (30) wherein a one-symbol delayed (312) input signal and a quantized (310) input signal are multiplied (316), and an unquantized input signal and a quantized (310) one-symbol delayed (314) input signal are multiplied (318). Signals produced by the multiplication are subtractively combined (320) to produce an output signal representing carrier phase error.

Description

PHASE DETECTOR IN A CARRIER RECOVERY NETWORK FOR A VESTIGIAL SIDEBAND SIGNAL Field of the Invention The present invention relates to a digital signal processing system. In particular, the invention relates to a phase detector in a carrier recovery network for a vestigial sideband (VSB) signal, such as can be modulated with high definition television (HDTV) information, for example.

BACKGROUND OF THE INVENTION The recovery of data from a vestigial sideband signal or Quadrature Modulated Amplitude (QAM) in a receiver, requires the implementation of three functions: time recovery for symbol synchronization, carrier recovery (frequency demodulation), and compensation. Time recovery is the process by which the receiver clock is synchronized (time base) with the transmitter clock. This allows the received signal to be sampled at the optimal time point to reduce the chance of a slicing error associated with the processing directed to the decision of the values of the received symbols. Carrier recovery is the process by which a received radiofrequency signal, after the frequency is changed to a lower intermediate frequency pass band, near the baseband, is finally changed in its frequency to the Baseband to allow the retrieval of information from the modulating baseband. Compensation is a process that compensates for the effects of the alterations of the transmission channel on the received signal. More specifically, the compensation removes the interference between symbols (ISI) of the baseband, caused by the alterations of the transmission channel, including the low-pass filtering effect of the channel. The interference between symbols causes the value of a symbol given by the values of the previous and following symbols to be distorted. For quadrature-modulated amplitude signals, time recovery is usually the first function implemented in a receiver. The time is recovered either from the intermediate passband signal or from a signal near the baseband, i.e., a baseband signal with a carrier phase shift that is corrected by a carrier recovery network. In any case, the time can be set before demodulation of the baseband. The carrier recovery demodulation process is usually a two-step process. First, the passband signal is changed up to almost the baseband by a frequency changer that uses a "better guess" on what frequency offset is between the input passband signal and the desired baseband signal. This frequency change is usually made by analog circuits; that is, before the analog-to-digital conversion in the receiver. Next, compensation is made on this signal near the baseband. Finally, the carrier recovery is performed, which removes any residual frequency offset from the signal near the baseband to produce a true baseband output signal. This function is performed by digital circuits of the receiver. The compensator is inserted between a first local oscillator that performs the change up to near the baseband and the carrier recovery cycle network. This is because the carrier recovery process is typically a decision-driven process (as it is known) that requires at least one partially open "eye" that is provided by the compensator function. A Quadrature Modulated Amplitude signal that transports digital information is represented by a constellation of two-dimensional data symbols defined by the Real and Imaginary axes. In contrast, a vestigial sideband signal is represented by a constellation of one-dimensional data symbols where only one axis contains quantized data to be retrieved at a receiver. The synchronized demodulation of a vestigial sideband signal has normally been carried out with the aid of a pilot signal. The pilot signal facilitates the demodulation of the vestigial sideband signal to the baseband in one step, typically without residual phase or frequency errors. The performance of time recovery, demodulation, and compensation functions in the order in which they are performed for quadrature modulated amplitude signals does not work for vestigial sideband signals using conventional techniques. For quadrature-modulated amplitude signals, various time recovery methods are known which are independent of the frequency offset between the signal near the baseband and the baseband signal. However, it is generally accepted that recovery of time regardless of frequency for vestigial sideband signals is not feasible. For this reason, in vestigial sideband systems, absolute demodulation has been historically implemented up to the baseband. An example of a vestigial sideband system that includes a pilot component is the Grand Alliance high definition television transmission system recently proposed for the United States. This system employs a vestigial sideband digital transmission format to transport a packet of data stream, and is being evaluated in the United States by the Federal Communications Commission through its Advisory Committee of Advanced Television Service (ACATS). A description of the Grand Alliance high definition television system as submitted to the ACATS Technical Subgroup on February 22, 1994 (project document) is in the 1994 Proceedings of the National Association of Broadcasters, 48th Annual Broadcast Engineering Conference Proceedings , March 20-24, 1994. The carrier recovery network in an 8-VSB system, such as in the transmission system used by the Grand Alliance, may not be able to track carrier errors in the presence of carrier noise. moderate phase, such as may be associated with the oscillators found in the tuners of consumer grade receivers. This requires the use of a secondary carrier recovery network to remove residual phase noise. In digital passband communications systems, the carrier is typically tracked with an Assured Phase Cycle (PLL). A phase detector is an important part of the assured phase cycle. The phase detector determines the amount of phase correction needed, and produces a corresponding error signal which, when multiplied with the input signal, demodulates the signal to the baseband. A decision-directed phase detector (DDPD) is often used in