Receiver for QAM digital television signals

[0002] This is a continuation-in-part of U.S. patent application Ser. No. 08/785,227 filed Jan. 17, 1997, which is a continuation-in-part of U.S. patent application Ser. No. 08/266,753 filed Jun. 28, 1994.

[0003] The invention relates to radio receivers having the capability of receiving digital television (DTV) signals transmitted using quadrature amplitude modulation (QAM) of the principal carrier wave.

[0004] A Digital Television Standard published Sep. 16, 1995 by the Advanced Television Subcommittee (ATSC) specifies vestigial sideband (VSB) signals for transmitting digital television (DTV) signals in 6-MHz-bandwidth television channels such as those currently used in over-the-air broadcasting of National Television Subcommittee (NTSC) analog television signals within the United States. The VSB DTV signal is designed so its spectrum is likely to interleave with the spectrum of a co-channel interfering NTSC analog TV signal. This is done by positioning the pilot carrier and the principal amplitude-modulation sideband frequencies of the DTV signal at odd multiples of one-quarter the horizontal scan line rate of the NTSC analog TV signal that fall between the even multiples of one-quarter the horizontal scan line rate of the NTSC analog TV signal, at which even multiples most of the energy of the luminance and chrominance components of a co-channel interfering NTSC analog TV signal will fall. The video carrier of an NTSC analog TV signal is offset 1.25 MHz from the lower limit frequency of the television channel. The carrier of the DTV signal is offset from such video carrier by 59.75 times the horizontal scan line rate of the NTSC analog TV signal, to place the carrier of the DTV signal about 309,877.6 kHz from the lower limit frequency of the television channel. Accordingly, the carrier of the DTV signal is about 2,690122.4 Hz from the middle frequency of the television channel. The exact symbol rate in the Digital Television Standard is (684/286) times the 4.5 MHz sound carrier offset from video carrier in an NTSC analog TV signal. The number of symbols per horizontal scan line in an NTSC analog TV signal is 684, and 286 is the factor by which horizontal scan line rate in an NTSC analog TV signal is multiplied to obtain the 4.5 MHz sound carrier offset from video carrier in an NTSC analog TV signal. The symbol rate is 10.762238*10⁶symbols per second, which can be contained in a VSB signal extending 5.381119 MHz from DTV signal carrier. That is, the VSB signal can be limited to a band extending 5.690997 MHz from the lower limit frequency of the television channel.

[0005] The ATSC standard for digital HDTV signal terrestrial broadcasting in the United States of America is capable of transmitting either of two high-definition television (HDTV) formats with 16:9 aspect ratio. One HDTV format uses 1920 samples per scan line and 1080 active horizontal scan lines per 30 Hz frame with 2:1 field interlace. The other HDTV format uses 1280 luminance samples per scan line and 720 progressively scanned scan lines of television image per 60 Hz frame. The ATSC standard also accommodates the transmission of DTV formats other than HDTV formats, such as the parallel transmission of four television signals having normal definition in comparison to an NTSC analog television signal.

[0006] DTV transmitted by vestigial-sideband (VSB) amplitude modulation (AM) during terrestrial broadcasting in the United States of America comprises a succession of consecutive-in-time data fields each containing 313 consecutive-in-time data segments. There are 832 symbols per data segment. So, with the symbol rate being 10.76 MHz, each data segment is of 77.3 microseconds duration. Each segment of data begins with a line synchronization code group of four symbols having successive values of +S, −S, −S and +S. The value +S is one level below the maximum positive data excursion, and the value −S is one level above the maximum negative data excursion. The initial line of each data field includes a field synchronization code group that codes a training signal for channel-equalization and multipath suppression procedures. The training signal is a 511 -sample pseudo-random noise sequence (or “PN511 sequence”) followed by three 63-sample pseudo-random noise sequences (or “PN63 sequences”). The middle PN63 sequence in the training signal is transmitted in accordance with a first logic convention in the first line of each odd-numbered data field and in accordance with a second logic convention in the first line of each even-numbered data field, the first and second logic conventions being one's complementary respective to each other. The other PN63 sequences and the PN511 sequence are transmitted in accordance with the first logic convention in the first line of each data field, whether odd-numbered or even-numbered. The initial line of each data field includes other information, such as a code for indicating the mode of the VSB transmission, and such as an indication of trellis code information at the close of the previous data field.

[0007] The remaining lines of each data field contain data that have been Reed-Solomon forward error-correction coded after having been randomized and subjected to diagonal byte interleaving. In over-the-air broadcasting the error-correction coded data are then trellis coded using twelve interleaved trellis codes, each a ⅔ rate punctured trellis code with one uncoded bit. Trellis coding results are parsed into three-bit groups for over-the-air transmission in eight-level one-dimensional-constellation symbol coding, which transmission is made without symbol pre-coding separate from the trellis coding procedure. Trellis coding is not used in cablecasting proposed in the ATSC standard. The error-correction coded data are parsed into four-bit groups for transmission as sixteen-level one-dimensional-constellation symbol coding, which transmissions are made without precoding.

[0008] The VSB signals have their natural carrier wave, which would vary in amplitude depending on the percentage of modulation, suppressed. The natural carrier wave is replaced by a pilot carrier wave of fixed amplitude, which amplitude corresponds to a prescribed percentage of modulation. This pilot carrier wave of fixed amplitude is generated by introducing a direct component shift into the modulating voltage applied to the balanced modulator generating the amplitude-modulation sidebands that are supplied to the filter supplying the VSB signal as its response. If the eight levels of 3-bit symbol coding have normalized values of −7, −5, −3, −1, +1, +3, +5 and +7 in the carrier modulating signal, the pilot carrier has a normalized value of 1.25. The normalized value of +S is +5, and the normalized value of −S is −5.

[0009] VSB signals using 8-level symbol coding will be used in over-the-air broadcasting within the United States, and VSB signals using 16-level symbol coding can be used in over-the-air narrowcasting systems or in cable-casting systems. However, certain cable-casting is likely to be done using suppressed-carrier quadrature amplitude modulation (QAM) signals instead, rather than using VSB signals. This presents television receiver designers with the challenge of designing receivers that are capable of receiving QAM transmissions of DTV as well as receiving VSB transmissions of DTV.

[0010] Radio receivers for receiving digital television signals, in which receiver the final intermediate-frequency signal is somewhere in the 1-8 MHz frequency range rather than at baseband, are described by C. B. Patel et alii in U.S. Pat. No. 5,479,449 issued Dec. 26, 1995, entitled DIGITAL VSB DETECTOR WITH BANDPASS PHASE TRACKER, AS FOR INCLUSION IN AN HDTV RECEIVER, and included herein by reference. The use of infinite-impulse response filters for developing complex digital carriers in such receivers is described by C. B. Patel et alii in U.S. Pat. No. 5,548,617 issued Aug. 20, 1996, entitled DIGITAL VSB DETECTOR WITH BANDPASS PHASE TRACKER USING RADER FILTERS, AS FOR USE IN AN HDTV RECEIVER, and incorporated herein by reference. The use of finite-impulse response filters for developing complex digital carriers in such receivers is described by C. B. Patel et alii in U.S. Pat. No. 5,731,848 issued Mar. 24, 1998, entitled DIGITAL VSB DETECTOR WITH BANDPASS PHASE TRACKER USING NG FILTERS, AS FOR USE IN AN HDTV RECEIVER, and incorporated herein by reference. The design of receivers for both VSB and QAM signals in which both types of signal are processed through the same intermediate-frequency amplifiers receivers is described by C. B. Patel et alii in U.S. Pat. No. 5,506,636 issued Apr. 9, 1996, entitled HDTV SIGNAL RECEIVER WITH IMAGINARY-SAMPLE-PRESENCE DETECTOR FOR QAM/VSB MODE SELECTION, and incorporated herein by reference. U.S. Pat. No. 5,606,579 issued Feb. 25, 1997 to C. B. Patel et alii and entitled DIGITAL VSB DETECTOR WITH FINAL I-F CARRIER AT SUBMULTIPLE OF SYMBOL RATE, AS FOR HDTV RECEIVER is incorporated herein by reference. U.S. Pat. No. 5,659,372 issued Aug. 19, 1997 to C. B. Patel et alii and entitled DIGITAL TV DETECTOR RESPONDING TO FINAL-IF SIGNAL WITH VESTIGIAL SIDEBAND BELOW FULL SIDEBAND IN FREQUENCY is incorporated herein by reference. U.S. patent application Ser. No. 08/266,753 filed Jun. 28, 1994 by C. B. Patel et alii and entitled RADIO RECEIVER FOR RECEIVING BOTH VSB AND QAM DIGITAL HDTV SIGNALS is incorporated herein by reference. U.S. patent application serial U.S. Pat. No. 5,748,226 issued May 5, 1998 to A. L. R. Limberg and entitled DIGITAL TELEVISION RECEIVER WITH ADAPTIVE FILTER CIRCUITRY FOR SUPPRESSING NTSC CO-CHANNEL INTERFERENCE is incorporated herein by reference. These patents and patent applications are all assigned to Samsung Electronics Co., Ltd., pursuant to employee invention agreements already in force at the time the inventions disclosed in these patents and patent applications were made.

[0011] In the radio receivers described in U.S. Pat. No. 5,506,636 the final intermediate-frequency signal is digitized, and synchrodyne procedures to obtain baseband samples are carried out in the digital regime. In radio receivers that are to have the capability of receiving digital TV signals no matter whether they are transmitted using VSB or QAM, frequencies of the local oscillators in the tuner can remain the same no matter whether VSB or QAM transmissions are being received. U.S. Pat. No. 5,506,636 indicates that this is possible because the differences in carrier frequency location within the channel can be accommodated in the synchrodyning procedures carried out in the digital regime. The radio receivers described in U.S. Pat. No. 5,506,636 and claimed in the claims following this specification do not synchronously detect the in-phase QAM sidebands and the quadrature-phase QAM sidebands in the analog regime and then digitize the two sets of baseband results in respective analog-to-digital conversion circuits for application to digital equalization filtering. Rather, to avoid the problems with accurately tracking two analog-to-digital conversion circuits at high sampling rates, a single analog-to-digital converter digitizes the final intermediate-frequency signal and the resulting stream of digital samples is converted to complex digital samples including a stream of real digital samples and a stream of imaginary digital samples. These complex digital samples and complex digital samples descriptive of the suppressed carrier are then multiplied in a complex digital multiplier to generate real and imaginary baseband results that are already in digital form circuits for application to digital equalization filtering.

