MXPA98004662A - Decoder for a trellis encoded signal corrupted by ntsc co-channel interference and white noise - Google Patents

Abstract

A receiver for decoding encoded data transmitted by a transmitting station including (i) sum (50) and difference (46) filters for filtering the encoded data (R(n)) and providing corresponding sum (V(n)) and difference (U(n)) filtered outputs, (ii) first and second variable amplifiers (54 and 56) for respectively controlling the sum and difference filtered outputs so as to provide corresponding sum and difference controlled filtered outputs that vary along a continuum in order to suppress co-channel interference and white Gaussian noise in the encoded data, and (iii) a Viterbi decoder (58) for decoding the sum and difference controlled filtered output to recover uncoded data.

Description

DECODER FOR A TRELLIS CODED SIGNAL THAT IS CORRUPTED BY NTSC CO-CHANNEL INTERFERENCE AND WHITE NOISE DESCRIPTION OF THE I vENCTÓW The present invention relates to decoding a trellis encoded signal that has been corrupted by color interference. A current implementation of an 8 VSB receiver for HDTV, is based on a switchable comb filter in order to reject co-channel interference, such as co-channel NTSC interference, which may be present for example in the marginal reception areas of an ATV service area. In this way, when co-channel NTSC interference is detected, the comb filter is activated (switched to on) in order to filter out the co-channel NTSC interference. When the comb filter is active, the decoding of the received data is more complex than when the comb filter is inactive. Accordingly, when no co-channel NTSC interference is detected, the comb filter is deactivated (i.e. switches off) in order to provide optimal performance against white Gaussian interference. The comb filter has been shown to be an effective filter in cost for interference rejection NTSC co-channel. The use of this switchable comb filter however has several disadvantages. First, while the REF: 27018 comb filter is good for rejecting NTSC interference from co-channel, the presence of filter comb degrades performance when white Gaussian interference is present. Accordingly, when only co-channel NTSC interference is present, the comb filter is an effective filter. When only white Gaussian interference is present, however, the comb filter becomes inactive. When both the NTSC co-channel interference and the white Gaussian interference are present, the comb filter is active such that a substantial amount of white Gaussian interference is allowed to pass. In this way, the comb filter may not be an appropriate filter under conditions of both co-channel NTSC interference and white Gaussian interference. A further disadvantage of the comb filter is that it is a switchable filter such that it is already active or inactive. The developed control circuits that determine when to switch the comb filter on or off are complicated and can make incorrect decisions. The present invention is directed to an adaptive decoder that adjusts appropriately to the relative amounts of co-channel NTSC interference and white Gaussian interference. In this way, the present invention achieves good performance under both NTSC co-channel interference and white Gaussian interference conditions. In accordance with one aspect of the present invention, a receiver for decoding interference encoded data received from a transmitting station comprises a filter means and a decoding means. The filter means variably filters the co-channel and noise interference in the encoded data depending on relative amounts of co-channel interference and noise in the encoded data. The decoding means decodes filtered encoded data. According to another aspect of the present invention, a receiver for decoding noise encoded data sent by a transmitting station comprises first and second filters and a decoding means. The first filter has a power to receive the coded data, the first filter is a difference filter, and the first filter has a first output to provide first coded data corresponding to a difference between the coded data and delayed coded data. The second filter has a power to receive the encoded data, the second filter is a sum filter and the second filter has a second output to provide second filtered encoded data, corresponding to a sum of the encoded data and the delayed encoded data. The decoding means decrypts the first and second filtered encoded data. According to yet another aspect of the present invention, a receiver for decoding encoded data sent by a transmission station comprises first and second filter means, a controller means, and decoding means. The first and second filter means filter the encoded data and provide corresponding first and second filtered outputs. The controlling means relatively control the first and second filtered outputs to provide relatively corresponding first and second controlled filtered outputs in order to suppress co-channel interference and noise in the encoded data. The decoding means decipher the first and second relatively controlled filtered outputs. These and other features and advantages of the present invention will become more apparent from the detailed consideration of the invention, when taken in conjunction with the drawings in which: Figure 1 illustrates an example of a trellis pre-coder and coder of the prior art coding data to be decoded by the present invention; Figure 2 illustrates an example of a Viterbi comb and decoder filter of the prior art for filtering and decoding data typically encoded by the trellis pre-decoder and encoder of Figure 1; Figure 3 illustrates a filter and decoder according to the present invention for filtering and decoding data typically encoded by the trellis pre-coder and coder of Figure 1; Figure 4 is a constellation composed of successively coded data