MXPA00012800A - Hdtv channel equalizer - Google Patents

Abstract

An adaptive channel equalizer (50) for processing a demodulated VSB signal containing terrestrial broadcast high definition television information includes an adaptive feed forward filter (FFF 20) and an adaptive decision feedback filter (DFF 30). The equalizer is preceded by a demodulator/carrier recovery network (18) and does not include a carrier recovery function in the equalizer control loop. The equalizer FFF and DFF elements operate adaptively in blind, training, and decision-directed modes. A direct connection provided from the equalizer output to the DFF facilitates coarse signal acquisition and equalization during the blind operating mode.

Description

HDTV CHANNEL EQUALIZER FIELD OF THE INVENTION This invention relates to the adaptive equalization of a video signal transmission channel that may contain high definition television information.

BACKGROUND OF THE INVENTION The recovery of data from modulated signals that carry digital information in the form of a symbol, usually requires three functions in the receiver: time recovery for symbol synchronization, carrier recovery (frequency demodulation in the base band) and a channel equalization. The recovery of time is a process by which the receiver clock (time base) is synchronized with the transmitter clock. This allows the received signal to be sampled at an optimal point in time to reduce the possibility of a section error associated with the decision-driven processing of the received symbol values. Carrier recovery is a process by which a received RF signal, after having been converted at a downward frequency to a lower intermediate frequency bandpass (eg, a nearby bandstand), is increased in frequency towards the base band to allow the retrieval of base band modulated information.

Many digital data communications systems employ adaptive equalization to compensate for the effects of channel change conditions and distortions in the signal transmission channel. The equalization process calculates a transfer function of the transmission channel and applies the inverse to the transfer function in the received signal to thereby reduce or eliminate the effects of distortion. Typically, channel equalization employs filters that withdraw from the amplitude of a received signal and the phase distortions that result from a variable time response, dependent on the frequency of the transmission channel, for example, in order to provide a decision capability of improved symbol. The equalization removes the interference between symbols of the base band (ISI) caused by the distortions of the transmission channel including the low pass filtering effect of the transmission channel. ISI causes the value of a symbol determined by the values of the predecessor and successor symbols to be distorted, and essentially represents "ghosts" of symbols since ISI includes advanced and delayed symbols with respect to the reference location of the symbol in a region of determined decision. An adaptive equalizer is essentially an adaptive digital filter. In systems that use an adaptive equalizer, it is necessary to provide a method to adapt the filter response so that channel distortions are compensated appropriately. Some algorithms are available for adapt the filter coefficients and therefore the filter response. A widely used method employs the Least Mean Squares (LMS) algorithm. In this algorithm, by varying the values of the coefficient as a function of a signal representative of error, the output signal of the equalizer is forced approximately to a sequence of reference data. This error signal is formed by subtracting the equalizer output signal from the reference data sequence. As the error signal approaches zero, the equalizer approaches convergence, whereby the equalizer output signal and the reference data sequence are approximately zero. When the equalizer operation is started, the coefficient values (filter tap weights) are usually not adjusted with values that produce an adequate compensation of the channel distortions. In order to force the initial convergence of the equalizer