[0012] The invention is embodied in apparatus for reproducing data from a selected digital television signal transmitted in packet form as quadrature-amplitude-modulation (QAM) of a carrier. This apparatus has a tuner for selecting said digital television signal and converting it to a final intermediate-frequency signal, which is digitized by an analog-to-digital converter before synchrodyning to baseband in the digital regime. QAM synchrodyning circuitry responsive to the digitized final intermediate-frequency signal generates real and imaginary digital sample streams of QAM symbol code, by synchrodyning the final IF signal to baseband in the digital regime. Symbol decoding circuitry responsive to the real and imaginary digital sample streams of QAM symbol coding generates a digital data stream. Data synchronization recovery circuitry recovers data synchronizing information included in the digital data stream. A video decoder of decompression type responsive to portions of the digital data stream, as selected in response to the data synchronizing information, generates digital signals descriptive of red, green and blue video signals. The video decoder can be an MPEG-2 decoder, for example, or one for decoding video compressed using wavelet descriptions rather than direct-cosine-transform descriptions, by way of further example.

[0013] FIG. 1 is a block schematic diagram of initial portions of a digital television (DTV) signal radio receiver of a type that embodies the invention, including circuitry for detecting symbols in a DTV signal of QAM type, circuitry for detecting symbols in a DTV signal of VSB type, and an amplitude-and-group-delay equalizer for symbols selected from the circuitry for detecting symbols in a DTV signal of QAM type and the circuitry for detecting symbols in a DTV signal of VSB type.

[0014] FIG. 2 is a block schematic diagram of the remaining portions of the DTV signal radio receiver of a type that embodies the invention, which are not shown in FIG. 1.

[0015] FIG. 3 is a detailed block schematic diagram of digital circuitry for synchrodyning QAM DTV signals to baseband in accordance with the invention, of digital circuitry for synchrodyning VSB DTV signals to baseband, and of circuitry associated with applying input signals to that QAM and VSB synchrodyning circuitry, as used in a DTV signal radio receiver of the type shown in FIGS. 1 and 2.

[0016] FIG. 4 is a detailed block schematic diagram of circuitry for providing the sample clock generator, the look-up table read-only memories (ROMS) for supplying digital descriptions of the complex carriers used in accordance with the invention for synchrodyning digital QAM signals at final IF signal frequencies to baseband (as well as for synchrodyning digital VSB signals at final IF signal frequencies to baseband), and the address generators for those ROMs.

[0017] FIG. 5 is a detailed block schematic diagram of circuitry similar to that of FIG. 4, modified so that the address generator for the ROMs supplying digital descriptions of the complex carrier used for synchrodyning digital QAM signals to baseband and the ROMs supplying digital descriptions of the complex carrier used for synchrodyning digital VSB signals to baseband share an address counter in common.

[0018] FIG. 6 is a detailed block schematic diagram of circuitry for converting digital samples to complex form in DTV signal radio receivers embodying the invention, which circuitry includes a Hilbert transformation filter for generating imaginary samples from real samples, and which includes delay compensation for the real samples equivalent to the latency of that filter.

[0019] FIG. 7 is a detailed block schematic diagram of a pair of all-pass digital filters of infinite-impulse-response (IIR) type designed based on Jacobian elliptic functions and exhibiting a constant π/2 difference in phase response for the digitized bandpass signals, as can be employed for converting digital samples to complex form in DTV signal radio receivers embodying the invention.

[0020] FIGS. 8 and 9 are block schematic diagrams of changes that can be made the filter circuitry of FIG. 7 to remove redundant delay.

[0021] FIG. 10 is a detailed block schematic diagram of a pair of all-pass digital filters of finite-impulse-response (FIR) type exhibiting a constant π/2 difference in phase response for the digitized bandpass signals, as can be employed for converting digital samples to complex form in DTV signal radio receivers embodying the invention.

[0022] FIG. 11 is a graph of the constraints on the final intermediate frequencies to which the carriers of a QAM DTV signal and a VSB DTV signal can be frequency translated, when the carrier of a VSB DTV signal is lower in frequency than the carrier of a QAM DTV signal in the final IF signal, so the full sideband of the VSB DTV signal is higher in frequency than its vestigial sideband in the final IF signal, and when the sample rate during digitization is constrained to 21.52*10⁶samples per second.

[0023] FIG. 12 is a graph of the constraints on the final intermediate frequencies to which the carriers of a QAM DTV signal and a VSB DTV signal can be frequency translated, when the carrier of a VSB DTV signal is higher in frequency than the carrier of a QAM DTV signal in the final IF signal, so the full sideband of the VSB DTV signal is lower in frequency than its vestigial sideband in the final IF signal, and when the sample rate during digitization is constrained to 21.52* 10⁶samples per second.

[0024] In the block schematic diagrams, clock or control signal connections are shown in dashed line, where it is desired to distinguish them from connections for the signals being controlled. To avoid overcomplexity in the block schematic diagrams, some shimming delays necessary in the digital circuitry are omitted, where a need for such shimming delay is normally taken into account by a circuit or system designer.

[0025] FIG. 1 shows a tuner 5 comprising elements 10-21, which tuner 5 selects one of channels at different locations in the frequency band for DTV signals and performs plural frequency conversion of the selected channel to a final intermediate-frequency signal in a final intermediate-frequency band. FIG. 1 shows a broadcast receiving antenna 6 arranged to capture the DTV signals for the tuner 5. Alternatively, the tuner 5 can be connected for receiving DTV signals from a narrowcast receiving antenna or from a cablecast transmission system.

[0026] More particularly, in the tuner 5 shown in FIG. 1, a channel selector 10 designed for operation by a human being determines the frequency of first local oscillations that a frequency synthesizer 11, which functions as a first local oscillator, furnishes to a first mixer 12 for heterodyning with DTV signals received from the antenna 6 or an alternative source of such signals. The first mixer 12 upconverts the received signals in the selected channel to prescribed first intermediate frequencies (e.g., with 920 MHz carrier), and an LC filter 13 is used to reject the unwanted image frequencies that accompany the upconversion result supplied from the first mixer 12. The first intermediate-frequency signal resulting from the upconversion, supplied as the filter 13 response, is applied as the input signal to a first intermediate-frequency amplifier 14, which supplies amplified first IF signal for driving a first surface-acoustic-wave (SAW) filter 15 or a filter constructed from ceramic resonators. The upconversion to the rather high-frequency first intermediate frequencies facilitates the SAW filter 15 having a large number of poles and zeroes. The SAW filter 15 passband is designed to pass those frequencies obtained by converting frequencies extending from the lower limit frequency of the television channel up to about 300 kHz of the upper limit frequency of the television channel. Preferably the SAW filter 15 is designed to reject the frequency-modulated sound carrier of any co-channel interfering NTSC analog TV signal. Second local oscillations from a second local oscillator 16 are supplied to a second mixer 17 for heterodyning with the response of the first SAW filter 15, to generate second intermediate frequencies (e. g., with 41 MHz carrier). A second SAW filter 18 is used for rejecting the unwanted image frequencies that accompany the downconversion result supplied from the second mixer 17. The skirts of the SAW filter 18 response may be designed for completing the roll-offs at the edges of the overall transmission channel frequency spectrum response that are needed to establish the Nyquist slopes desired for minimizing intersymbol interference. During the period of transition from NTSC television transmissions to digital television transmissions, the second SAW filter 18 will usually include traps for sound and video carriers of adjacent-channel NTSC television transmissions. The second IF signal supplied as the response of the second SAW filter 18 is applied as input signal to a second intermediate-frequency amplifier 19, which generates an amplified second IF signal response to its input signal. Oscillations from a third local oscillator 20 are heterodyned with the amplified second IF signal response in a third mixer 21. The plural-conversion tuner 5 as thusfar described resembles those previously proposed by others, except that the frequency of the oscillations from the third local oscillator 20 is chosen such that the third mixer 21 supplies a third intermediate-frequency signal response.

[0027] This third IF signal response is the final intermediate-frequency output signal of the tuner 5, which is supplied to a subsequent analog-to-digital converter (ADC) 22 for digitization. This final IF signal occupies a frequency band 6 MHz wide, the lowest frequency of which is above zero frequency. The lowpass analog filtering of the third mixer 21 response done in the ADC 22 as a preliminary step in analog-to-digital conversion suppresses the image frequencies of the third intermediate frequencies, and the second SAW filter 18 has already restricted the bandwidth of the third intermediate-frequency signals presented to the ADC 22 to be digitized; so the ADC 22 functions as a bandpass analog-to-digital converter. The sampling of the lowpass analog filter response in the ADC 22 as the next step in analog-to-digital conversion is done responsive to pulses in a first clock signal supplied from a sample clock generator 23.

[0028] The sample clock generator 23 preferably includes a crystal oscillator capable of frequency control over a relatively narrow range for generating cissoidal oscillations at a multiple of symbol rate. A symmetrical clipper or limiter generates a square-wave response to these cissoidal oscillations to generate the first clock signal, which the ADC 22 uses to time the sampling of the final IF signal after filtering to limit bandwidth. The frequency of the cissoidal oscillations generated by the crystal oscillator in the sample clock generator 23 can be determined by an automatic frequency and phase control (AFPC) signal developed in response to symbol frequency components of the received DTV signal, for example, as will be described in detail further on in this specification. The pulses in the first clock signal recur at a 21.52*10⁶samples-per-second rate, twice the 10.76*10⁶symbols-per-second symbol rate for VSB signals and four times the 5.38*10⁶symbols-per-second symbol rate for QAM signals. At this 21.52*10⁶samples-per-second clock rate, placing the final IF signal so its mid-frequency is above 5.38 MHz reduces the number of 21.52*10⁶samples-per-second rate samples in the QAM carrier to less than four, which undesirably reduces the uniformity of synchrodyne response supplied for symbol decoding.