points and is useful to explain the present invention; Figure 5 is a composite constellation of Figure 4 that has been rotated in accordance with the principles of the present invention; Figure 6 is a rotated composite constellation of Figure 5 showing the effects of reduced gain on the V (n) axis according to the principles of the present invention; Figure 7 is a rotated composite constellation of Figure 5, showing the zero gain effect on the V (n) axis according to the principles of the present invention; Figure 8 illustrates a trellis coder diagram that is descriptive of the operation of the trellis coder of Figure 1; Figure 9 illustrates the subset constellations implemented by the Viterbi decoder of Figure 3, - Figure 10 illustrates a trellis decoder diagram that is descriptive of the operation of the Viterbi decoder of Figure 3; Figure 11 illustrates a diagram that is descriptive of the decoding of the bit not encoded by the Viterbi decoder of Figure 3; Figure 12 illustrates the impact of the subset constellations of Figure 9, when the gain on the V (n) axis is set to zero; Figure 13 illustrates a simplification of the subset constellations of Figure 12; and Figure 14 illustrates an apparatus that can be implemented in order to determine the gains for the variable gain amplifiers of Figure 3. Figure 1 illustrates an encoder 10 which can for example be an 8 VSB coder. The encoder 10 includes a precoder 12, a trellis coder 14, and a symbol scraper 16. This precoder 12 includes an adder 18 and a delay element bit D 20. The value "D" is chosen to correspond to the delay of the comb filter of rejection of interference, which is present in the receiver. The adder 18 is a MOD-2 adder. The adder 18 receives a first feed data bit X2 (n) of a data bit pair X2Xl (n) and adds this first feed bit X2 (n) to the adder output 18 that occurs prior to D bits. The output of the adder 18 forms the intermediate bit Y2 (n) which is provided as the output bit Z2 (n) to the symbol mapper 16. The function (n) denotes that an associated value is at a discrete time n. A second feed data bit XI (n) of the data bit pair X2Xl (n) is provided to the trellis encoder 14 which includes an adder 22 and two bit delay elements D 24 and 26. The adder 22 is an adder MOD-2. The trellis encoder 14 provides the second feed data bit Xl (n) as the output bit Zl (n) directly to the symbol mapper 16. The second feed data bit XI (n) is also provided to a first feed of the adder 22 of the trellis encoder 14. The output of the adder 22 is connected to the bit delay element D 24. The bit delay element D 24 delays the adder output 22 and supplies this delayed output as the output bit ZO (n ) to the symbol mapper 16. The delayed output of the bit delay element D 24 is also connected back to the bit delay element D 26. The output of the bit delay element D 26 is connected to a second supply of the adder 22. Accordingly, the power data bits XI (n) and X2 (n) of the bit pair X2Xl (n) enter the encoder 10. The encoder 10 (i) differentially encodes the feed data bit X2 (n) ) as the output bit Z2 (n), (ii) passes the feed data bit XI (n) through as the output bit Zl (n) without modification, and (iii) convolutionally encodes the feed data bit XI (n) as the output bit ZO (n). The symbol mapper 16 maps each set of output bits Z2Z1Z0 (n) as a corresponding symbol S (n). Each symbol S (n) can therefore have one of eight possible signal levels. The symbols S (n) are then transmitted in a conventional manner through a transmission channel where they can pick up co-channel NTSC interference and white Gaussian noise before they are received by a receiver as received symbols R (n). The bit delay elements D 20, 24 and 26 illustrated in Figure 1, for example, represent twelve-bit time delays (ie D is twelve). As described above, the information bits X2X1 (n) are coded differentially and convolutionally as symbols. These symbols are separated by multiples of D (where D can for example be twelve). Accordingly, the coded bit stream, in effect, consists of twelve independently coded data streams. A decoder must decode these coded data streams independently of each other. Figure 1 also includes a supplemental state variable Q2 (n) that represents the previous value of Yl (n) such that Q2 (n) = Yl (n-D). The value Q2 (n) does not currently exist in the encoder 10 and does not impact the operation of the encoder 10. The value Q2 (n) is simply a convenient annotation for the subsequent description of the present invention. A receiver including a Viterbi decoder 28, a switch 30 and a comb filter 32, receives the transmitted symbols R (n). The Viterbi decoder 28 decodes the received symbols R (n) to recover the original pair of bits X2X1 (N). The switch 30 detects the presence of co-channel NTSC interference. If the co-channel NTSC interference is present, the switch 30 goes to its upper position in order to connect the comb filter 32 to the Viterbi decoder 28 in such a way that the comb filter 32 filters the received symbols R (n). The comb filter 32 includes an adder 34 and a symbol delay element D 36. The adder 34 has a first feed that receives the symbol R (n) directly and a second feed that receives the symbols R (n) delayed by D times of symbol. The value of D determines the position and quantity of the notches that the comb filter places on the pass band. A selection of D = 12 produces a set of notches that are close to the chromatic and visual aural NTSC carriers of interference. If the symbol delay element D 36 is a twelve-symbol delay element, the delay elements D of bits 20, 24 and 26 of FIG. 1 are twelve-bit delay elements. If the co-channel NTSC interference is not detected by the receiver, the switch 30 goes to its lower position, where the received symbols R (n) are connected directly to the Viterbi decoder 28. Although the filter comb 