coefficients, a "trainer" signal, known as the reference signal, can be used. This signal is programmed in both the transmitter and the receiver. The error signal is formed in the receiver by subtracting a locally generated copy of the training signal from the output of the adaptive equalizer. The training signal helps to open the "eye" initially obstructed of the received signal, as it is known. After adaptation of the training signal, the "eye" opens considerably and the equalizer is switched to the decision-driven operation mode. In this mode, convergence End of the filter take weights is achieved by using the actual values of the symbols of the equalizer output instead of using the training signal. The decision-driven equalizer mode has the ability to track and cancel channel distortions with variable time faster than methods that use training signals transmitted periodically. In order for decision-driven equalization to provide reliable convergence and stable coefficient values, approximately 90% of the decisions must be correct. The training signal helps the equalizer reach this 90% level of correct decision. However, in some systems, a training signal is not available. In such a case, a "blind" equalization is often used to provide an initial convergence of the values of the equalizer coefficient and to force the eye to open. In blind mode, the filter coefficients are roughly adjusted in response to the error signal that is calculated using a known function or algorithm. Among the most popular blind equalization algorithms is the Constant Module Algorithm (CMA) and the Reduced Constellation Algorithm (RCA). These algorithms are described, for example, in Proakis, Digital Communications, McGraw-Hill: New York, 1989 and Godard, Self-recovery Equalization and Carrier Tracking in two Data Communication Dimension Systems, "IEEE Transactions in Communications" , November 1980. Briefly, the CMA is based on the fact that, at the decision instants, the modules of the detected data symbols must lie in a locus of points that define one of several (constellation) circles of different diameters. The RCA is based on forming "super constellations" within a main constellation transmitted. The data signal is first forced to fit within a super constellation, then the super constellation is subdivided to include the complete constellation. In a conventional system that uses a direct feed filter (FFF) and a decision feedback filter (DFF) as an equalizer, the FFF typically performs the adaptive blind (non-decision driven) equalizer during the initial signal acquisition interval. The DFF does not provide the equalization at this time. At the end of the blind equalization interval, the DFF is activated for decision-driven equalization. In the decision-driven mode, the filter coefficients are updated to finer values by using the decision error signal, which is calculated by using the known decision function. At this time, both the FFF and the DFF have their coefficients adapted (updated) in response to the control signals generated locally in the decision-driven mode, for example, based on the differences between the symbol samples that appear at the entrance and exit of a sectioning network. These measures have certain disadvantages. If there are ghost effects and important ISI effects, it will be difficult for the FFF to achieve the equalization because the central intake of the filter will be contaminated by the symbol "ghosts". To equalize the predecessor and successor ghosts, the FFF uses both the predecessor and the successor takes. The successor shots of the FFF overlap with the predecessor shots of the DFF, which is not an effective use of the filter intakes. This limitation is avoided by a system of the type described in U.S. Patent No. 5,712,873 to Shlue et. to the. In that system, a digital signal processor includes a decision feedback filter (DFF) that has different modes of operation before and during decision-driven equalization. Specifically, the DFF operates as a linear feedback filter during the blind equalization, and as a non-linear filter in the decision-driven mode after the blind equalization.