[0029] The ADC 22 supplies real digital responses of 10-bit or so resolution to the samples of the band-limited final IF signal, which digital responses are converted to complex digital samples by the circuitry 24. There are various ways to construct the circuitry 24. The imaginary digital samples at the QAM carrier frequency may be generated using a Hilbert transformation filter, for example, as described in U.S. Pat. No. 5,479,449. If the frequency band 6 MHz wide occupied by the final IF signal has a lowest frequency of at least a megaHertz or so, it is possible to keep the number of taps in the Hilbert transformation filter reasonably small and thus keep the latency time of the filter reasonably short. Other ways to construct the circuitry 24 described in U.S. Pat. No. 5,548,617 rely on the differential delay between the responses of two infinite-impulse-response (IIR) filters being substantially equal to phase shift at all frequencies. Still other ways to construct the circuitry 24 rely on the differential delay between the responses of two finite-impulse-response (FIR) filters being substantially equal to 90° phase shift at all frequencies.

[0030] In the FIG. 1 receiver circuitry the complex digital samples of final IF signal supplied from the circuitry 24 are applied to circuitry 25 for synchrodyning the QAM signal to baseband. The circuitry 25 supplies a stream of real samples and a stream of imaginary samples in parallel to a symbol de-interleaver 26, to provide baseband description of the QAM modulating signal. The QAM synchrodyning circuitry 25 receives complex-number digital descriptions of two phasings of the QAM carrier, as translated to final intermediate frequency and in quadrature relationship with each other, from read-only memory 27. ROM 27, which comprises sine and cosine look-up tables for QAM carrier frequency, is addressed by a first address generator 28. The first address generator 28 includes an address counter (not explicitly shown in FIG. 1) for counting the recurrent clock pulses in the first clock signal generated by the sample clock generator 23. The resulting address count is augmented by a symbol phase correction term generated by QAM de-rotator correction circuitry, thereby to generate the addressing for the ROM 27. The QAM synchrodyne circuitry 25, the first address generator 28, and the operation of each will be explained in greater detail further on in this specification.

[0031] In the FIG. 1 receiver circuitry the complex digital samples of final IF signal supplied from the circuitry 24 are also applied to circuitry 30 for synchrodyning the VSB signal to baseband. The VSB synchrodyning circuitry 30 supplies streams of samples descriptive of real and imaginary components of the vestigial-sideband modulating signal as synchrodyned to baseband. The VSB synchrodyning circuitry receives complex-number digital descriptions of two phasings of the VSB carrier, as translated to final intermediate frequency and in quadrature relationship with each other, from read-only memory 31. ROM 31, which comprises sine and cosine look-up tables for VSB carrier frequency, is addressed by a second address generator 32. The second address generator 32 includes an address counter (not explicitly shown in FIG. 1) for counting the recurrent clock pulses in the first clock signal generated by the sample clock generator 23. In preferred embodiments of the invention this address counter is the same address counter used by the first address generator 28. The resulting address count is augmented by a symbol phase correction term generated by symbol phase correction circuitry, thereby to generate the addressing for the ROM 31. The VSB synchrodyne circuitry 30, the second address generator 32, and the operation of each will be explained in greater detail further on in this specification.

[0032] A digital-signal multiplexer 33 functions as a synchrodyne result selector that selects as its response either a first or a second one of two complex digital input signals supplied thereto, the selection being controlled by a detector 34 for detecting the zero-frequency term of the real samples from the VSB synchrodyne circuitry 30. When the zero-frequency term has essentially zero energy, indicating the absence of pilot carrier signal that accompanies a VSB signal, the multiplexer 33 selectively responds to its first complex digital input signal, which is the de-interleaved QAM synchrodyne-to-baseband result supplied from the de-interleaver 26. When the zero-frequency term has substantial energy, indicating the presence of pilot carrier signal that accompanies a VSB signal, the multiplexer 33 selectively responds to its second complex digital input signal comprising the real and imaginary components of the baseband response of the VSB synchrodyne circuitry 30.

[0033] The response of the synchrodyne result selection multiplexer 33 is resampled in response to a second clock signal from the sample clock generator 23 in 2:1 decimation circuitry 35, to reduce the sample rate of complex baseband response down to the 10.76 MHz VSB symbol rate, which is twice the 5.38 MHz QAM symbol rate. The 2:1 decimation of the multiplexer 33 response prior to its application as input signal to an amplitude-and-group-delay equalizer 36 reduces the hardware requirements on the equalizer. Alternatively, rather than 2:1 decimation circuitry 35 being used after the synchrodyne result selection multiplexer 33, the baseband responses of the QAM synchrodyne circuitry 25 and of the VSB synchrodyne circuitry 30 can each be resampled in response to a second clock signal from the sample clock generator 23 to carry out 2:1 decimation before the synchrodyne result selection multiplexer 33.

[0034] FIG. 2 shows the amplitude-and-group-delay equalizer 36, which converts a baseband response with an amplitude-versus-frequency characteristic that tends to cause inter-symbol error to a more optimum amplitude-versus-frequency characteristic that minimizes the likelihood of inter-symbol error. The amplitude-and-group-delay equalizer 36 can be a suitable one of the monolithic ICs available off-the-shelf for use in equalizers. Such an IC includes a multiple-tap digital filter used for amplitude-and-group-delay equalization, the tap weights of which filter are programmable; circuitry for selectively accumulating training signal and temporarily storing the accumulation results; and a microcomputer for comparing the temporarily stored accumulation results with an ideal training signal as known a priori and for calculating updated tap weights of the multiple-tap digital filter used for amplitude-and-group-delay equalization.

[0035] The response of the amplitude-and-group-delay equalizer 36 is applied as input signal to a two-dimensional trellis decoder 37, which performs the symbol decoding that recovers a digital data stream from a QAM-origin signal. The response of the amplitude-and-group-delay equalizer 36 is also applied as input signal to a one-dimensional trellis decoder 38, which performs the symbol decoding that recovers a digital data stream from a VSB-origin signal. A digital-signal multiplexer 39 functions as a data source selector that selects as its response either a first or a second one of two digital input signals thereto, the selection being controlled by the detector 34 for detecting the zero-frequency term of the real samples from the VSB synchrodyne circuitry 30. When the zero-frequency term has essentially zero energy, indicating the absence of pilot carrier signal that accompanies a VSB signal, the multiplexer 39 selectively responds to its first digital input signal, selecting as the source of its digital data output the two-dimensional trellis decoder 37 that decodes the symbols received in the QAM signal. When the zero-frequency term has substantial energy, indicating the presence of pilot carrier signal that accompanies a VSB signal, the multiplexer 39 selectively responds to its second digital input signal, selecting as the source of its digital data output the one-dimensional trellis decoder 38 that decodes the symbols received in the VSB signal.

[0036] The data selected by the data source selection multiplexer 39 are applied to a data de-interleaver 40 as its input signal, and the de-interleaved data supplied from the data de-interleaver 40 are applied to a Reed-Solomon decoder 41. The data de-interleaver 40 is often constructed within its own monolithic IC and is made so as to respond to the output indications from the pilot carrier presence detector 34 to select the de-interleaving algorithm suitable to the DTV signal currently being received, whether it be of QAM or VSB type; this is a mere matter of design. The Reed-Solomon decoder 41 is often constructed within its own monolithic IC and is made so as to respond to the output indications from the pilot carrier presence detector 34 to select the appropriate Reed-Solomon decoding algorithm for the DTV signal currently being received, whether it be of QAM or VSB type; this also is a mere matter of design. Error-corrected data are supplied from the Reed-Solomon decoder 41 to a data de-randomizer 42, which regenerates packets of data for a packet sorter 43. The data de-randomizer 42 is made so as to respond to the output indications from the pilot carrier presence detector 34 to select the appropriate data de-randomizing algorithm for the DTV signal currently being received, whether it be of QAM or VSB type; this is a mere matter of design, too.

[0037] First data synchronization recovery circuitry 44 recovers the data synchronizing information included in the data output of the two-dimensional trellis decoder 37, and second data synchronization recovery circuitry 45 recovers the data synchronizing information included in the data output of the one-dimensional trellis decoder 38. A data sync selector 46 selects between the data synchronizing information as provided by the data sync recovery circuitry 44 and as provided by the data sync recovery circuitry 45, the selection being controlled by the detector 34 for detecting the zero-frequency term of the real samples from the VSB synchrodyne circuitry 30. When the zero-frequency term has essentially zero energy, indicating the absence of pilot carrier signal that accompanies a VSB signal, the data sync selector 46 selects for its output signals the data synchronizing information provided by the data sync recovery circuitry 44. When the zero-frequency term has substantial energy, indicating the presence of pilot carrier signal that accompanies a VSB signal, the data sync selector 46 selects for its output signals the data synchronizing information provided by the data sync recovery circuitry 45.

[0038] When the data sync selector 46 selects for its output signals the data synchronizing information provided by the data sync recovery circuitry 45, the initial data lines of each data field are selected for application to the amplitude-and-group-delay equalizer 36 as training signal. The occurrences of two consecutive 63-sample PR sequences are detected within the data sync recovery circuitry 45 to provide data-field indexing information to the data sync selector 46.