quadrature-modulated amplitude systems. This phase detector measures the phase angle as the angular error between the input signal and a quantized version of the input signal. In a quadrature-modulated amplitude signal, both phase and quadrature components carry data symbols, and both components are created using Nyquist filtering in a transmitter. This explains that, when there is no intersymbol interference (ISI) present, the expected values of both in-phase and quadrature components of the quadrature-modulated amplitude signal are separate quantizer values. Accordingly, a decision-directed phase detector can accurately measure the phase error by measuring the angular difference between the decisions of the quantizer and the input signal from the compensator to the quantizer. This technique is not directly applicable to vestigial sideband signals. Unlike a quadrature-modulated amplitude signal, in a complex vestigial sideband signal, only the phased component (the Real component of the complex signal) has been subjected to Nyquist filtering in the transmitter. The quadrature component (the Imaginary component) undergoes vestigial side-band filtration, which is generally not Nyquist filtration. The implication of this leak is that, even in the absence of inter-symbol interference, the quadrature channel contains a continuum of expected values at the optimum sampling point for the in-phase component. The use of a conventional decision-directed phase detector (such as can be used in a quadrature modulated amplitude system) is not applicable to a vestigial sideband system, since there are no separate symbol values, for example, Quantified samples, available from the quadrature channel. In accordance with the foregoing, a phase detector is described herein that is particularly useful in the carrier recovery network of vestigial sideband data communications systems.

SUMMARY OF THE INVENTION In accordance with the principles of the present invention, a phase detector suitable for use in a carrier recovery network of vestigial sideband signal processing systems is disclosed. The phase detector includes a quantizer, a delay network for providing delayed symbols, and first and second multipliers for producing first and second symbols in response to combinations of quantized and delayed input symbols. The first and second signals combine to produce a phase error signal. In a preferred embodiment described, the phase detector responds to a real only vestigial sideband signal. The phase error signal is applied to a control input of a compensator that is preceded on the signal line by another carrier recovery network.

BRIEF DESCRIPTION OF THE DRAWINGS In the drawings: Figure 1 is a block diagram of a portion of an advanced television receiver, such as a high definition television receiver, which includes a phase detector apparatus in accordance with the principles of the invention. Figure 2 is a block diagram of another embodiment of a receiver system that includes a carrier recovery phase detecting apparatus in accordance with the principles of the present invention. Figure 3 shows the details of a phase detector circuit in accordance with the principles of the present invention. Figure 4 illustrates a symbol decision process associated with a blind compensation process described in relation to Figures 1 and 2.

Detailed Description of the Drawings In Figure 1, an analogue high definition television signal modulated in vestigial sideband broadcasting received by an antenna 10, is processed by an input network 14 that includes the radio frequency tuning circuits, a tuner double conversion to produce an intermediate frequency pass band output signal, and appropriate gain control circuits, for example. The received vestigial sideband signal is illustratively an 8-VSB signal with a symbol rate of 10.76 Msymbols / second, occupying a 6 MHz frequency spectrum of conventional NTSC, in accordance with the Grand Alliance high definition television specification. . The Nyquist bandwidth for this system is 5.38 MHz, with an excess bandwidth of 0.31 MHz on each band edge. The passband output signal from the input processor 14 is converted from the analog form to the digital form by an analog-to-digital converter 16, which operates at a sample rate of 2 samples / symbol, for example. The vestigial sideband signal received in this example does not include a pilot component or a training component, and has been processed by unit 14, such that the center of the 6 MHz band is nominally 5.38 MHz. frequency spectrum of this signal at the input of analog-to-digital converter 16, occupies a scale of 2.38 MHz to 8.38 MHz. When the time synchronization has been established by means of a time recovery network 17, the converter unit from analog to digital 16 samples this signal at 21.52 MHz, which is twice the symbol rate. The time recovery network 17 provides an output symbol clock (CLK) that is synchronized with a corresponding clock generated in a transmitter. The CLK clock is applied to the analog-to-digital converter unit 16 and other elements of the receiver system. The techniques to achieve the recovery of time are well known. A particularly convenient time recovery technique suitable for use by the network 17 is described in a pending United States Patent Application Serial Number (RCA 87)., 588) by C. Strolle et al, entitled Carrier Independent Timing Recovery System for a Vestigial Sideband Modulated Signal. In the system to be discussed, the carrier frequency of the transmitted signal is nominally 5.38 MHz, the frequency of transmitted symbols is nominally 10.76 Msymbols / second, and the sampling frequency of the receiver is nominally 21.52 MHz. Time clock, the sampling frequency of the receiver is twice the frequency of transmitted symbols. In carrier insurance, when baseband demodulation results, the recovered carrier frequency is one quarter of the receiver's sampling frequency. The digital signal from the analog-to-digital converter unit 16 is applied to a carrier processor 18. The processor 18 includes a carrier recovery network of a conventional design, to provide a vestigial sideband output signal that is demodulated up near the baseband. Carrier recovery networks suitable for this purpose are known in the art. A carrier recovery network particularly suitable for use in unit 18 is described in a pending United States patent application by C. Strolle et al., Serial Number.