32 illustrated in the Figure 2 is effective for rejecting co-channel NTSC interference, its performance is poor when both co-channel NTSC interference and white Gaussian noise are present. In addition, the circuits that determine when to operate the switch 30 to activate the comb filter 32 are complicated and can occasionally make a correct decision. A decoder 40 of FIG. 3 is more effective in rejecting both co-channel NTSC interference and white Gaussian noise, does not require a switching decision and does not appreciably contribute to receiver complexity. The decoder 40 includes a first filter 42 and a second filter 44. The first filter 42 which is a difference filter, includes an adder 46 having a first feed receiving the received symbols R (n) directly, and a second feed for receiving the received symbols R (n) through a symbol delay element D 48. The symbol delay element D 48 may impose a delay of twelve symbols on the received symbols R (n). The second filter 44, which is a sum filter, includes an adder 50 having a first feed that receives the received symbols R (n) directly, and a second feed to receive the received symbols R (n) through a symbol delay element D 52. The symbol delay element D 52 can also impose a delay of twelve symbols on the received symbols R (n). The difference output of the first filter 42 is designated U (n), and the summation output of the second filter 44 is designated V (n). The difference output U (n) represents the difference between a symbol and a corresponding symbol that due to the symbol delay element D 48, occurred D previous symbol such that U (n) = R (n) - R (n - D ). Similarly, the sum output V (n) represents the sum of a symbol and a corresponding symbol which, due to the symbol delay element D 52, occurred from previous symbols, such that V (n) = R (n) + R ( n - D). The difference output U (n) is processed through a first variable gain amplifier 54 and the summation output V (n) is processed through a second variable gain amplifier 56. The first variable gain amplifier 54 has a gain g0 and the second variable gain amplifier 56 has a gain g-_. In this way, the output of the first variable gain amplifier 54 is g0U (n), and the output of the second variable gain amplifier 56 is g? V (n). The outputs of the first variable gain amplifier 54 and the second variable gain amplifier 56 are connected to a bank of Viterbi decoders D 58. Each of the Viterbi decoders of the decoder bank Viterbi D 52, decodes one of the coded currents interspersed with time using the outputs g0U (n) and g-_V (n) in order to recover the corresponding pair of power bits X2X1 (n). The symbols R (n) entering the decoder 40 may be corrupted by additive noise. The decoder 40 filters the symbols R (n) with the first and second filters 42 and 44 to give the difference and sum outputs U (n) and V (n). The first and second variable gain amplifiers 54 and 56 apply the corresponding gains g0 and gx to the difference and sum outputs U (n) and V (n) in order to produce the outputs g0U (n) and g? V (n) . The bank of decoders Viterbi D 58 extracts the information bits Xl / X2 (n) from the outputs g0U (n) and g? V (n). The first filter 42 and the second filter 44 can already be before or after the usual equalizer and phase follower (not shown) which are conventional in a receiver 8 VSB. As previously discussed, each symbol that results from mapping the output bits Z2Z1Z0 (n), can take one of eight possible signal levels independently of any other symbols. When two symbols (a newly received symbol R (n) and a symbol R (n - D) that was received D symbol times before) are processed together by a comb filter, a two-dimensional constellation of sixty-four points results, which is illustrated by way of example in Figure 4. The newly received symbol R (n) is illustrated on the horizontal axis of Figure 4, and the previously received symbol R (n-D) is illustrated on the vertical axis. Because each of these two symbols can take eight possible levels, the combination of these two symbols processed together results in sixty-four possible points as illustrated in Figure 4. When the received symbols R (n) are corrupted only by white Gaussian noise, white Gaussian noise is independent of sample-to-sample. Therefore, the distribution around each of the constellation points shown in Figure 4, due to white Gaussian interference is circularly symmetric. (That is, if the symbols are in pairs in the manner described above and seen on an oscilloscope, the points would be blurred circles). When the received symbols R (n) are corrupted only by co-channel NTSC interference, the NTSC co-channel interference does not depend on sample-to-sample and in fact, there is high correlation between symbols spaced by times of twelve symbols. Therefore, the distribution around each of the constellation points illustrated in Figure 4, because the NTSC co-channel interference is elliptical and has a principal axis aligned on the line defined by the equation R (n - D) = R (n). When the received symbols R (n) are corrupted by both co-channel NTSC interference and white Gaussian noise, the distribution around each of the constellation points illustrated in Figure 4 varies between elliptical and circular, according to these relative amounts of co-channel NTSC interference and white Gaussian noise. Due to the first filter 42 and the second filter 44 of Figure 3, the two-dimensional constellation of sixty-four points of Figure 4 is rotated counterclockwise by 45 ° to a new coordinate system as shown in FIG. illustrated in Figure 5. The difference output U (n) is now on the horizontal axis and the summation output V (n) is on the vertical axis. The decoding algorithm described here operates directly using this rotated two-dimensional constellation. It should be noted that the techniques described herein are different from