[0039] The standards for a QAM DTV signal are not as well defined at this time as the standards for a VSB DTV signal. A 32-state QAM signal provides sufficient capacity for a single HDTV signal, without having to resort to compression techniques outside MPEG standards, but commonly some compression techniques outside MPEG standards are employed to encode the single HDTV signal as a 16-state QAM signal. The occurence of a prescribed 24-bit word is detected by the data sync recovery circuitry 44 to generate data-field indexing information for application to the data sync selector 46. A multiplexer within the data sync selector 46 selects between the data-field indexing information respectively supplied by the data sync recovery circuitry 44 and the data sync recovery circuitry 45; the data-field indexing information thus selected is supplied to the data de-interleaver 40, the Reed-Solomon decoder 41, and the data de-randomizer 42. At the time this specification is written there is no training signal included in the QAM DTV signal. Accordingly, the amplitude-and-group-delay equalizer 36 is arranged to provide a flat amplitude-versus-frequency characteristic in response to the VSB pilot carrier presence detector 34 indicating the absence of pilot carrier, and the VSB training signal selected by the data sync recovery circuitry 45 is wired through the data sync selector 46 without need for a multiplexer. Also, there is no data line synchronization signal for QAM DTV transmission, at least not one selected as a standard. The data sync recovery circuitry 44 includes counting circuitry for counting the samples in each data field to generate intra-data-field synchronizing information. This intra-data-field synchronizing information and the intra-data-field synchronizing information (such as data line count) generated by the data sync recovery circuitry 45 are selected between by appropriate multiplexers in the data sync selector 46, for application to the data de-interleaver 40, the Reed-Solomon decoder 41, and the data de-randomizer 42, as required.

[0040] The packet sorter 43 sorts packets of data for different applications, responsive to header codes in the successive packets of data. Packets of data descriptive of the audio portions of the DTV program are applied by the packet sorter 43 to a digital sound decoder 47. The digital sound decoder 47 supplies left-channel and right-channel stereophonic sound signals to a plural-channel audio amplifier 48 that drives the plurality of loudspeakers 49, 50. Packets of data descriptive of the video portions of the DTV program are applied by the packet sorter 43 to an MPEG decoder 51, such as of MPEG-2 type. The MPEG decoder 51 supplies horizontal (H) and vertical (V) synchronizing signals to kinescope deflection circuitry 52 that provides for the raster scanning of the viewing screen of a kinescope 53. The MPEG decoder 51 also supplies signals to the kinescope driver amplifiers 54 for applying amplified red (R), green (G) and blue (B) drive signals to the kinescope 53. In variations of the DTV receiver shown in FIGS. 1 and 2, a different display device may be used instead of or in addition to the kinescope 53, and the sound recovery system may be different, consisting of but a single audio channel, or being more elaborate than a simple stereophonic reproduction system.

[0041] Referring back to FIG. 1, in order that ROMs 27 and 31 can be used to generate digital complex-number descriptions of the QAM and VSB signal carriers as translated to respective final intermediate frequencies, in response to addressing generated by counting first clock signals, provision must be made to lock the one those final intermediate frequencies that is the carrier of the currently received DTV signal to a submultiple of a multiple of the first clock signal frequency. That is, those final intermediate frequencies must be in whole number ratios with the first clock signal frequency. An automatic phase and frequency control (AFPC) signal is developed in the digital circuitry following the analog-to-digital converter 22 and is used to control the frequency and phase of one of the local oscillators 11, 16 and 20 in the tuner. Using a fixed-frequency third local oscillator 20, and controlling the frequency and phase of the oscillations the second local oscillator 16 provides, is preferred in that alignment of the second IF signal with the second SAW filter 18 can be readily assured. The second SAW filter 18 usually includes traps for adjacent-channel signal components, in which case proper alignment of the second IF signal between these traps is important for preserving its integrity. The symbol clocking is made to exhibit a high degree of frequency stability. By locking the carrier of the final intermediate-frequency (IF) signal in frequency and phase to a submultiple of a multiple of the symbol clock frequency, the AFPC for correcting frequency and phase error in the carrier as translated to a final intermediate frequency invariably operates to correct dynamic symbol phase error as well, eliminating the need for a separate phase tracker to correct dynamic symbol phase error.

[0042] FIG. 1 denominates a digital multiplexer 55 as “AFPC selector”. The multiplexer 55 responds to the pilot carrier presence detector 34 indicating that a pilot carrier is included in the currently received DTV signal for selecting, as an input signal for a digital lowpass filter 56, the imaginary output signal of the baseband response of the VSB synchrodyne circuitry 30. The response of lowpass filter 56 is a digital AFPC signal supplied as input signal to a digital-to-analog converter (DAC) 57. The output signal from the DAC 57 is an analog AFPC signal, which is subjected to further lowpass filtering in an analog lowpass filter 58, the response of which filter 58 is used for controlling the frequency and phase of the oscillations that the second local oscillator 16 provides. Analog lowpass filtering is advantageous to use for realizing long-time-constant lowpass filtering because there is reduced need for active devices as compared to digital lowpass filtering. Since the shunt capacitor of a resistance-capacitance lowpass filter section can be at the interface between a tuner 5 IC and the IC containing the digital synchrodyning circuitry, the analog lowpass filtering can be done without any cost in IC pin-out. Doing some digital lowpass filtering is advantageous, however, since the digital lowpass filter response can be subsampled to the DAC 57; the reduced speed requirements on the digital-to-analog conversion reduces the cost of the DAC 57. This procedure is similar to that used in the AGC circuitry described at the end of this specification with reference to FIG. 12 of the drawing, and the third clock signal developed for the AGC circuitry can be used by the DAC 57 and can be used to reset an accumulator the digital lowpass filter 56 includes for averaging samples of filter input signal.

[0043] The multiplexer 55 responds to the pilot carrier presence detector 34 indicating that a pilot carrier is not included in the currently received DTV signal for selecting the input signal for the digital lowpass filter 56 from the circuitry for processing a QAM DTV signal. FIG. 1 shows a digital multiplier 29 connected for supplying the digital multiplexer 55 used as “AFPC selector” an unfiltered digital AFPC input signal during the times a QAM signal is being received. The digital multiplier 29 multiplies together the real and imaginary output signals of the QAM synchrodyne circuitry 25 to generate an unfiltered digital AFPC signal. The generation of the unfiltered digital AFPC signal is very similar to that in the well-known Costas loop. In the Costas loop the AFPC signal is used to control the frequency and phase of the digital local oscillations used for synchrodyning received signals to baseband. The FIG. 1 arrangement departs from this procedure, the AFPC signal being used instead to control the frequency and phase of the analog oscillations generated by the second local oscillator 16. This regulates the frequency and phase of the final IF signal supplied to the ADC 22 for digitization and for subsequent synchrodyning to baseband in the digital regime. As is the case with the Costas loop, the multiplier 29 is preferably of especial design in which the real signal is converted to a ternary signal for multiplying the imaginary signal; this simplifies the digital multiplier and improves the pull-in characteristics of the AFPC loop.

[0044] In other embodiments of the invention the digital multiplier 29 is replaced by a more sophisticated AFPC detector in which the real output signal from the QAM synchrodyne circuitry 25 is converted to a ternary signal for multiplying the imaginary output signal from the QAM synchrodyne circuitry 25 to develop a first partial product, in which the imaginary output signal from the QAM synchrodyne circuitry 25 is converted to a ternary signal for multiplying the real output signal from the QAM synchrodyne circuitry 25 to develop a second partial product, and in which the first and second partial products are differentially combined to develop a final product. The generation of the unfiltered digital AFPC signal is similar to that described by Edward E. Lee and David G. Messerschmidt in the text DIGITAL COMMUNICATION, 2^ndEd., published in 1994 by Kluwer Academic Publishers Boston,/Dordecht/London), more particularly in “Section 16.1 Decision-Directed Carrier Recovery” with reference to FIG. 16-6 therein.

[0045] In the FIG. 1 arrangement the AFPC loop does not include an equalizer. The AFPC loop is not preceded by the passband equalizer customarily used in digital QAM receivers, however. Instead, baseband equalization is employed. The AFPC loop for the second local oscillator 16 does not have to include the equalizer, since AFPC signal can be developed from the real and imaginary components of the QAM synchrodyne circuitry 25 output signal without them having been equalized. The slicing procedures for developing ternary signals for AFPC are independent of symbol decoding, permitting soft-decision symbol decoding procedures that are more slowly completed. This arrangement operates even in the presence of multipath distortion. The resultant tight relatively-fast tracking of the second local oscillator 16 reduces the tracking requirements on the symbol synchronization circuitry when multipath distortion afflicts the received QAM signal, so the equalization filtering can be contained within the AFPC loop for symbol synchronization without encountering too much difficulty. Using baseband equalization for QAM reception simplifies the design of decision-feedback loops for adjusting the coefficients of adaptive equalization filtering.

[0046] The second intermediate-frequency amplifier 19, the third local oscillator 20 (except for its outboard crystal and other frequency selection components), and the third mixer 21 are advantageously constructed within the confines of a monolithic IC; since the output signal of the third mixer 21 is at a different frequency than the input signal to the second IF amplifier 19, the second IF amplifier 19 can have high gain without attendant high risk of unwanted regeneration. The first IF amplifier 14, the second local oscillator 16 (except for its outboard crystal and other frequency selection components) and the second mixer 17 can be constructed within the confines of the same IC, or they may be constructed otherwise—e.g., within other integrated circuitry. The analog-to-digital converter (ADC), as customary, will be a flash type with at least ten bits resolution and is preferably constructed within the confines of a different monolithic IC than the IF amplifiers. The analog lowpass filter at the input of the converter isolates the sampling circuitry, with its associated switching transients, from the IC in which the high-gain second IF amplifier 19 is located (and in some cases, in which the first IF amplifier 14 is also located). This reduces the likelihood of unwanted regeneration in the tuner 5. Considerable die area is required for the resistance ladder used in establishing the quantizing levels and for the large number of analog comparators involved in an ADC of flash type, so often such an ADC does not share a monolithic IC with other elements anyway.

[0047] The elements 23-35, 55 and 56 are advantageously constructed within the confines of a single monolithic integrated circuit (IC), to reduce the number of wiring connections made outside the confines of a monolithic IC. The synchrodyning circuits 25 and 30 both receive input signals from the real-to-complex sample converter 24, and portions of their respective address generators 28 and 32 can usually be provided by circuitry shared in common. It is advantageous that this single monolithic IC and the circuitry that follows this IC include all the circuitry for automatically selecting the appropriate mode of reception for the DTV transmission currently being received. Such practice avoids the need for operating the third local oscillator at two markedly different frequencies, depending on whether a DTV signal is of QAM type or is of VSB type. Operation of the third local oscillator at two markedly different frequencies is normally associated with the use of two different crystals for setting those frequencies. Operating the third local oscillator at essentially the same frequency, no matter whether the DTV signal is of QAM type or is of VSB type, saves the cost of the extra crystal and of the electronic switching circuitry involved with the use of two crystals. Furthermore, the reliability of the tuner 5 is improved by the reduction in the amount of circuitry located outside the monolithic integrated circuitry.