(RCA 87,862) entitled Carrier Recovery System for a Vestigial Sideband Signal. In the system to be discussed, the demodulation to the absolute baseband is performed by means of a blind compensating network together with a second carrier recovery network, without relying on a pilot signal to assist in carrier recovery, or in a training signal to assist in compensation. The vestigial input sideband signal to be processed is a complex signal with real and imaginary components, and may be of the time used by the Grand Alliance high definition television transmission system. Only the actual component of the vestigial sideband signal contains the data symbols to be recovered. The vestigial sideband output signal near the baseband from the processor 18 contains digital data as well as intersymbol interference (ISI) caused by the alterations of the transmission channel and the artifices. This signal is applied to an input of a complex, adaptive, forward feedband compensator 20, for example, a fractionally spaced compensator, which in this case is implemented as a digital FIR filter. The compensator 20 operates in a "blind" mode during signal acquisition, and then operates in a decision-directed mode. The coefficient values (derivation weights) of the compensator 20 are controlled in an adaptive manner by an error signal "E" applied to a control input, as will be discussed. The initial blind compensation of the vestigial input sideband signal from the processor 18 is performed on the constellation of vestigial sideband symbols by what can be considered a modified version of the algorithm of the reduced constellation algorithm. In a specific manner, the inventors have realized that blind compensation of a vestigial sideband signal can be performed by utilizing a one-dimensional version of the reduced constellation algorithm that is appropriate for a vestigial sideband signal. . The algorithm used determines the appropriate decision regions for a vestigial sideband decision device, to generate decisions that allow an adaptive compensator to converge without using a training signal. It will be useful to define several terms before describing the blind compensation process in more detail. A "decision region" is a continuous portion of the real numerical range, and has upper and lower limits. A "non-limited decision region" is a decision region with either a positive infinity for an upper limit, or a negative infinity for a lower limit. A symbol point is placed in a decision region if it has a value less than the upper limit and greater than the lower limit. A decision region "extends" at a symbol point if the symbol point is located in the decision region. A "decision device", such as a quantifier, determines in which decision region an input symbol point is, and produces a symbol corresponding to that decision region. A "step" is the distance between two adjacent symbols in a complete constellation. As noted above, a vestigial sideband signal is essentially a one-dimensional constellation of data, wherein only one axis contains quantized symbol data to be retrieved at a receiver. In a vestigial sideband system, a decision region typically extends a data symbol of the entire constellation. The upper and lower limits of each decision region are set midway between the sampling points of the constellation. If these decision regions are used for the initial convergence of the compensator, convergence will not occur, because, due to the presence of inter-symbol interference, significantly less than 90 percent of the decisions from the decision device will be correct. A blind compensation algorithm, as will be discussed, determines new upper and lower decision region boundaries in the process of forcing some correct decisions. The complete vestigial sideband constellation is grouped into several series, and the upper and lower limits for the decision regions are determined. These first series are subdivided into smaller series until each series contains only one symbol, and the decision regions correspond to the typical vestigial sideband decision regions. Decision boundaries are typically located halfway between the symbols within the decision regions. Each decision stage, for example, a quantifier, allows a number of decisions to be correct, in such a way that the compensator approaches convergence. Accordingly, each decision stage in the blind compensation process serves to progressively open the "eye" of the vestigial sideband signal as it approaches convergence. The upper and lower limits of each decision region are determined as follows. For a given group of symbols, the lower limit of a given decision region is set to a value that is half a step less than the value of the smallest symbol in that group. However, if the smallest symbol is the smallest value symbol of the constellation, then the lower limit is set to the negative infinity. The upper limit of the decision region is set to a value that is half a step higher than the value of the largest symbol in the group (unless the symbol is the largest value symbol in the constellation, in which case the upper limit is set to a positive infinity value). If an output symbol from the compensator resides in one of these decision regions, the output of the decision device is taken as the arithmetic measure of the data symbols of the associated group. When a locally generated error signal is less than a predetermined threshold level