trellis-dimensional coding because the transmitted signal is encoded as a one-dimensional constellation. The two-dimensional constellation arises in this technique by forming (superimposing) pairs of symbols transmitted on the receiver. The counterclockwise rotation of the sixty-fourth point bi-dimensional constellation of Figure 4 to the new coordinate system illustrated in Figure 5 is fortuitous because the NTSC co-channel interference It has also been turned by the same amount, so that its main axis is on the vertical axis. That is, said rotation is useful because if there is co-channel NTSC interference present, the decoder 40 should put less weight on errors in the V direction because these errors are known as noisy. Additionally, if only white Gaussian noise is present, the noise distribution is circulated and consequently equal weight will be given by the decoder 40 to errors in both dimensions U (n) and V (n). In this way, the function of the gains g0 and g-_ is to appropriately weight the errors in the outputs of the first and second filters 42 and 44 according to the changing conditions of co-channel NTSC interference and white Gaussian noise, in order of putting the correct weight on the errors in each of the dimensions U (n) and V (n). Both of the gain values g0 and g1 are adjusted in order to react to the co-channel NTSC interference and white Gaussian noise statistics. Alternatively, the gain value g0 can be set to unity and the gain value g? it can be adjusted between 0.0 and 1.0 depending on the relative amounts of co-channel NTSC interference and white Gaussian noise. This simplification is appropriate in those cases where only co-channel NTSC interference and white Gaussian noise are assumed. If the interference is a more general continuous wave interference having frequency or unknown frequencies, then adjustment of both gains is appropriate. Figure 6 illustrates the impact of a change in the gain value g-_ in the two-dimensional constellation of sixty-four points illustrated in Figure 5. As the gain value gx is reduced to less than 1.0, the two-dimensional constellation of sixty-four points is compressed in dimension V. At the optimum value of gx, the interference around each point of the two-dimensional constellation of sixty-four points is equal in both dimensions, and the branch metric of the Viterbi decoder bank D 58 is based on the UV point distances (n) to each point in the compressed constellation. As the gain value gx is reduced to zero, all the points in the two-dimensional constellation of sixty-four points are projected on the U (n) axis as illustrated as Figure 7. Accordingly, the two-dimensional constellation of sixty-four points is condensed to a one-dimensional constellation of fifteen points. Under this condition, the bank of decoders Viterbi D 58 makes its decisions based only on the output of the first filter 42. The trellis coder diagram of Figure 8 describes the convolutional coding process performed by the trellis encoder 14 of the Figure 1. The trellis coder diagram of Figure 8 shows the current state of trellis coder 14 in column Q1 0 (n) and the next resulting state of trellis coder 14 in column QiQOtn + D). The transition from a state in the column QlQO (n) to a state in the column Q1Q0 (n + D) depends on the intermediate data bit Yl (n). The label Z1Z0 (n) in Figure 8 shows which of the four subsets (00, 01, 10 and 11) is transmitted as a symbol. For example, if the current state of the trellis encoder 14 is such that Q1Q0 (n) = 00, and if Yl (n) is 0, then Zl (n) is 0 because Yl (n) is 0, Z0 (n ) is 0 because Q0 (n) is 0, and Q1Q0 (n + D) becomes 00. On the other hand, if the current state of the trellis encoder 14 is such that Q1Q0 (n) = 00, and if Yl ( n) is 1, then Zl (n) is 1 because Yl (n) is 1, Z0 (n) is 0 because Q0 (n) is 0, and Q1Q0 (n + D) becomes 01. eight points of the selected constellation by the "combinations of the Z2Z1Z0 (n) bits are divided into four subsets, where each subset has the same bits Z1Z0 (n). For example, the subset Z1Z0 (n) = 00 contains the points Z2Z1Z0 (n) = 000 and 100. The selection between these two points in the subset Z1Z0 (n) = 00 for transmission, is determined by the uncoded bit Z2 (n) and is not shown in Figure 8. Sin However, as it is known, if Z2 (n) is zero, the point Z2Z1Z0 (n) = 000 is chosen and if Z2 (n) is one, the point Z2Z1Z0 (n) = 1 00 is chosen. The bank of decoders Viterbi D 58 (Figure 3) performs its decoding operation using the outputs g0U (n) and gxV (n) of the corresponding first and second filters 42 and 44, based on the sixteen sub-assemblies illustrated in Figure 9 of the two-dimensional constellation of sixty-four points. The sixteen subsets shown in Figure 9 are defined by the sixteen possible combinations of the output bits ZlZO for times of two nyn-D symbols. The four rows of Figure 9 correspond to ZlZO (n) = 00, 01, 10 and 11 and the four columns of Figure 9 correspond to ZlZO (n -D) = 00, 01, 10 and 11. The subsets are labeled in the upper right corner as SO, SI, S2, .... SF according with the hexadecimal representation of the bits Zl (n), Z0 (n), Zl (n - D), and Z0 (n - D). The sixteen subsets shown in Figure 9 form all possible combinations of subsets that may occur between two symbols. There are four possible subsets for the current symbol [ZlZO (n)] and there are four possible subsets for the previous symbol [ZlZO (n - D)]. For example, if the previous subset is ZlZ0 (n - D) = 01 and the current subset is ZlZO (n) = 10, then the UV output (n) of the first and second filters 42 and 44 must be in the subset S9 of the sixteen two-dimensional constellations. The bank of decoders Viterbi D 58 performs decoding according to the subsets shown in Figure 9 and the trellis decoder diagram shown in Figure 10. This trellis decoder diagram is constructed by using the trellis