[0048] If the ADC is not constructed within an IC, all or substantially all its own, it is advantageous to include it in the IC that contains the circuitry for synchrodyning VSB DTV signals and the circuitry for synchrodyning QAM DTV signals to baseband, since the signals for clocking the sampling of the final IF signal by the ADC are to be generated within that IC. Furthermore, the analog lowpass filter at the input of the converter still isolates the sampling circuitry, with its associated switching transients, from the IC(s) in which high-gain IF amplification is done.

[0049] FIG. 3 shows in more detail the digital circuitry 25 for synchrodyning QAM DTV signals to baseband. The QAM synchrodyning circuitry 25 includes the QAM in-phase synchronous detector 250 for generating the real portion of its output signal and the QAM quadrature-phase synchronous detector 255 for generating the imaginary portion of its output signal. The QAM synchrodyning circuitry 25 includes a digital adder 256, a digital subtractor 257, and respective first, second, third and fourth digital multipliers 251-254. The QAM in-phase synchronous detector 250 includes the multiplier 251, the multiplier 252, and the adder 256 for adding the product output signals of the multipliers 251 and 252 to generate the real portion of the output signal of the QAM synchrodyning circuitry 25. The first digital multiplier 251 multiplies the real digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the cosine of the QAM carrier that are read from the look-up table 271 in the ROM 27, and the second digital multiplier 252 multiplies the imaginary digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the sine of the QAM carrier that are read from the look-up table 272 in the ROM 27. The QAM quadrature-phase synchronous detector 255 includes the multiplier 253, the multiplier 254, and the subtractor 257 for subtracting the product output signal of the multiplier 253 from the product output signal of the multiplier 254 to generate the imaginary portion of the output signal of the QAM synchrodyning circuitry 25. The third digital multiplier 253 multiplies the real digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the sine of the QAM carrier that are read from the look-up table 272 in the ROM 27, and the fourth digital multiplier 254 multiplies the imaginary digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the cosine of the QAM carrier that are read from the look-up table 271 in the ROM 27. To increase multiplication speeds the digital multipliers 251, 252, 253 and 254 may be implemented using read-only memory (ROM). This advantageously reduces delay in the AFPC loop controlling the local oscillator 16 when QAM signals are being synchrodyned to baseband.

[0050] FIG. 3 also shows in more detail the digital circuitry 30 for synchrodyning VSB DTV signals to baseband. The VSB synchrodyning circuitry 30 includes the VSB in-phase synchronous detector 300 for generating the real portion of its output signal and the VSB quadrature-phase synchronous detector 305 for generating the imaginary portion of its output signal. The VSB synchrodyning circuitry 30 includes a digital adder 306, a digital subtractor 307, and respective first, second, third and fourth digital multipliers 301-304. The VSB in-phase synchronous detector 300 includes the multiplier 301, the multiplier 302, and the adder 306 for adding the product output signals of the multipliers 301 and 302 to generate the real portion of the output signal of the VSB synchrodyning circuitry 30. The first digital multiplier 301 multiplies the real digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the cosine of the VSB carrier that are read from the look-up table 311 in the ROM 31, and the second digital multiplier 302 multiplies the imaginary digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the sine of the VSB carrier that are read from the look-up table 312 in the ROM 31. The VSB quadrature-phase synchronous detector 305 includes the multiplier 303, the multiplier 304, and the subtractor 307 for subtracting the product output signal of the multiplier 303 from the product output signal of the multiplier 304 to generate the imaginary portion of the output signal of the VSB synchrodyning circuitry 30. The third digital multiplier 303 multiplies the real digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the sine of the VSB carrier that are read from the look-up table 312 in the ROM 31, and the fourth digital multiplier 304 multiplies the imaginary digital samples of final IF signal supplied from the real-to-complex-sample converter 24 by digital samples descriptive of the cosine of the VSB carrier that are read from the look-up table 311 in the ROM 31. To increase multiplication speeds the digital multipliers 301, 302, 303 and 304 may be implemented using read-only memory (ROM). This advantageously reduces delay in the AFPC loop controlling the local oscillator 16 when VSB signals are being synchrodyned to baseband.

[0051] FIG. 4 shows in detail a representative construction of the sample clock generator 23. This construction includes a voltage-controlled oscillator 230 that generates cissoidal oscillations nominally of 21.52 MHz frequency. The oscillator 230 is a controlled oscillator, the frequency and phase of its oscillations being controlled by an automatic frequency and phase control (AFPC) signal voltage. This AFPC signal voltage is generated by an automatic frequency and phase control (AFPC) detector 231, which compares frequency-divided response to the oscillations of the oscillator 230 with a 10.76 MHz reference carrier supplied via an analog 10.76 MHz bandpass filter 232. Preferably, oscillator 230 is of a type using a crystal for stabilizing the natural frequency and phase of its oscillations. A symmetrical clipper or limiter 233 generates an essentially squarewave response to these cissoidal oscillations, which is used as the first clock signal for timing the sampling of the final IF signal in the ADC 22. A frequency-divider flip-flop 234 responds to transitions of the first clock signal in a prescribed sense for generating another square wave with a fundamental frequency of 10.76 MHz, half the frequency of the oscillations of the oscillator 230. This frequency-divided response to the oscillations of the oscillator 230 is supplied to the AFPC detector 231 for comparison with the 10.76 MHz reference carrier supplied via the 10.76 MHz bandpass filter 232. The frequency-divider flip-flop 234 also supplies squarewave output signal with a fundamental frequency of 10.76 MHz to an AND circuit 235 to be ANDed with the first clock signal for generating a second clock signal used by the 2:1 decimator 35 shown in FIG. 1.

[0052] The AFPC signal voltage for controlling the VCO 230 is generated by dividing the frequency of the 21.52 MHz reference carrier supplied from the VCO 230 to obtain a subharmonic thereof for comparison with a signal generated by frequency multiplier circuitry. A component of the received DTV signal as synchrodyned to baseband, which component is of a frequency that is a subharmonic of the symbol frequency (or baud frequency), is extracted by frequency-selective filtering. The frequency multiplier circuitry multiplies the frequency of that subharmonic of the symbol frequency by an appropriate factor to generate the signal for comparison with the subharmonic of the 21.52 MHz reference carrier. The details of this procedure will now be specifically described, first presuming the received DTV signal is a VSB signal with a baud rate of 10.76*10⁶symbols per second, and then presuming the received DTV signal is a QAM signal with a baud rate of 5.38*10⁶symbols per second.

[0053] A digital multiplexer 236 responds to the pilot carrier presence detector 34 detecting pilot carrier accompanying the received DTV signal, which is indicative that the received DTV signal is a VSB signal, to select the real samples of this signal supplied from the VSB in-phase synchronous detector 300 for application to a bandpass FIR digital filter 237 that provides a selective response centered at 5.38 MHz, which selects the first subharmonic of symbol frequency from the VSB signal. Further frequency multiplication of the 5.38 MHz is performed in the analog regime to avoid undersampling problems that arise when attempting to perform further frequency multiplication in the digital regime. An digital-to-analog converter (DAC) 238 converts the filter 237 response to analog form for application to full-wave rectification circuitry 239, which generates harmonics of the filter 237 response including a strong 10.76 MHz component as second harmonic of 5.38 MHz. The analog bandpass filter 232 responds to this 10.76 MHz second harmonic to supply the AFPC detector 231 with 10.76 MHz reference carrier input signal.

[0054] The digital multiplexer 236 responds to the pilot carrier presence detector 34 not detecting pilot carrier accompanying the received DTV signal, which is indicative that the received DTV signal is a QAM signal, to select the output signal of a squaring circuit 23A for application to the bandpass filter 237 that provides a selective response centered at 5.38 MHz. A bandpass FIR digital filter 23B that provides a selective response centered at 2.69 MHz for selecting the 2.69 MHz first subharmonic of the symbol frequency of a baseband QAM signal supplies input signal to the squaring circuit 23A, which generates harmonics of the filter 23B response including a strong 5.38 MHz component. This baseband QAM signal can be supplied either from the QAM in-phase synchronous detector 40, as shown in FIG. 4, or from the QAM quadrature-phase synchronous detector 45.

[0055] The squaring circuit 23A is shown in FIG. 7 as a digital multiplier receiving the filter 23B response both as multiplier and multiplicand. The squaring circuit 23A can be constructed from logic gates as a digital multiplier, but for the sake of speedier operation is better provided by a ROM storing a look-up table of squares. An absolute-value circuit can be used as a substitute for the squaring circuit in generating harmonics of the response of a preceding filter, but produces weaker second harmonics and so is not preferred.

[0056] FIG. 4 also shows in more detail a representative construction of the first address generator 28, which supplies addresses to a cosine look-up table portion 271 and a sine look-up table portion 272 of the ROM 27 that provides complex-number digital descriptions of two phasings of the QAM carrier, as translated to a final intermediate frequency and in quadrature relationship with each other. Transitions of the first clock signal are counted by a first address counter 281 in the first address generator 28 to generate a basic first address signal. This basic first address signal is applied as a first summand to a digital adder 282. A first address correction signal, which is applied to the adder 282 as a second summand, adds to the basic first address signal in the adder 282 for generating as a sum output signal a corrected first address signal for addressing both the cosine look-up table portion 271 and the sine look-up table portion 272 of the ROM 27. A symbol-clock-rotation detector 283 responds to the sequence of real samples of QAM signal as synchrodyned to baseband by the QAM in-phase synchronous detector 250 and to the sequence of imaginary samples of QAM signal as synchrodyned to baseband by the QAM quadrature-phase synchronous detector 255. The symbol-clock-rotation detector 283 detects the misphasing between symbol clocking done at the receiver in accordance with the first clock signal and symbol clocking done at the transmitter, as evidenced in the received QAM signal heterodyned to a final intermediate frequency that is a submultiple of its symbol frequency. Several types of symbol-clock-rotation detector 283 are described and background literature describing certain of them are catalogued in U.S. Pat. No. 5,115,454 issued May 19, 1992 to A. D. Kucar, entitled METHOD AND APPARATUS FOR CARRIER SYNCHRONIZATION AND DATA DETECTION, and incorporated herein by reference. A digital lowpass filter 284 averages over many samples (e. g., several million) the misphasing of the symbol clocking done at the receiver as detected by the symbol-clock-rotation detector 283 to generate the first address correction signal supplied to the adder 282 to correct the basic first address. Averaging over so many samples can be done by procedures which accumulate lesser numbers of samples and dump them forward at a reduced sample rate for further accumulation, accumulation and subsampling being repeated a few times with progressively lower subsampling rates.