of the quantifier, meaning that the evaluation of the decision region can be refined, the decision regions are changed by dividing each group of symbols in half. The upper and lower limits of the new decision regions and the output of the decision device are recalculated in the manner described above. The process described above is illustrated by the following example for an 8-VSB signal. The signal format adopted by the Grand Alliance high-definition television system employs an 8-VSB signal that has a one-dimensional data constellation defined by the following eight data symbols: -7 -5 -3 -1 +1 +3 +5 +7 This constellation of a dimension is transported by the actual component in phase of the vestigial sideband signal. With this symbol configuration, the symbols are all uniformly separated by two separate units, and the data bits can be mapped into symbols without incurring a direct current offset. The example of the blind compensation process given above encompasses three stages, or levels, in which the input data symbols are grouped or "assembled" in three different ways, and are respectively subjected to progressively finer quantization steps by the devices of associated quantification decision. The first group (thick) of the vestigial sideband constellation points of eight symbols is presented in a first compensation level, involving a thick quantization step, and produces two groups of symbols: [-7, -5, -3 , -1] and [1, 3, 5, 7] For this operation, the slicing point of the quantizer is set to zero, and the data sign (+ or -) is detected. The decision regions of the thick quantization step for each of these groups are respectively: [-infinity, 0] and [0, + infinity] In this case, the outputs of the thick quantifier decision devices are respectively: [-4] [+4] The next level of grouping (thinner) at the next level of compensation produces the following four groups of symbols: [-7, -5] [-3, -1] [1, 3] [3, 5] The decision regions of the finest quantization step for these groups are respectively: [-inf, -4] [-4, 0] [0, 4] [4, inf] In this case, the outputs of the finest resolution decision devices are respectively: [-6] [-2] [2] [6] The last level of refinement in the last level of compensation produces the groups of symbols: [-7] [-5] [-3] [-1] [1] [3] [5] [7] with the finest decision regions: [-inf, -6] [-6, -4] [-4, -2] [-2, 0] [0, 2] [2, 4] [4, 6] [6, inf] The outputs of the finer resolution decision device are therefore the complete vestigial sideband constellation: -7 -5 -3 -1 1 3 5 7 The decision outputs produced by the quantifiers are provided by an input to output mapper (look-up table). The use of this mapper is well known in the design of the quantifier. This example for an 8-VSB signal started with two groups of four symbol samples. It could also have started with a group of eight symbols. An analogous operation belongs to a 16-VSB signal. A 16-VSB signal can start with four groups of four symbols or with two groups of eight symbols. As between the successive coarse and fine regions, the values of the decision region are typically related by a factor of half, but this relationship is not critical. The process described above is summarized in Figure 4, which shows groups, decision regions, and outputs of the decision device for blind compensation of an 8-VSB signal. These operations are performed by a network 50 of FIG. 1, as will be discussed, which includes the quantizers 52, 54, and 56, and a multiplexer (MUX) 58 to provide a data stream of symbol-multiplexed time outputs. Sometimes a modification of the processes described above may be necessary for vestigial side-band signals. A problem arises when some, but not all decision regions of a number of decision regions, are not limited. For vestigial sideband signals, the most external positive and negative decision regions are not limited. Due to the alterations of the transmission channel, it is possible that more points may fall in the non-limited regions than,.,. Normally it would be the case without the distortions of the channel. This situation creates a polarization in the output of the decision device. To overcome this polarization, the scale of the decision regions not limited slightly is shortened, 5 and the scale of the limited decision regions is simultaneously increased. These regions are shortened and lengthened by an amount necessary to reach the optimal phase shift values mentioned in the following paragraph. These values are usually a small percentage of the region of total decision. This adjustment makes the decision of all the decision regions equally equiprobable. This polarization adjustment procedure is illustrated by the following example, in the context of the 8-VSB system described above. In the stage of four groups, for example, the values of the decision region are modified by multiplying by a scalar factor out of phase "?" which has a value slightly greater than unity, for example. The value of the phase shift can vary with the nature and requirements of a particular system. He The purpose of the phase shift is to narrow the range of intermediate decision regions. The phase shift is not used with the outermost values at the positive and negative ends of a decision region, for example, positive or negative infinity. Therefore, in the case of the second group of symbols discussed above, the decision regions are modified as follows: [-inf., -4 *?] [-4 *? 0] [0, 4 *?] [4 *?, + inf.].