decoder diagram of Figure 8 with the state variable Q2 (n) and the subsets of Figure 9. This trellis decoder diagram of Figure 10 can be used by the Viterbi decoder bank D 58 in determining which of the sixteen subsets results from each state transition ( the state transition is defined by the state Q2Q1Q0 (n - D) of the encoder 10. According to this, the bank of decoders Viterbi D 58 has eight states, with each state corresponding to the status bits Q2Q1Q0 of the encoder 10 shown in Figure 1. State bit Q2 is simply used for accounting to track the value of the bits of previous feed data Yl (n - D). Accordingly, Figure 10 shows the state transitions and associated subsets S0-SF of the data received by the decoder bank Viterbi D 58. Using the trellis decoding diagram of Figure 10, the decoder bank Viterbi D 58 decodes the g0U (n) and g-_V (n) signals in its normal form with the Viterbi 58 decoder bank that generates branch metrics for each S0-SF subset, using these branch metrics to update a cumulative path metric for each decoder state and maintaining a survivor path associated with each state. The surviving path of the lowest path metric is the path of the decoded information bit XI (n) and the associated subset (SO-SF). The generation of the branching metric performed by the decoder bank Viterbi D 58 is done by measuring the distances between each received point (g0U (n), g? V (n)) and the points in each subset of Figure 9, with each of these subsets adjusted in scale by the gains g0 and g1 on the U (n) axis and the V (n) axis, respectively. Alternatively, the branch metric can be calculated by directly measuring the distance between the received point (g0U (n), gxV (n)) before adjusting in scale and each point in the subset of Figure 9 and applying the g0 and g-_ gains to the distance measurements U and V, appropriately. As illustrated in Figure 9, each of the two-dimensional subsets contains four points. Once the bank of decoders Viterbi D 58 has determined in which subset the transmitted point is located, the bank of decoders Viterbi D 58 uses that subset in order to find the information bit differentially encoded X2. Which of the four points in this subset is received, is determined by the bit Z2 (Figure 1) for the pair of consecutive symbols Z2 (n) Z2 (n-D). The decoding of information bit X2 is done according to which of the four points in the subset is closest to the received point g0U (n) g1V (n). The four points of an arbitrary subset and the bit values Z2 (n) Z2 (n - D) that correspond to each point are illustrated in Figure 11. The four possible permutations of the bits Z2 (n) Z2 (n - D ) uniquely identify the point within the subset. In this way, the decoding of the information bit X2 is performed by determining which Euclidean distance between the received point g0U (n) gxV (n) and the four points Z2 (n) Z2 (n-D) of the subset is shorter. The point Z2 (n) Z2 (n - D) is closest to the received point g0O (n) g ^ and di) determines the bit X2 (n) according to the labeling shown in Figure ll. Due to the differential coding of bit X2, both of the upper and lower bits have a label X2 (n) = 0 and the left and right points both have a label X2 (n) = 1. Accordingly, if one of If the gain values g0 or gx are reduced to zero, the differentially coded bit can still be determined. For example if gx = 0, then Z2 (n) Z2 (n - D) = 11 and 00 points merge into a single point. On the other hand, if g0 = 0, then the points Z2 (n) Z2 (n - D) = 01 and 10 merge into a single point. Unfortunately, in each case, each of the merged points has the same corresponding value. As a result, the decoding of bit X2 is performed correctly by decoder bank D Viterbi D 58 without catastrophic error propagation. When the gain parameter gt is reduced to zero, the decoder bank D Viterbi 58 operates only on the difference filter output U (n). Figure 12 shows the impact on the sixteen two-dimensional subsets as the gain gx is adjusted to zero. The SO-SF subsets merge into one-dimensional constellations. According to this, as can be seen in Figure 12, there are only seven unique subsets that are labeled (in the lower left corner) A, Bl, B2, Cl, C2, DI and D2 in Figure 12. Figure 13 shows The seven subsets of the comb filter output constellation. In this way, Figures 12 and 13 show that, when the first filter 42 acts alone (without the second filter 44), the decoder bank D Viterbi 58 operates identically to the procedure described in the US patent application. Serial No. 08 / 272,181 filed July 8, 1994. In this manner, the gain values g0 and gx can be adjusted over a continuum from 0.0 to 1.0, depending on the relative amount of NT? C co-channel interference and White Gaussian noise, present in the received symbols R (n). Consequently, instead of simply switching the first filter 42 in and out, depending on the amount of co-channel NTSC interference present, the first filter 42 and the second filter 44 are relatively adjusted, depending on the relative amounts of NTSC interference of co-channel and Gaussian noise. Figure 14 illustrates a gain adjustment structure 60. The gain adjustment structure 60 establishes the gain values g0 and gx by measuring the interference and noise present in a training channel, which are periodically transmitted and received as a signal training received TR (n). The training signal, before transmission is identical to an ideal reference training signal TS. The training signal is generated by a transmitter and transmitted to the receiver that contains the decoder 40 and which receives the received training signal TR (n). On the transmission path between the transmitter and the receiver, the transmitted training signal picks up any NTSC co-channel interference and white Gaussian noise that are present between the transmitter and the receiver. The ideal reference training signal TS is