BRIEF DESCRIPTION OF THE INVENTION In accordance with the principles of the present invention, an equalizer for digital channel for processing a demodulated VSB signal containing high definition video information comprises a direct feed filter (FFF) and a decision feedback filter. (DFF) Both the FFF and the DFF operate in an adaptive manner, both in the blind mode and in the decision-directed mode.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a block diagram of a portion of an advanced television receiver, such as a high definition television receiver (HDTV), which includes an adaptive equalizer system, in accordance with the principles of present invention. Figure 2 illustrates a data frame format for a VSB signal, in accordance with the HDTV Grand Alliance system.

DETAILED DESCRIPTION OF THE INVENTION In Figure 1, an analog modulated HDTV signal received by an antenna is processed by an input network 14, which includes RF tuning circuits, a double conversion tuner to produce a band-out signal. intermediate frequency step and gain control circuits, for example. The received signal is a modulated VSB signal as proposed for use by the HCTV Grand Alliance system in the United States. Such signal VSB is represented by a one-dimensional data symbol constellation where only one axis contains quantized data to be retrieved by the receiver. For simplicity, Figure 1, no signals are shown for synchronizing the illustrated functional blocks or a time recovery network (as shown) to derive time and clock signals from the received signal. As described in the Grand HDTV System Specification Alliance dated April 14, 1994, the transmission system VSB transmits the data with a prescribed data box format, as shown in Figure 2. A small pilot signal on the suppressed carrier frequency is added to the transmitted signal to help the carrier to get hooked on the VSB receiver. With reference to Figure 2, each data frame comprises two fields, each field includes 313 segments of 832 multi-level symbols. The first segment of each field is called a field segment and the remaining 312 segments are data segments. The data segments contain data packets compatible with MPEG. Each data segment comprises a four-symbol segment synchronization character followed by 828 data symbols. Each field segment comprises a four-symbol segment synchronization character followed by the field synchronization component, which comprises a predetermined sequence of pseudo-random numbers (PN) of 511 symbols and three predetermined PN sequences of 63 symbols, whose intermediate sequence it is invested in successive fields. A mode control signal VSB (which defines the size of the VSB symbol constellation), follows the last PN 63 sequence, which in turn is followed by 96 reserved symbols and 12 symbols copied from the previous field. The bandpass output signal from the unit 14 is converted into a baseband signal by the demodulator VSB and the carrier recovery network 18. In this example, the network 18 contains arranged circuits as described in the Grand Alliance HDTV System Specification and in an article by W. Bretl et. al., "Design of a VSB modem sub-system for Grand Alliance Digital Television Receivers", IEEE Transactions for Consumable Electronics, August 1995. Briefly, carrier recovery can be carried out through a frequency and phase lock cycle using the small pilot signal component included in the broadcast HDTV signal VSB. The output band base signal from network 18 contains only l-channel data symbols recovered along the real axis. The symbol information demodulated from the network 18 is converted into a digital data stream by means of an analog to digital converter. The recovery of the data segment synchronization and the clock recovery (time) are carried out by means of the unit 15., which may include networks as described in the Grand Alliance HDTV System Specification and in the Bretl et. to the. previously mentioned. A Detection Signal Segment Synchronization is provided when data segment synchronization and time recovery have been achieved. The output of the analog-to-digital converter 19 is also applied with a field synchronization component detector (frame) of data. Adequate networks to provide detection of the field synchronization component (frame) of data are also described in the HDTV Grand Alliance Specification and in the article by Bretl et. to the. The detector 17 provides a Field Synchronization Detection output signal for the microprocessor 66 when the data field synchronization component has been detected. The digital data from the unit 19 is processed by an adaptive equalizer network 50, as will be described below. A band-based output signal equalized from the network 50 is decoded by the unit 60 and processed by the output network 64. The decoder 60 includes, for example, cross-linked decoding, data deinterleaver, Reed Solomon error correction and audio / video decoder networks as known and described, for example, in the aforementioned Bretl article. The output processor 64 includes audio / video processors and audio / video playback devices. The segment synchronization and field synchronization detection circuits in units 15 and 17 provide Field Synchronization Detection and Segment Synchronization Detection input signals for the control signal generator 66 (e.g., including a microprocessor). ), when these synchronization components of the received signal are detected. The microprocessor 66 responds to these signals by providing output control signal and a PN signal (sequence of pseudorandom numbers) of output reference to the equalizer 50 as will be described. This PN training signal sequence is a repetitive pattern fixed, binary data as specified in the Grand Alliance HDTV Specification, and is a pre-programmed reference signal acquired by the control signal generator 60 from the memory 70. Since the data pattern of the PN signal As it is known, an exact error is generated by obtaining the difference between the stored reference PN signal and the training signal component PN of the data stream received during the field synchronization interval. The control signals control the switching of the multiplexers 26, 28 and 29 in the blind, training and decision-driven modes as will be described. The output signal from the unit 19 contains digital data as well as symbol interference (ISI) caused by the distortions and interferences in the transmission channel. The signal is applied to a real direct filter (FFF) 20 (the opposite of a complex) that operates as an equalizer, for example, a separate symbol speed equalizer (separate-T), which in this case, It is installed as a digital FIR filter. The coefficient values (weights of the tap) of the equalizer filter 20 are controlled in an adaptive manner by a coefficient control signal from the multiplexer 26 as will be described later.