[0057] FIG. 4 also shows in more detail a representative construction of the second address generator 32, which supplies addresses to a cosine look-up table portion 311 and a sine look-up table portion 312 of the ROM 31 that provides complex-number digital descriptions of two phasings of the VSB carrier, as translated to a final intermediate frequency and in quadrature relationship with each other. Transitions of the first clock signal are counted by a second address counter 321 in the second address generator 32 to generate a basic second address signal. This basic second address signal is applied as a first summand to a digital adder 322. A second address correction signal, which is applied to the adder 322 as a second summand, adds to the basic second address signal in the adder 322 for generating as a sum output signal a corrected second address signal for addressing both the cosine look-up table portion 311 and the sine look-up table portion 312 of the ROM 31.

[0058] FIG. 4 shows a clocked digital delay line 323 for delaying the samples from the in-phase synchronous detector 300 by a prescribed number of sample periods prior to their being applied as input signal to a quantizer 324, which supplies the quantization level most closely approximated by the sample currently received by the quantizer 324 as input signal. The quantization levels can be inferred from the energy of the pilot carrier accompanying the VSB signal or can be inferred from the result of envelope detection of the VSB signal. The closest quantization level selected by the quantizer 324 as its output signal has the corresponding quantizer 324 input signal subtracted therefrom by a digital adder/subtractor 325, which is operated as a clocked element by including a clocked latch at its output. The difference output signal from the adder/ subtractor 325 describes the departure of the symbol levels actually recovered from those that should be recovered, but whether the polarity of the departure is attributable to symbol misphasing being leading or lagging remains to be resolved.

[0059] The samples from the in-phase synchronous detector 300 applied as input signal to the clocked digital delay line 323 are applied without delay as input signal to a mean-square-error gradient detection filter 326. The filter 326 is a finite-impulse-response (FIR) digital filter having a (−½), 1, 0, (−1), (+½) kernel, the operation of which is clocked by the first sampling clock. The prescribed number of sample periods of delay provided by the clocked digital delay line 323 is such that filter 326 response is in temporal alignment with the difference signal from the adder/subtractor 325. A digital multiplier 327 multiplies the difference signal from the adder/subtractor 325 by the filter 326 response to resolve this issue. The sign bit and the next most significant bit of the two's complement filter 326 response suffice for the multiplication, which permits simplification of the digital multiplier 327 structure. The samples of the product signal from the digital multiplier 327 are indications of the misphasing of the symbol clocking done at the receiver that are averaged over many samples (e. g., several million) by a digital lowpass filter 328 for generating the second address correction signal supplied to the adder 322 to correct the basic second address.

[0060] The symbol synchronization techniques used in the second address generator 32 shown FIG. 4 are of the same general type as S. U. H. Qureshi describes for use with pulse amplitude modulation (PAM) signals in his paper “Timing Recovery for Equalized Partial-Response Systems”, IEEE Transactions on Communications, Dec. 1976, pp. 1326-1330. These symbol synchronization techniques as used in connection with symbol synchronization for VSB signals are specifically described by C. B. Patel et alii in U.S. Pat. No. 5,548,617 issued Aug. 20, 1996, entitled DIGITAL VSB DETECTOR WITH BANDPASS PHASE TRACKER USING RADER FILTERS, AS FOR USE IN AN HDTV RECEIVER, and incorporated herein by reference.. In preferred designs of the general type of second address generator 32 shown FIGS. 4 and 5, the clocked digital delay line 323 does not exist as a separate element; instead, an input signal to the quantizer 324 with the requisite number of sample periods of delay for the difference signal from the adder/subtractor 325 being temporally aligned with the filter 326 response is taken from the tapped digital delay line included in the filter 326 for supplying differentially delayed samples to be weighted by the (−½), 1, 0, (−1), (+½) kernel before being summed to generate the filter 326 response.

[0061] The carrier of a QAM DTV signal and the carrier of a VSB DTV signal are translated to respective final intermediate frequencies that are at a 2.69 MHz remove from each other, since the carrier of the QAM DTV signal is at the center of a 6-MHz-wide TV channel, but the carrier of the VSB DTV signal is only 310 kHz above the lowest frequency of a 6-MHz-wide TV channel. The frequencies of the local oscillators 11, 16 and 20 in the tuner 5 of FIG. 1 can be chosen so that the intermediate frequency to which the carrier of a VSB DTV signal is translated is higher than that to which the carrier of a QAM DTV signal is translated, with the vestigial and full sidebands of the VSB DTV signal being respectively above and below its carrier. Alternatively, the frequencies of the local oscillators 11, 16 and 20 can be chosen so that the intermediate frequency to which the carrier of a VSB DTV signal is translated is lower than that to which the carrier of a QAM DTV signal is translated, with the vestigial and full sidebands of the VSB DTV signal being respectively below and above its carrier.

[0062] Preferably the lowest frequency of the final IF signal is above 1 MHz, to keep the ratio of the highest frequency of the final IF signal thereto substantially below 8:1 and thereby ease the filtering requirements for the real-to-complex-sample converter 24. To satisfy this preference in regard to the QAM signal alone, the lowest carrier frequency for the QAM carrier in the final IF signal is 3.69 MHz. To satisfy this preference in regard to the VSB signal alone, the lowest the carrier frequency for the VSB carrier in the final IF signal could be is 1.31 MHz, presuming its full sideband to be above its vestigial sideband in frequency, or 6.38 MHz, presuming its full sideband to be below its vestigial sideband in frequency. Presuming the full sideband of the VSB signal to be above its vestigial sideband in frequency, since the carrier frequency of the VSB carrier most be at least 1.31 MHz, the carrier frequency of the QAM carrier must be at least 4.00 MHz. Presuming the full sideband of the VSB signal to be below its vestigial sideband in frequency, since the carrier frequency of the VSB carrier most be at least 6.38 MHz, the carrier frequency of the QAM carrier must still be at least 3.69 MHz.

[0063] If the sample rate in the ADC 22 is established by the first clock signal from the sample clock generator 23 to be 21.52*10⁶samples per second, preferably the intermediate frequency to which the carrier of a QAM DTV signal is translated is not higher than 5.38 MHz, so that it can be sampled at least four times per cycle. Presuming the full sideband of the VSB signal to be above its vestigial sideband in frequency, this preference constrains the lowest frequency in the final IF signal to being no higher than 2.38 MHz and the carrier of the VSB signal being no higher than 2.69 MHz. FIG. 11 illustrates how, for these presumed conditions, the VSB carrier is constrained to the band 1.31-2.69 MHz, and the QAM carrier is constrained to the band 4.00-5.38 MHz.

[0064] Presuming the full sideband of the VSB signal to be below its vestigial sideband in frequency, the QAM carrier is constrained to the band 3.69-5.38 MHz. Accordingly, the carrier of the VSB signal is constrained to the band 6.38-8.07 MHz in order that the 2.69 MHz offset between carriers is maintained. FIG. 12 illustrates the case where the QAM carrier is constrained to the band 3.69-5.38 MHz and the VSB carrier is constrained to the band 6.38-8.07 MHz.

[0065] The final intermediate frequency to which the QAM carrier is translated must be a submultiple of a multiple of the 21.52 MHz sampling rate in order that this carrier can be described on a continuous basis relying on a sine-cosine look-up table in the ROM 27. The final intermediate frequency to which the VSB carrier is translated must be a submultiple of a multiple of the 21.52 MHz sampling rate in order that this carrier can be described on a continuous basis relying on a sine-cosine look-up table in the ROM 31. The final intermediate frequency (m/n) times the 21.52 MHz sampling rate, to which the carrier is translated, preferably has a small value of n, to keep the number of values in the sine-cosine look-up tables stored in ROM reasonably small.

[0066] One can search for respective intermediate frequencies to which the carrier of a QAM DTV signal and the carrier of a VSB DTV signal are to be translated, which frequencies meet the criteria set forth above, by following procedure taught in U.S. Pat. No. 5,506,636. A table of subharmonics of successive harmonics of the 10.76 MHz VSB symbol rate, which the sampling clock rate is harmonically related to, is constructed for the frequency ranges of interest. Then pairs of subharmonics of the same harmonic which exhibit the desired 2.69 MHz difference in frequency between them are considered with regard to their relative advantages as carriers.

[0067] The fourth and eighth subharmonics of 21.52 MHz at 5.38 MHz and at 2.39 MHz exhibit substantially the desired 2.69 MHz offset and are appropriate for use as QAM carrier and a VSB carrier with its full sideband above its vestigial sideband in frequency. The 2.69 MHz offset between these subharmonics is one-quarter the symbol rate or 2,690,559.4 Hz, rather than 2,690122.4 Hz offset between the QAM and VSB carriers required to offset the VSB carrier from a co-channel interfering NTSC video carrier by 59.75 times the nominal NTSC horizontal scanning frequency during color transmissions. This small 437 Hz frequency discrepancy is easily accommodated by the automatic frequency and phase control of the controlled local oscillator 16 in the tuner 5 of FIG. 1. The addressing of ROMs 27 and 31 is greatly simplified when the QAM and VSB carriers are translated to be close to the fourth and eighth subharmonics of 21.52 MHz in final IF signals, since advantage can be taken of repetitive symmetries in the stored sine and cosine functions, to reduce ROM address bitwidth.