generated locally by the receiver. The ideal reference training signal TS is supplied to the first and second summers 62 and 64 directly and also to the first and second summers 62 and 64 via corresponding symbol delay elements D and 66. The output of the first adder 62, TU (n) represents the difference between the reference training signal TS (n) and the delayed reference training signal TR (n - D). Accordingly, the output of the first adder 62, TU (n), represents the ideal output U (n). The output of the second adder 64, TV (n), represents the sum of the reference training signal TS (n) and the delayed reference training signal TS (n-D). Accordingly, the output of the first adder 62, TV (n) represents the ideal output V (n). The outputs of the first and second summers 62 and 64 are supplied to third and fourth corresponding adders 70 and 72. The received training signal TR (n) is supplied directly to the fifth and sixth adders 74 and 76 and is also supplied indirectly to fifth and sixth. sixth adders 74 and 76 through corresponding symbol delay elements D 78 and 80. The output of first adder 74 represents the difference between the received training signal TR (n) and the delayed received training signal TR (n-D) ). The output of the sixth adder 76 represents the sum of the received training signal TR (n) and the delayed received training signal TR (n-D). The output of the fifth adder 74 is supplied to the third adder 70, and the output of the sixth adder 76 is supplied to the fourth adder 72. The error eu of the third adder 70 represents the interference and noise on the U (n) axis collected by the signal of training transmitted on the transmission path. The error eu is transformed into a square signal by a square 82 and accumulated by an accumulator 84 to produce a signal Pu representing the power or average energy of the interference and noise represented by the error eu. The error ev of the fourth adder 62 represents the interference and noise collected by the training signal transmitted over the transmission path. The error ev is converted by a quadrat 86 and processed by the accumulator 88 in order to produce a Pv signal representing the average energy of interference and noise represented by the error ev. The average energies Pu and Pv are supplied to a block 90 that determines a parameter? . The parameter? is supplied to a block 92 that determines the weighting of g0 as g0 = eos? and to a block 94 that determines the weighting of gx as gx = sin? . Agree? varies from 0 to p / 2, the minimum distance of the trellis code changes according to the following equation: d 2 (?) = 9 - \ 4 cos2? \ for 0 <;? < p. (i) min 2 The minimum distance of a trellis code is a measure of the coding gain of the convolutional code. The total interference energy at the output of the first and second variable gain amplifiers 54 and 56 is given by the following expression: NT0TAL (T) = Pucos2? + Pvsen2? (2) Block 90 finds the value of? which maximizes the following equation: d 2 (?) min = 9 - 14cos2? \ for 0 < ? < p (3) NT0TAL (T) Pucos2? + P? Se ^? 2 Equation (3) balances the gain decrement of confrme code? it is varied against a reduction of the output interference energy. Under a single white noise condition, Pu and Pv will be equal, and equation (3) is brought to a maximum for? = p / 4. Consequently, gQ = cos? = COSE = / 2 (4) 4 2 And g? = sen? = sinp = 2 (5) 4 2 and the gains are equal in both dimensions U and V.

Under a more general interference condition, when Pu and Pv are not equal, block 90 finds the value of? which adjusts the relative gains g0 and g1 and balances the minimum distance of the code against the interference energy in each dimension. Accordingly, the gain values g0 and gL relatively adjust the outputs of the first filter 42 and the second filter 44, respectively, depending on the relative amounts of co-channel NTSC interference and white Gaussian noise that are collected by the signal of training during its transmission to the receiver that incorporates the gain adjustment structure 60. This training signal, for example may be the frame sync portion of the conventional VSB frame 8. Alternatively, an algorithm-directed decision can be implemented so that an eight-level slice signal, or U (n) and V (n) slices, is used to produce the TU (n) and TV (n) signals. The first and second filters 42 and 44 are well suited for co-channel NTSC interference, because the interference correlation has a peak at a delay of D symbols (where D can be twelve for all delay elements of Figures 1 and 3). Certain modifications of the present invention have been discussed above. Other modifications will occur to those practicing the technique of the present invention. For example, the present invention has been described in terms of co-channel NTSC interference, which is substantially ellipsoidal around the constellation points, and white Gaussian noise, which is substantially circular around the constellation points. However, it will be understood that the present invention is useful in terms of other interference, which is generally not circular around the constellation points and other noise, which is generally circular around the constellation points. Also, filters other than the first and second filters 42 and 44 illustrated in Figure 3 may be employed depending on the noise statistics that are likely to be present. For example, pairs of filters that are not exactly sum and difference filters, but that provide a slightly different rotation of the coordinate axes can be used where the noise statistics are different from those described in relation to Figure 3. Even more , the present invention is described in terms of discrete components. These components can be analog and / or digital components. Furthermore, the functions of these components, on the other hand, can be performed on a computer.