The signal equalized from the filter 20 is combined by an adder 24 with an equalized signal from a decision feedback filter 30, which operates as an equalizer. The DFF 30 removes the interference between symbols not removed by the FFF . The coefficient values (tap weights) of the equalizer filter 30 are also controlled in an adaptive manner by the coefficient control signal (ie, the switched error signal) from the multiplexer 26. The input signal to be equalized by the DFF 30 is provided from the multiplexer 28. Both the FFF 20 and the DFF 30 have coefficient values adapted (updated) in response to the coefficient control signal during the blind and decision-driven modes of operation. Both the FFF 20 and the DFF 30 are digital FIR filters that perform the equalization functions individually. When they are considered together, these filters represent an equalizer 50 added to equalize the input signal to the decoder 60. The FFF 20 equalizes the pre-ghost components, while the DFF 30 equalizes the post-ghost components. The FFF 20 and the DFF 30 operate in a linear infinite impulse (MR) response mode from the moment the input signal is initially received. Both the FFF 20 and the DFF 30 are FIR devices, but the feedback operation causes the DFF 30 to operate as an MR device. The output signal from the adder 24 is the output signal of the equalization 50. The signal of the adder 24 is coupled with a network including multiplexers 26 and 28, a disconnector 40, a subtraction combiner 21 and a source 25, which provides a blind adaptation CMA algorithm. The multiplexer 26 provides one of two signals to the inputs of the coefficient control of the FFF 20 and the DFF 30 in response to a control signal produced by the microprocessor 66 for different operating modes, when the segment synchronization and field synchronization components are detected, as will be explained later . These signals from the multiplexer 26 include the blind adaptation algorithm CMA from the unit 25, which responds to the output signal of the equalizer and an error signal from the output of the subtraction combiner 21. The error signal represents the difference between the input signal of the switch 40 and the output of a third multiplexer 29. The output of the combiner 21 is either an error signal or a training error signal where: Error of the switch = output of the disconnector 40 - equalizer output Training error = reference PN signal - equalizer output When the training error signal is generated, the equalizer output is the PN component of the received data stream. The multiplexer 28 provides one of three input signals to a signal input of the DFF 30, in response to the control signal from the microprocessor 66. These signals include the output signal of the equalizer 50 as applied by a direct connection to a first input (1) of the multiplexer 28, the output signal from the disconnector 40 applied to a second input (2) of the multiplexer 28 and the reference PN signal stored from the memory 70 and the unit 66 applied to a third input (3) of the multiplexer 28. The multiplexer 29 responds to a switching control signal from the microprocessor 66 and receives as inputs the PN signal sequence of reference training during the field synchronization intervals, and the output signal from the switch 40 at other times. The output of the multiplexer 29 is applied to the sub-combiner 21 where it is subtracted with the output signal from the equalizer 50, to produce the error signal. The error signal represents the difference between the disconnector 40 and the output signals of the equalizer 50 or the difference between the reference PN signal and the PN signal component of the received data stream, as contained in the output signal of the Equalizer 50. During operation, the equalizer 50 exhibits an initial condition, a blind operation mode, a data-driven training mode, a decision-driven mode and an equalized steady-state condition. Blind mode occurs when the characteristic of the eight level "eye" pattern of the received 8-VSB signal exhibits a closed eye pattern. The training operation and directed by decision occurs later, when the "eye" exhibits an open eye pattern. It should be noted that it is not necessary that the "eye" of the pattern be opened in case the training signal component is detected immediately. In that case, The training signal component is used as soon as it is detected, even before the "eye" of the pattern is opened. In the initial condition, before the time lock is reached (time synchronization), the FFF 20 and the DFF 30 are reactivated while the demodulator 18 attempts to block the received signal with respect to the automatic gain control (AGC), the time and the carrier. At this time, the control signals applied to the multiplexers 26 and 28 cause the coefficient values of all the taps of FFF 20 and of the DFF 30 to be reset and maintained at a zero value, except for a tap value, the which is reset to a predetermined initial value that is not zero. This action of the control signals freezes the values of the filter coefficient to avoid unwanted random changes in coefficient values before the equalization process is initiated. Alternatively, the FFF 20 and the DFF 30 can be pre-loaded with the last known valid coefficient values. In this initial state, both multiplexer 26 and multiplexer 28 exhibit a zero output. The multiplexer output 29 is a "unimportant" condition at this time. The process of blind equalization that uses an algorithm CMA starts after the rough time is reached. This