[0068] The second harmonic of the 21.52 MHz sampling frequency is 43.05 MHz, and its subharmonics can be searched, looking for a pair offset from each other in frequency by an amount substantially equal to 2.69 MHz. The eighth and sixteenth subharmonics of 43.05 MHz are the fourth and eighth subharmonics of 21.52 MHz which have already been considered. The tenth and twenty-seventh subharmonics of 43.05 MHz at 4.305 MHz and at 1.594 MHz exhibit a 20 kHz or 0.74% error in regard to the desired 2.69 MHz offset and could respectively serve as QAM carrier and as VSB carrier. This error is within the 30 kHz or so mistuning tolerated in past commercial designs for NTSC TV receivers. The ROM 31 storing sine-cosine look-up tables for the twenty-seventh subharmonic of 43.05 MHz has to store an excessive number of samples, however; and the ROM 27 storing sine-cosine look-up tables for the tenth subharmonic of 43.05 MHz has to store an appreciable number of samples, too.

[0069] The third harmonic of the 21.52 MHz sampling frequency is 64.57 MHz, and its subharmonics can be searched, looking for a subharmonic offset in frequency from a subharmonic of 43.05 MHz or from another subharmonic of 64.57 MHz by an amount substantially equal to 2.69 MHz. The thirteenth subharmonic of 64.57 MHz, 4.967 MHz, and the nineteenth subharmonic of 43.05 MHz, 2.265 MHz, exhibit a 12 kHz or 0.45% error in regard to the desired 2.69 MHz offset and could respectively serve as QAM carrier and as a VSB carrier with its full sideband above its vestigial sideband in frequency. This error is well within the 30 kHz or so mistuning tolerated in past commercial designs for NTSC TV receivers. However, the ROM 27 storing sine-cosine look-up tables for the thirteenth subharmonic of 64.57 MHz has to store an excessive number of samples; and the ROM 31 storing sine-cosine look-up tables for the nineteenth subharmonic of 43.05 MHz has to store an excessive number of samples, too.

[0070] The eighth subharmonic of 64.57 MHz is 8.07 MHz, offset almost exactly the desired 2.69 MHz from the fourth subharmonic of 21.52 MHz. This fourth subharmonic of 21.52 MHz, 5.38 MHz, and the eighth subharmonic of 64.57 MHz, 8.07 MHz, are appropriate for use as QAM carrier and a VSB carrier with its full sideband below its vestigial sideband in frequency.

[0071] It appears preferable that the frequencies of the local oscillators 11, 16 and 20 in the tuner 5 of FIG. 1 be chosen so that the intermediate frequency to which the carrier of a QAM DTV signal is translated is 5.38 MHz, the presumed symbol rate for the QAM DTV signal and half the standard symbol rate for the VSB DTV signal. Accordingly, if the VSB carrier is translated in frequency so as to have its full sideband above its vestigial sideband in frequency in the final IF signal, the preferred frequency of the VSB carrier in the final IF signal is 2.69 MHz. Alternatively, if the VSB carrier is translated in frequency so as to have its full sideband below its vestigial sideband in frequency in the final IF signal, the preferred frequency of the VSB carrier in the final IF signal is 8.07 MHz.

[0072] It is noted in passing that all the subharmonics of 43.05 MHz and all the subharmonics of 64.57 MHz are subharmonics of 129.15 MHz, the third harmonic of 43.05 MHz and the second harmonic of 64.57 MHz. The 2.69 MHz, 5.68 MHz and 8.07 MHz frequencies are the forty-eighth, twenty-fourth and sixteenth subharmonics, respectively, of 129.15 MHz. It is also noted that while the harmonic relationship between carriers have been considered in terms of harmonics of the 21.52 MHz sampling rate that is the second harmonic of the 10.76 MHz VSB symbol rate, the consideration thusfar can be viewed as involving the even harmonics of the 10.76 MHz symbol rate. A more complete consideration of the possible harmonic relationships between carriers also includes consideration of odd harmonics, at least third, of the 10.76 MHz VSB symbol rate. The 2.69 MHz, 5.68 MHz and 8.07 MHz frequencies are respectively the twelfth, sixth and fourth subharmonics of 32.29 MHz, 32.29 MHz being three times the 10.76 MHz symbol rate of the VSB signal.

[0073] One skilled in the art of designing analog-to-digital conversion circuitry for digital systems will appreciate that the sampling of analog signals for digitization can use various widths of sampling window. Thusfar, it has been presumed that 21.52*10⁶samples per second are taken with the duration of each sampling window extending over half a cycle of 21.52 MHz. The pulses from the limiter 233 can be stretched to nearly twice this duration, if desired. Another alternative that is possible is to design the analog-to-digital converter to use two staggered sets of sampling windows with each sampling window extending over half a cycle of 21.52 MHz and to digitize on a staggered-phase basis at a 43.05*10⁶samples per second combined rate. The digitization of final IF signal at a 43.05*10⁶samples per second improves automatic phase and frequency control accuracy.

[0074] FIG. 5 shows a modification of the FIG. 4 circuitry that is possible when the fourth and the eighth subharmonics of 21.52 MHz are used as the final intermediate frequencies to which the QAM and VSB DTV carriers are respectively converted. In a modification 320 of the second address generator 32 described above, second address counter 321 is arranged to count modulo eight when sampling rate is 21.52*10⁶samples per second, thereby to generate two cycles of ROM 27 addressing and the one cycle of addressing for a ROM 310 that replaces the ROM 31; and the less significant bits of the output count from the second address counter 321 are made available for replacing the basic first address from the first address counter 281. In a modification 280 of the first address generator 28 described above, the first address counter 281 is dispensed with, and the less significant bits of the second address counter 321 are applied to the adder 282 as basic first address instead of the count from the first address counter 281. The VSB complex carrier ROM 31 is replaced with a ROM 310 that comprises a portion 313 that stores only one-half cycle of VSB carrier cosine values and a portion 314 that stores only one-half cycle of VSB carrier sine values. These portions 313 and 314 of the ROM 310 are addressed by the less significant bits of the adder 322 sum output signal. A selective bits complementor 315 exclusive-ORs the most significant bit of the adder 322 sum output signal with each of the bits of the VSB carrier cosine values read from the portion 313 of the ROM 310 for generating a first summand input for a digital adder 317, and the most significant bit of the adder 322 sum output signal is provided with zero extension in the direction of increased significance for generating a second summand input for the adder 317. The sum output from the adder 317 provides the full cycle of VSB carrier cosine values over eight first clock periods. A selective bits complementor 316 exclusive-ORs the most significant bit of the adder 322 sum output signal with each of the bits of the VSB carrier sine values read from the portion 314 of the ROM 310 for generating a first summand input for a digital adder 318, and the most significant bit of the adder 322 sum output signal with zero extension in the direction of increased significance is also applied as a second summand input for the adder 318. The sum output from the adder 318 provides the full cycle of VSB carrier sine values over eight first clock periods.

[0075] The FIG. 5 of the FIG. 4 circuitry can also be used when the sixth and fourth subharmonic of 32.29 MHz are used as the final intermediate frequencies to which the QAM and VSB DTV carriers are respectively converted. The contents of the portions 313 and 314 of the ROM 310 are modified for the higher-frequency 8.07 MHz VSB carrier, of course.

[0076] One skilled in the art of digital circuit design will understand that other hardware savings can be made in the FIG. 4 read-only memory circuitry taking advantage of symmetries in the cosine and sine functions or the 900 offset in the respective phases of these two functions. FIG. 6 shows a form that the circuitry 24 can take, which comprises:

[0077] (a) a linear-phase, finite-impulse-response (FIR) digital filter 60 that generates imaginary (Im) digital samples as a Hilbert transform response to the real (Re) digital samples; and

[0078] (b) compensating, clocked digital delay of the real digital samples to compensate for the latency time of the Hilbert transformation filter 60, which clocked digital delay can be provided by clocked latch elements 61-66 included in the Hilbert transformation filter 60.

[0079] The use of such circuitry for implementing in-phase and quadrature-phase sampling procedures on bandpass signals is described by D. W. Rice and K. H. Wu in their article “Quadrature Sampling with High Dynamic Range” on pp. 736-739 of IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS, Vol. AES-18, No. 4 (November 1982). Since the frequency band 6 MHz wide occupied by the final IF signal has a lowest frequency of at least a megaHertz or so, it is possible to use as few as seven non-zero-weighted taps in the FIR filter 60 used for Hilbert transformation.

[0080] The seven-tap Hilbert transformation filter 60 includes a cascade connection of one-sample delay elements 61, 62, 63, 64, 65 and 66 from which samples taken to be weighted and summed to generate the Hilbert transform response. The Hilbert transform is linear phase in nature so the tap weights of the FIR filter 60 exhibit symmetry about median delay. Accordingly, a digital adder 67 sums the input signal to delay element 61 and the output signal from the delay element 66 to be weighted in common, a digital adder 68 sums the output signal from the delay element 61 and the output signal from the delay element 65 to be weighted in common, and a digital adder 69 sums the output signal from the delay element 62 and the output signal from the delay element 64 to be weighted in common. The output signal from the delay element 64 is applied as input address to a read-only memory 70, which multiplies that signal by an appropriate weight W₀magnitude. The sum output signal from the digital adder 69 is applied as input address to a read-only memory 71, which multiplies that signal by an appropriate weight W₀magnitude. The sum output signal from the digital adder 68 is applied as input address to a read-only memory 72, which multiplies that signal by an appropriate weight W₂magnitude. The sum output signal from the digital adder 67 is applied as input address to a read-only memory 73, which multiplies that signal by an appropriate weight W₃magnitude. The use of the ROMs 70, 71, 72 and 73 as fixed-multiplicand multipliers keeps the delay associated with multiplication negligibly short. The output signals of the ROMs 70, 71, 72 and 73 are combined by a tree of signed digital adders 74, 75 and 76 operated as adders or subtractors, as required to appropriately assign signs to the magnitudes of the weights W₀, W₁W₂and W₃stored in the ROMs 70, 71, 72 and 73. The adders 67, 68, 69, 74, 75 and 76 are assumed to be clocked adders each exhibiting one-sample latency, which results in the seven-tap FIR filter 60 exhibiting a six-sample latency. Delay of the filter 60 input signal that compensates for this latency is provided by the cascade connection of the six one-sample delay elements 61, 62, 63, 64, 65 and 66. The input address to the read-only memory 70 is taken from the output of the delay element 64, rather than from the output of the delay element 63, so the one-sample delay of delay element 64 compensates for the one-sample delays in the adders 67, 68 and 69.