Furthermore, although separate symbol delay elements D 48 and 52 are illustrated for the first and second filters 42 and 44 respectively, the symbol delay elements D 48 and 52 can be replaced by a single symbol delay element D which delays the symbols received R (n) to both of the adders 46 and 50. Accordingly, the description of the present invention will be considered as illustrative only and is for the purpose of showing those skilled in the art the best mode for carrying out the invention. The details may be varied substantially, without departing from the spirit of the invention and the exclusive use of all modifications that fall within the scope of the appended claims is reserved. It is noted that in relation to this date, the best method known by the applicant to carry out the aforementioned invention, is the conventional one for the manufacture of the objects to which it relates. Having described the invention as above, the content of the following is claimed as property: transmitted training signal and delayed transmitted training signal, and wherein the output of the fifth adder provides a first representative error of interference and noise in the transmitted training signal; a sixth adder having first and second feeds and one output, wherein the first feed of the sixth adder receives the reference training signal, wherein the second feed of the sixth adder receives the sum of the training signal transmitted and the signal of delayed transmitted training, and wherein the output of the sixth adder provides a second representative error of interference and noise in the transmitted training signal; means for determining an average energy f ^ of the first error; means to determine an average energy "^ of the second error, means to determine a substantial maximum value, based on the following: d 2 (?) min \ 4cos2? \ for 0 =? = f NTOTA (T) P "uc * os2? + P ^ sin2? Y; means for determining the gains of the first and second variable gain amplifiers based on? 18. The receiver according to claim 1, characterized in that the

Claims (17)

1. CLAIMS 1. A receiver for decoding encoded data, with noise, received from a transmission station, wherein the receiver includes a filtering means for filtering the encoded data, with noise and decoding means for decoding the filtered encoded data, characterized in that the filtering medium variably filters co-channel interference and noise in the coded data depending on the relative amounts of co-channel interference and noise in the coded data, where the filtering media filters the co-channel interference and the noise over the corresponding substantially continuous intervals.
2. 2. The conformance receiver with claim 1, characterized in that the filtering medium inverse filters the co-channel interference and the noise over the corresponding substantially continuous intervals.
3. The receiver according to claim 1, characterized in that the filtering means comprises filter controlling means for controlling relative amounts of filtering performed by the first and second filters depending on the relative amounts of co-channel interference and noise in the encoded data.
4. The receiver according to claim 3, characterized in that the first and second filters have a corresponding power supply or input and first and second outputs, wherein the power supply or input of the first and second filters is arranged to receive the coded data, in wherein the filter control means comprise first and second filter controls, wherein the first filter control is between the first output and the decoding means, and wherein the second filter control is between the second output and the means of decoding
5. 5. The receiver according to claim 4, characterized in that the filter control means comprise: first means for level adjustment, for adjusting a control level of the first filter control over a first continuous interval; and second level adjusting means for adjusting a control level of the second filter control over a second continuous interval; wherein the control level of the first filter control and the control level of the second filter control are relatively varied by the first and second means for level adjustment. The receiver according to claim 1, characterized in that the filtering means comprises: a delay element having a power supply and an output, wherein the power supply of the delay element receives the encoded data and where the output of the element Delay provides delayed coded data; wherein the first filter comprises a first adder having first and second feeds and an output, wherein the first feed or input of the first adder receives the coded data, wherein the second feed of the first adder receives the delayed coded data and wherein the output of the first adder provides a difference between the encoded data and the delayed encoded data; and wherein the second filter comprises a second adder having first and second feeds and an output, wherein the first feed of the second adder receives the coded data, wherein the second feed of the second adder receives the delayed coded data, and wherein the output of the second adder provides a sum of the encoded data and the delayed coded data. The receiver according to claim 6, characterized in that the filtering means comprises first and second variable gain amplifiers, wherein the first variable gain amplifier is between the output of the first adder and the decoding means, and where the second variable gain amplifier is between the outputs of the second adder and the decoding means. 8. The receiver according to claim 7, characterized in that the filtering means comprises means for adjusting gains, for variably adjusting a gain of the first variable gain amplifier with respect to a gain of the second variable gain amplifier, for variable filtering of co-channel interference and noise in the encoded data. The receiver according to claim 8, characterized in that the delay element is a first delay element, and wherein the gain adjustment means comprise: a second delay element having a power and an output wherein the delay element power receives a training signal transmitted, and where the output of the delay element provides a delayed transmitted training signal; a third adder having first and second feeds and an output, wherein the first feed of the third adder receives the transmitted training signal, wherein the second feed of the third adder receives the delayed transmitted training signal, and wherein the output of the third adder provides a difference between the transmitted training signal and the delayed transmitted training signal; a fourth adder having first and second limentations and an output, wherein the first feed of the fourth adder receives the transmitted training signal, wherein the second feed of the fourth adder receives the delayed transmitted training signal, and wherein the output of the fourth adder provides a sum of the transmitted training signal and the delayed transmitted training signal; a fifth adder having a pirmera and second feeds and an output, wherein the first feed of the fifth adder receives a reference training signal, wherein the second feed of the fifth adder receives the difference between the training signal transmitted and the signal of delayed transmitted training, and wherein the output of the fifth adder provides a first representative error of interference and noise in the transmitted training signal; a sixth adder having first and