occurs when the segment synchronization component of the received signal is detected. Locking the carrier and AGC blocking are present. At that moment the Segment Synchronization Detection signal is directed to the microprocessor 66, which in turn generates adequate control signals. The blind equalization process involves the use of a CMA algorithm before the field synchronization component of the received signal is detected. Specifically, the control signal applied to the multiplexer 26 causes the multiplexer 26 to drive the CMA algorithm from its input (1) to the control inputs of the coefficient of the FFF 20 and the DFF 30, and a control signal applied to the multiplexer 28 causes that the multiplexer 28 conducts the output signal of the equalizer from its input (1) to the signal input of the DFF 30. The output of the multiplexer 29 is a "no matter" condition during the blind equalization interval. The processes of decision-directed equalization and training occur later, when the blocking of the time after detection is achieved by the field synchronization component. The training mode occurs when the received PN signal component is available during the field synchronization interval of each data frame. The decision-driven mode occurs at other times during each data frame. The presence of the field synchronization component initiates the start of the training mode of the PN sequence. At that time the Field Synchronization Detection signal is conducted to the microprocessor 66, which in turn generates suitable control signals. During the field synchronization intervals when the training component PN is available and the reference signal PN is acquired from the memory 70, the control signals respectively applied to the multiplexers 26, 28 and 29 cause (a) a training error signal to be coupled with the control inputs of the coefficient of the FFF 20 and the DFF 30 through the multiplexer 26, (b) the reference signal PN to be conducted to the signal input of the DFF 30 through the multiplexer 28, and (c) the reference signal PN to be coupled to the combiner 21 through the multiplexer 29. During the intervals of non-field synchronization of each data frame when the decision-driven equalization is carried out based on the disconnector, the control signals respectively applied to the multiplexers 26, 28 and 29 cause (a) an error signal of the disconnector to be coupled with the control inputs of the coefficient of the FFF 20 and of the DFF 30 through the multiplexer 26, (b) the output of the disconnector 40 to be conducted to the signal input of the DFF 30 through the multiplex r 28, and (c) the output of the switch 40 to be coupled with the combiner 21 through the multiplexer 29. During the steady state operation, after the equalization has been reached, the signal conditions described above prevail for the operation directed by decision. The operation of the equalizer 50 described above is summarized in the following table. Mode of operation Input signal control of Multiplexer coefficient output 28 to DFF multiplexer 29 multiplexer 26 a The exposed system including the adaptive equalizer 50 advantageously exhibits reduced physical equipment costs and complexity. The equalizer 50 comprises 20 and 30 real filters, rather than complex ones and does not require the use of rotator and detuner circuits (for example, in the equalizer control lock). A rotator and detuner is basically a circuit for circularly shifting the data symbol constellation to compensate for unwanted frequency and phase deviations in a received signal. The adaptive equalizer 50 operates as a linear IIR filter, thereby improving the equalizer's capabilities since the DFF 30 provides some equalization even in the blind mode when the DFF 30 operates as a linear feedback filter, before operating as a filter non-linear in the decision-driven mode that follows the blind equalization. The initial operation of the DFF 30 as a linear feedback filter produces some convergence that facilitates the equalization of the system, particularly in the presence of Signal important ghosts. In particular, at this time the DFF 30 exhibits the ability of a feedback filter to cancel distant phantom components. In addition, the exposed system exhibits a smoother transition from a linear mode of operation to a non-linear decision mode after blind equalization, compared to a conventional system using the equalization FFF and DFF. This is because the DFF 30 starts the operation in the non-linear mode after having been pre-conditioned to operate in the linear mode, that is, many of its coefficients have been adapted in the direction of the final values. Essentially, Equalizer 50 is physical equipment and a recurrent linear efficient filter that takes advantage of all available data to achieve equalization for a high definition modulated VSB signal as soon as possible. Both the FFF 20 and the DFF 30 operate at a symbol rate, and a PN sequence is used to facilitate rapid equalization. The data processing occurs continuously online in real time and with certain advantages uses a direct connection from the output of the equalizer 50 for the DFF 30 through the multiplexer 28 to allow the DFF 30 to facilitate the coarse acquisition of the signal during the blind operation mode. In this way, the decision feedback filter 30 is advantageously used in a linear mode during blind equalization, then responds to the PN signal in the training mode and to the output of the disconnector during the directed mode by decision, as described.