The outputs of the decision device are modified in a similar way: -6 *? -2*? 2*? 6 *? The scalar value of phase shift can be determined by experimentation. The optimal offset values for each compensation stage (grouping level) are found by minimizing the transients in the RMS error when the quantizer switches from two groups to four groups and from four groups to 8 groups. These values are often determined in an empirical way. In some cases, the offset value of the output device and the phase shift value of the decision region may be different. Analogous observations are applied to a 16-VSB signal. The operation of the system shown in Figure 1 will now be discussed. In this embodiment, the compensator 20 is implemented as an FIR filter with adjustable leads, although other adaptive filter structures can be used. The compensator 20 is a complex unit with complex inputs and outputs. However, the compensator 20 can be a real-only filter that processes only the actual component in phase of the input signal, and that has a single real output. In Figure 2 a real compensator filter configuration is only shown, as will be discussed. The output signal from the compensator 20 is applied to a first (des) rotator 22, which is of a conventional configuration, and operates in a well-known manner to compensate for phase errors of an input signal in response to a control signal . The rotator 22, a complex multiplier, is included in a secondary carrier recovery network that also includes a phase detector network 30 and a network 24 for separating the actual components in phase and imaginary quadrature of the output signal from the rotator 22. Networks to separate the real and imaginary components of a complex signal are well known. The secondary carrier recovery network typically removes residual phase errors in the output signal of the compensator 20, to produce a baseband signal. The secondary carrier recovery network conveniently augments a previous carrier recovery network in the processor 18, which typically removes the frequency offset, but may lack the power to remove all frequency and phase phase shifts. The real component separated from the network 24 is processed by a real phase detector 32 in the network 30, as will be shown and discussed in connection with Figure 3. An output signal from the detector 32 is representative of a phase error of the detector input signal, which is related to the output signal of the compensator 20. The output signal of the detector 32 is filtered by the cycle filter 34 (e.g., an integrator) to produce a voltage proportional to the error of phase. A Controlled Voltage Oscillator (VCO) 36 produces a frequency proportional to this voltage. Therefore, the output of the controlled voltage oscillator 36 is a complex signal whose frequency and phase are proportional to the phase error of the output signal from the adaptive compensator. The output signal from the controlled voltage oscillator 36 controls the operation of the unloader 22 to compensate for phase errors in the output signal of the compensator 20. Specifically, the rotator 22 modifies the phase of the input signal as a function of the output signal of the controlled voltage oscillator to reduce the phase error to zero. Using well-known signal processing techniques, the control signal from the network 30 is conjugated by the unit 64 to remove the imaginary component of the complex control signal from the network 30. The actual control signal only resulting from the unit 64 it is applied to a control input of a second (des) rotator 62, which will be discussed subsequently. The real component separated from the unit 24 is applied to an input of the network 50 for processing. The unit 67 processes the separate imaginary component from the unit 24 with an actual output signal from the network 50 to reconstitute a complex signal. The network 50 is a multi-stage decision device comprising three parallel stages of the quantizer (decision devices) 52, 54, and 56, which provide quantized data to a time multiplexer of 3: 1 58. The network 450 provides the symbol grouping, decision regions, and decision outputs described above, as summarized in Figure 4. The two-level quantizer 52 is initially used during the first (thick) level of the compensation. When the RMS value of a passband error signal E developed at the output of the second rotator 62 exceeds a predetermined threshold, as detected by a comparator network on a sensor 66, a Multiplexer Control signal is generated ( MUX) by the sensor 66. This control signal causes the multiplexer 58 of the network 50 to select the output from the next (finer) level quantizer, for example, from the quantizer of four levels 54 at the second level of compensation. The compensator responds to the information derived from the use of this quantifier, until the RMS error exceeds a previously determined second threshold, also detected by the sensor 66. A control signal of the multiplexer, generated for this condition, causes the network 50 select the output from the next and last level quantifier (finest), the eight level quantizer 56 in the third final compensation level in this example. The quantizer 56 covers the entire 8-VSB constellation. At this