[0081] C. M. Rader in his article “A Simple Method for Sampling In-Phase and Quadrature Components”, IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS, Vol. AES-20, No. 6 (November 1984), pp. 821-824, describes improvements in complex synchronous detection carried out on digitized bandpass signals. Rader replaces the Hilbert-transform FIR filter and the compensating-delay FIR filter of Rice and Wu with a pair of all-pass digital filters designed based on Jacobian elliptic functions and exhibiting a constant π/2 difference in phase response for the digitized bandpass signals. A preferred pair of such all-pass digital filters, which are of infinite-impulse-response (IIR) type, has the following system functions: $\begin{array}{cc}{H}_1\left(z\right)=z^{-1}\left(z^{-2}-a^2\right)/\left(1-a^2{z}^{-2}\right)& a^2=0.5846832\\ H_2\left(z\right)=-\left(z^2-b^2\right)/\left(1-b^2{z}^2\right)& b^2=0.1380250\end{array}$

[0082] Rader describes filter configurations which require only two multiplications, one by a²and one by b².

[0083] FIG. 7 shows an alternative form that the circuitry 24 can take, which comprises a pair of all-pass digital filters 80 and 90 of a type described by C. M. Rader and designed based on Jacobian elliptic functions. The filters 80 and 90 exhibit a constant π/2 difference in phase response for digitized bandpass signals. Since oversampled real samples better provide for symbol synchronization when synchrodyning VSB signals, it is preferable not to use the all-pass filters described by Rader that exploit sub-sampling to provide further reductions in the delay network circuitry.

[0084] The construction of the filter 80, which provides the system function H₁(z)=z⁻¹(z⁻²−a²)/(1−a²z⁻²), where a²=0.5846832 in decimal arithmetic, is shown in FIG. 7 to be as follows. The samples from the ADC 22 are delayed by one ADC sample clock duration in a clocked delay element 88 for application to a node 89. The signal at node 89 is further delayed by two ADC sample clock durations in cascaded clocked delay elements 81 and 82, for application as first summand signal to a digital adder 83. The sum output signal of the adder 83 provides the real response from the filter 80. The sum output signal of the adder 83 is delayed by two ADC sample clock durations in cascaded clocked delay elements 84 and 85, for application as minuend input signal to a digital subtractor 86 that receives the signal at node 89 as its subtrahend input signal. The resulting difference output signal from the digital subtractor 86 is supplied as multiplier input signal to a digital multiplier 87 for multiplying an a²multiplicand signal, using a binary arithmetic. The resulting product output signal is applied to the digital adder 83 as its second summand signal.

[0085] The construction of the filter 90, which provides the system function −H₂(z)=(z⁻²−b²)/(1−b²Z⁻²), where b²=0.1380250 in decimal arithmetic, is shown in FIG. 7 to be as follows. The samples from the ADC 22 are delayed by two ADC sample clock durations in cascaded clocked delay elements 91 and 92, for application as first summand signal to a digital adder 93. The sum output signal of the adder 93 provides the imaginary response from the filter 90. The sum output signal of the adder 93 is delayed by two ADC sample clock durations in cascaded clocked delay elements 94 and 95, for application to a digital subtractor 96 as its minuend signal. The subtractor 96 receives the samples from the ADC 22 as its subtrahend input signal. The resulting difference output signal from the digital subtractor 96 is supplied as multiplier input signal to a digital multiplier 97 for multiplying a b²multiplicand signal, using a binary arithmetic. The resulting product output signal is applied to the digital adder 93 as its second summand signal.

[0086] FIG. 8 shows a complex-signal filter resulting from modifying the FIG. 7 complex-signal filter as follows. The position of the clocked delay element 88 is shifted so as to delay the sum output signal of the adder 83, rather than to delay the digital output signal of the ADC 22, and the digital output signal of the ADC 22 is applied to the node 89 without delay, thereby to cause real response to be provided at the output port of the shifted-in-position clocked delay element 88. The real response provided at the output port of the shifted-in-position clocked delay element 81 is the same as the response provided at the output port of the clocked delay element 84. So, the real response is provided from the output port of the clocked delay element 84 instead of from the output port of the shifted-in-position clocked delay element 81; and the shifted-in-position clocked delay element 81, being no longer required, is dispensed with.

[0087] FIG. 9 shows a complex-signal filter resulting from modifying the FIG. 8 complex-signal filter as follows. The first summand signal for the adder 83 is then taken from the cascaded clocked delay elements 91 and 92, rather than from the cascaded clocked delay elements 81 and 82. The cascaded clocked delay elements 81 and 82, being no longer required, are dispensed with. The FIG. 9 complex-signal filter is preferred over the complex-signal filters of FIG. 7 and 8 in that redundant clocked delay elements are eliminated.

[0088] FIG. 10 is a detailed block schematic diagram of a complex-signal filter developing a constant π/2 difference in phase between a real response Re and an imaginary response Im to the digitized bandpass signals, that is similar the complex-signal filter described by T. F. S. Ng in United Kingdom patent application 2 244 410 A published Nov. 27, 1991 and entitled QUADRATURE DEMODULATOR. The Ng filters are finite-impulse-response (FIR) digital filters, rather than IIR filters as described by Rader. The FIG. 10 complex-signal filter differs from the filters described by Ng in that 2:1 decimation is done following filtering, rather than before.

[0089] This permits the real and imaginary filtering to be supported by a shared tapped delay line. FIG. 10 shows this shared tapped delay line composed of cascaded single-clock-delay elements 100-114, such as latches that like the ADC 22 are clocked at four times symbol transmission rate. The single-clock-delay element 100 may be dispensed with or subsumed into the ADC 22 in some designs. Digital adders and subtractors in the FIG. 6 complex filter are assumed to be clocked at four times symbol transmission rate, with each having a single-clock-duration latency. The digital multipliers are assumed to be a wired place shift in the case of a multiplication by an integral power of two or to be provided from read-only memory (ROM), so there is zero latency in each of the multiplications insofar as clocked operation is concerned. At least the eight-bit resolution in the filter results per Ng is presumed.

[0090] In order to generate the real response H₁(z), the real-response filter is presumed to apply tap weights W₀=4, W₁=0, W₂=−12, W₃=−72, W₄=72, W₅=12, W₆=0 and W₇=−4 per the example described by Ng. The real-response filter, in addition to the single-clock-delay elements 100-114, includes a digital subtractor 121 for subtracting the response of the delay element 114 from the response of the delay element 100, a digital multiplier 122 for weighting the differential response of the subtractor 121 by a factor of four, a digital subtractor 125 for subtracting the response of the delay element 103 from the response of the delay element 109, a digital multiplier 126 for weighting the differential response of the subtractor 125 by a factor of twelve, a digital subtractor 127 for subtracting the response of the delay element 105 from the response of the delay element 107, a digital multiplier 128 for weighting the differential response of the subtractor 127 by a factor of seventy-two, a digital adder 129 for summing the products from the digital multipliers 126 and 128, a digital adder 130 for summing the product from the digital multiplier 122 with the sum output signal from the adder 129, and a 2:1 decimator 131 for generating the real filter response Re in decimated response to the sum output signal from the adder 130.

[0091] The subtractor 121 subtracts the response of the delay element 114 from the response of the delay element 100, rather than subtracting the response of the delay element 113 from the output signal of the ADC 22, to introduce single-clock-duration delay to compensate for the latency of the adder 129. Since W₁=0 and W₆=0, there is no digital subtractor 123 for subtracting the response of the delay element 111 from the response of the delay element 101 or digital multiplier 124 for weighting the differential response of the subtractor 123. Consequently, there is no digital adder for summing product from the multiplier 124 with the product from the multiplier 122. This gives rise to the need to compensate for the latency of the adder 129.

[0092] In order to generate the imaginary response H₁(z), the imaginary-response filter is presumed to apply tap weights W₈=8, W₉=14, W,₁₀=22, W₁₁=96, W₁₂=22, W₁₃=14, W₁₄=8 corrected from the example described by Ng. The imaginary-response filter, in addition to the single-clock-delay elements 100-112, includes a digital adder 141 for adding the response of the delay element 112 with the response of the delay element 100, a digital multiplier 142 for weighting the sum response of the adder 141 by a factor of eight, a digital adder 143 for adding the response of the delay element 110 with the response of the delay element 102, a digital multiplier 144 for weighting the sum response of the adder 143 by a factor of fourteen, a digital adder 145 for adding the response of the delay element 108 with the response of the delay element 104, a digital multiplier 146 for weighting the sum response of the adder 145 by a factor of twenty-two, a digital multiplier 147 for weighting the response of the delay element 107 by a factor of ninety-six, a digital adder 148 for summing the products from the digital multipliers 142 and 144, a digital adder 149 for summing the products from the digital multipliers 146 and 147, a digital adder 150 for summing the sum output signals from the adders 148 and 149, and a 2:1 decimator 151 for generating the imaginary filter response Im in decimated response to the sum output signal from the adder 150.

[0093] The digital multiplier 147 weights the response of the delay element 107 by a factor of ninety-six, rather than the response of the delay element 106, in order to introduce single-clock-duration delay to compensate for the single-clock-duration latency of each of the adders 141143 and 145.

[0094] Television engineers are currently considering using the digital transmission system for HDTV for transmitting other types of television signals—for example, four television signals with resolution similar to present-day NTSC signals that are simultaneously transmitted. The invention is suitable for use in receivers for these alternative transmission schemes, and the claims which follow should be construed broadly enough to include such receivers within their scope.

[0095] In the claims which follow, the word “said” is used whenever reference is made to an antecedent, and the word “the” is used for grammatical purposes other than to refer back to an antecedent.