Claims (15)

1. CLAIMS 1. A system for processing a modulated signal (VSB) Vestigial Sideband containing high definition video information represented by a VSB constellation of symbols and subject to exhibiting unwanted interference, which comprises: a demodulator that responds to the VSB signal modulated received to produce a demodulated baseband signal; and an adaptive equalizer that has an input to receive the demodulated baseband signal and an output in which the equalized signal is produced, the adaptive equalizer includes: (a) an adaptive direct feed filter (FFF) to equalize the signal demodulated; the FFF exhibits: (1) blind operation directed by no decision, linear in the first mode of operation; and (2) decision directed operation in a second subsequent mode of operation; and (b) an adaptive decision feedback filter (DFF) to equalize the demodulated signal, the DFF exhibits: (1) blind operation directed by non-decision, linear in a first mode of operation; and (2) operation directed by non-linear decision in the second mode of operation.
2. 2. A system according to claim 1, wherein: the demodulator includes a carrier recovery network; Y the FFF and the DFF do not perform the operation of rotation or carrier desrotation; and the DFF is directly connected to the output of the equalizer without an intervening carrier rotation operation.
3. 3. A system according to claim 1, wherein: the signal input of the DFF is directly connected (DIRECT CONNECTION) to the signal equalized at the output of the equalizer during the blind operation in the first mode.
4. 4. A system according to claim 3, wherein: the signal input of the DFF receives an output signal from the disconnector during the decision-directed operation in the second mode.
5. 5. A system according to claim 1, wherein: the direct feed filter and the decision feedback filter operate at a symbol rate, in real-time online.
6. 6. A system according to claim 1, wherein the signal having a data frame format (Figure 2) is comprised of a succession of data frames, which comprise a field synchronization component before a plurality of data segments with an associated segment synchronization component.
7. 7. A system according to claim 6, wherein: in blind mode the DFF signal input is directly connected to the equalized signal at the equalizer output, and the control inputs of the FFF and DFF coefficient respond to a blind adaptation algorithm (CMA 25); and in the decision mode the signal input of the DFF responds to an output signal from the disconnector (ERROR) and the control inputs of the coefficient of the FFF and the DFF respond to an error signal representing the difference between the output signal of the disconnector and the output signal of the equalizer.
8. 8. A system according to claim 7, wherein: the DFF responds to a training signal (PN) during the field synchronization intervals and responds to the output signal of the disconnector during the non-field synchronization intervals; during the non-field synchronization intervals of the data frame, the control inputs of the coefficient of the FFF and the DFF respond to an error signal of the disconnector representing the difference between the output signal of the disconnector and the output signal of the equalizer; and during the field synchronization intervals of the data frame, the coefficient control inputs of the FFF and the DFF respond to a training error signal, which represents the difference between the training signal component of the equalizer output signal and a reference training signal.
9. 9. A system according to claim 8, where: the training signal is a PN sequence.
10. 10. A system according to claim 6, wherein: the direct feed filter and the decision feedback filter operate at a symbol rate, in real-time online.
11. 11. In a system for processing a Vestigial Sideband modulated signal (VSB) containing high definition video information represented by a VSB constellation of symbols and subject to exhibiting unwanted distortions, the system includes a channel equalizer comprising a filter of direct feed (FFF) and a decision feedback filter (DFF) to produce an equalized output signal, a method for processing the signal which comprises the steps of: demodulating the modulated VSB signal to produce a demodulated baseband signal; driving the demodulated baseband signal towards the equalizer; adapt the FFF linearly during the first blind mode of operation; «« »» 25 adapt linearly the DFF during the blind mode of operation; adapt non-linearly the FFF in a subsequent mode of operation directed by decision; and adapt non-linearly the DFF in a subsequent decision-directed mode.
12. 12. A method according to claim 11, wherein the step of linear adaptation of the DFF during the blind operation mode includes the step of: directly connecting the output signal equalized with a signal input of the DFF without the intervention of Carrier rotation during blind operation mode to allow coarse equalization by DFF during blind mode. The method according to claim 11, wherein the signal having a data frame format is comprised of a succession of data frames comprising a field synchronization component, preceding a plurality of data segments with an associated segment synchronization component. The method according to claim 13, which also comprises the steps of: generating an output signal of the switch in response to an output signal of the equalizer; coupling the output signal of the disconnector with a signal input of the DFF in the decision-driven mode; Y coupling the error signal representing the difference between the output signal of the disconnector and the output signal of the equalizer with the coefficient control inputs of the FFF and DFF during the decision-driven mode. 15. In a system to process a modulated signal (VSB) Vestigial Sideband received, which contains high definition video data represented by a VSB constellation of symbols and that is subject to exhibit unwanted interference, the data has a data box format (Figure 2) consisting of a succession of frames of data having a field synchronization component prior to a plurality of the data segments with an associated segment synchronization component, the system includes a direct-feed filter (FFF) and a decision-feed filter (DFF), the which constitute a channel equalizer for producing an equalized output signal, the method for signal processing comprises the steps of: demodulating the demodulated data signal VSB to produce a demodulated baseband signal; applying the demodulated baseband signal to the channel equalizer constituted by the FFF and the DFF; (a) during a non-directed, blind decision-based equalization interval; (1) apply the equalizer output signal to a signal input of the DFF; (2) adapt the FFF in a linear fashion; f «. ' * t 27 (3) adapt the DFF in linear form; (b) during the subsequent interval of training equalization during intervals of field synchronization components; (1) apply a training signal to the signal input of the DFF; (2) adapting the FFF and the DFF in response to a training error signal which represents the difference between a reference training signal and a training signal component of the equalizer output signal; (c) during the decision-directed equalization interval following the training interval and comprising intervals of synchronization component not belonging to a field; (2) adapt the FFF and the DFF in response to an error signal from the disconnector which represents the difference between an output signal of the symbol switch and the output signal of the equalizer.