point, it is expected that the compensator 20 can converge completely. The input to the non-inverting input (+) of the combiner 60 is the complex signal before quantization, and the input to the inverting (-) input of the combiner is the complex signal after the quantization of the real component. Accordingly, the output signal from the combiner 60 represents the quantization difference before and after, or the offset / error from the desired quantization level. This signal represents a baseband phase error. Rotator 22 and rotator 62 are similar complex rotators that rotate in opposite directions (ie clockwise and counterclockwise). The difference in the direction of rotation is caused by the conjugation of the signal applied to the control input of the rotator 62 compared to the rotator 22. The error signal E developed at the output of the rotator 62, represents the band phase error in step which is intended to remove the compensator 20 by adjusting its coefficients in response to the error signal E. The network 50 can use a single adaptive quantizer with a controllable quantization level instead of three separator quantifiers 52, 54, and 56, as shown. A compensated baseband signal is decoded by the unit 76, and processed by an output processor 78. The decoder 76 may include, for example, the deinterleaver, the Reed-Solomon error correction, and the audio decoder networks. video, as it is known. The output processor 78 may include audio and video processors and audio and video playback devices. In a system using a trellis decoder, an input to the trellis decoder can be taken from the TI terminal at the output of the first rotator 22. The system shown in Figure 2 also performs blind compensation of a vestigial sideband signal near the baseband, but uses a real compensator only rather than a complex compensator as used in Figure 1. In Figure 2, the actual component of a received vestigial sideband signal is applied to an input of a real adaptive compensator only , feed forward 210. The coefficients of compensator 210 are adjusted in response to an error signal E (as will be discussed). The actual output signal of the compensator 210 is combined in the adder 212 with an actual output signal from a decision feedback compensator 214. A filter network 216 reconstructs the imaginary quadrature phase component of the sideband signal component. real vestigial, from the actual output of adder 212. This reconstruction is performed using known Hilbert transformation techniques, and is based on the fact that the real component in phase and the imaginary component in quadrature of a sideband signal vestigial, form approximately a pair of Hilbert transformation. The unit 218 combines the quadrature component reconstructed from the filter 216, and the real component from the unit 212, to produce a complex vestigial sideband signal reconstructed with a compensated real component. The delay element 220 compensates for a time delay associated with the operation of the reconstruction filter 216, to ensure that the input signals arrive at the adder 218 with a time match. The complex vestigial sideband signal from the unit 218 is processed by a multiplier (rotator) 224 which operates in the same manner as the rotator 22 of FIG. 1, and in the same manner responds to a complex output signal produced by a voltage oscillator controlled in a secondary carrier recovery network 226 corresponding to the unit 30 in Figure 1. As in the case of Figure 1, the carrier recovery network 226 responds to the actual component separated from the output signal from the rotator 224, as provided by a real / imaginary component separator 228. The actual component from the unit 228 is processed by a decision network of the multistage quantizer 230 corresponding to the network 50 of Figure 1. It appears a baseband vestigial sideband signal compensated at the output of quantizer 230, and transported to subsequent signal processing circuits (not shown). The actual input and output signals of the quantizer 230 are differentiated by a subtracter 232, and the resulting real signal is applied to an input of the multiplier 234. The output signal of the subtracter represents the difference between the actual input signal to the quantizer 230 and the actual output signal quantized from the quantizer 230. Another input of the multiplier 234 receives the output of the real signal from the subtracter 232. The imaginary component separated by the unit 228 and the real component compensated from the output of the quantizer 230 are combined by the unit 236 to produce a complex vestigial sideband signal, which is applied to a signal input of the multiplier 240. Another input of the multiplier 240 receives a real signal from a conjugating network 236, which inverts the imaginary component of the complex output signal of unit 226. The output of multiplier 240 is a real passband signal (only the actual output of the complex multiplier is used). This signal is applied to a signal input of the decision feedback compensator 214, and a control input of the compensator 214 receives an error output signal (E) from the multiplier 234. This error signal represents an error signal of passband, and also applies to the input buffer 210 as a control signal of the coefficient. The output of the compensator 214 is real and is combined in a unit 212 with the compensated actual output signal of the compensator 210. The feedback compensator 214 removes the remaining intersymbol interference not removed by the forward compensator 210. The feedback compensators of FIG. decision are well known. The multistage quantizer 230 can be controlled by means of a Multiplexer control signal in the same manner as shown in Figure 1, by detecting the Error signal to develop the Multiplexer Control signal, which is applies to a Multiplexer, which is applied to a multiplexer associated with the quantizer 230, as described in relation to Figure 1. The phase detector 32 of the network 30 in the secondary carrier recovery cycle of Figure 1, and in the corresponding network 226 of Figure 2, is shown in detail in Figure 3. Phase detector 32 measures the phase error of the carrier using only the actual component of the vestigial sideband signal, and produces a signal output that is proportional to the sine of the carrier phase error. The phase detector 32 detects essentially any quadrature phase component of the actual component input to the phase detector. Any quadrature distortion of the real component represents a phase shifting error that is manifested at the output of the phase detector 32. The phase detector includes a quantizer 310, symbol delay elements 312 and 314, multipliers 316 and 318, and a subtraction combiner 320 configured as shown. The quantizer 310 is a quantizer of eight levels in the case of an 8-VSB signal, a quantizer of sixteen levels in the case of a 16-VSB signal, and so on. The delay elements 312 and 314 compensate for a transient time delay associated with the operation of the quantizer 310, such that the signals reach the multipliers 314 and 316 in time synchronization. The phase detector 32 is a low latency phase detector with a small delay (a symbol) between the input and the output, which produces good noise tracking. The phase detector generates a Phase Error Output signal Ph (t) that is proportional to the sine of the phase error (angular) of the input signal h (t). This signal is a rotated version of the output signal of the adaptive compensator, as can be seen in Figure 1. The output signal of the phase detector Ph (t) is defined by the expression: Ph (t) = h (t) * h '(t-T) - h * (t) * h (t-T) where h '(t) is the output of the decision device of the quantizer 310, h (t) is the output of the adaptive quantizer after rotation, and T is a symbol period. The output signal of the phase detector Ph (t) is proportional to the sine of its input signal, and not to the time offset. The sine function is not a mathematical sine function by itself, but results from the shape of the input-output transfer function of the phase detector 32.

Claims (7)

1. In a system for receiving a vestigial sideband (VSB) video signal formatted as a constellation of a dimension of data symbols representing data from a digital image, and subject to the display of a carrier phase shift, an apparatus that comprises: a carrier recovery network (22,30) to change the received vestigial sideband signal to the baseband, - and a phase detector (32) in the carrier recovery network, which includes: an input for receiving a vestigial sideband signal near the baseband, - a quantizer (310) that responds to the vestigial input sideband signal to produce a quantized vestigial sideband signal; a symbol delay network (312, 314) for delaying the vestigial input sideband signal, and for delaying the quantized signal; a first multiplier (316) that responds to the quantized vestigial sideband signal and to a signal delayed by a delay network symbol, to produce a first signal; a second multiplier (314) that responds to the vestigial input sideband signal and a quantized signal delayed by a symbol from the delay network, to produce a second signal; and a combiner for combining the first and second signals in a subtractive manner to produce a signal representative of the phase error.

2. An apparatus according to claim 1, wherein: the vestigial input sideband signal exhibits a real component exclusive of an imaginary component. An apparatus according to claim 1, and further comprising: a time recovery network for providing a symbol clock synchronized with a transmitter clock, such that the input signal to the phase detector exhibits the second of time. An apparatus according to claim 1, and further comprising: a signal compensator (20) having an input for receiving a transmitted vestigial sideband signal, an output coupled with the carrier recovery network, and a control input to receive a control signal that is a function of the error signal. An apparatus according to claim 4, and further comprising: an additional carrier recovery network (18) with an input for receiving a transmitted vestigial sideband signal and an output coupled with the input of the compensator. 6. An apparatus according to claim 1, wherein: the vestigial input sideband signal is a vestigial sideband signal of level N; and the quantifier exhibits N levels of quantification. 7. An apparatus according to claim 1, wherein: the delay network exhibits a delay of a symbol.