second feeds and one output, wherein the first feed of the sixth adder receives the reference training signal, wherein the second feed of the sixth adder receives the sum of the training signal transmitted and the signal of delayed transmitted training, and wherein the output of the sixth adder provides a second representative error of interference and noise in the transmitted training signal; means for determining an average energy 1 = of the first error; means to determine an average energy? of the second error; means to determine a substantial maximum value? Based on the following: 2 (?) Min 9 - I 4cos2? for 0 < ? < p, NTOT? T. (?) Pucos2? + P S n2 ^ Y; means for determining the gains of the first and second variable gain amplifiers based on? The receiver according to claim 1, characterized in that the decoding means comprises a Viterbi decoder, wherein said decoder is arranged to generate branch metrics for each subset of a plurality of subsets according to a trellis, wherein each The subset contains points that correspond to possible transitions of the encoded data and where the trellis determines paths defined by the possible transitions of the encoded data. The receiver according to claim 10, characterized in that the filter means comprises filter controlling means for controlling relative amounts of filtering performed by the first and second filters depending on the relative amounts of co-channel interference and noise in the encoded data. The receiver according to claim 11, characterized in that the first and second filters have a corresponding power and first and second outputs, wherein the output of the first and second filters is arranged to receive the encoded data, wherein the means of filter control comprises first and second filter controls, wherein the first filter control is between the first output and the Viterbi decoder, and wherein the second filter control is between the second output and the Viterbi decoder. The receiver according to claim 12, characterized in that the filter control means comprises: a first means for level adjustment, for adjusting a control level of the first filter control over a first continuous interval; and a second level adjusting means, for adjusting a control level of the second filter control over a second continuous interval; wherein the control level of the first filter control and the control level of the second filter control are varied in a relative manner by the first and second level adjustment means. The receiver according to claim 10, characterized in that the filter means comprises: a delay element having a power supply and an output, wherein the power supply of the delay element receives the data encoded and wherein the output of the delay element provides delayed coded data; wherein the first filter comprises a first adder having first and second inputs and an output, wherein the first feed of the first adder receives the coded data, wherein the second input of the first adder receives the delayed coded data, and wherein the The output of the first adder provides a difference between the coded data and the delayed coded data and wherein the second filter comprises a second adder having first and second feeds and an output, wherein the first feed of the second adder receives the coded data, in where the second power of the second adder receives the delayed coded data, and wherein the output of the summing second provides a sum of the coded data and the delayed coded data. The receiver according to claim 14, characterized in that the filter means comprises first and second variable gain amplifiers, wherein the first variable gain amplifier is between the output of the first adder and the Viterbi decoder, and wherein the second variable gain amplifier is between the output of the second adder and the Viterbi decoder. 16. The receiver according to claim 15, characterized in that the filtering means further comprises gain adjusting means for variably adjusting a gain of the first variable gain amplifier with respect to a gain of the second variable gain amplifier., in order to filter in a variable way the co-channel and noise interference in the encoded data. The receiver according to claim 16, characterized in that the delay element is a first delay element, and wherein the gain adjustment means comprise: a second delay element having a power and an output, wherein the power of the delay element receives a training signal transmitted, and where the output of the delay element provides a delayed transmitted training signal; a third adder having first and second feeds and an output, wherein the first feed of the third adder receives the transmitted training signal, wherein the second feed of the third adder receives the delayed transmitted training signal, and wherein the output of the third adder provides a difference between the transmitted training signal and the delayed transmitted training signal; a fourth adder having first and second feeds and an output, wherein the first feed of the fourth adder receives the transmitted training signal, wherein the second feed of the fourth adder receives the delayed transmitted training signal, and wherein the output of the fourth adder provides a sum of the transmitted training signal and the delayed transmitted training signal; a fifth adder having a pirmera and second feeds and an output, wherein the first feed of the fifth adder receives a reference training signal, wherein the second feed of the fifth adder receives the difference between the filtering means comprises first and second filters , wherein the first filter has a power to receive the encoded data, wherein the first filter is a difference filter, wherein the first filter has a first output to provide first filtered encoded data corresponding to a difference between the encoded data and the delayed coded data, wherein the second filter has a power to receive the coded data, wherein the second filter is an adder filter and wherein the second filter has a second output to provide second filtered coded data corresponding to a sum of the encoded data and the delayed encoded data. The receiver according to claim 1, characterized in that the filtering means comprises: subtraction means for subtracting the delayed encoded data from the encoded data in order to provide the first filtered output; and, addition means for adding the encoded data and the delayed encoded data in order to provide the second filtered output. The receiver according to claim 1, characterized in that the filtering means comprises: first and second filters, wherein the first filter is arranged to filter co-channel interference in the encoded data, wherein the second filter is arranged to filter the noise in the encoded data; a first variable gain amplifier connected to an output of the first filter; a second vapable gain amplifier connected to an output of the second filter; and, means for adjusting gain to adjust in variable sleep a first gain of the first variable gain amplifier with respect to a second gain of the second variable gain amplifier so that the co-channel interference and noise in the encoded data. The receiver according to claim 1, characterized in that the filtering means comprises: a co-channel interference filter which is arranged to filter the co-channel interference of the coded data; a noise filter that is arranged to filter the noise of the encoded data; and, a controller that is arranged to control in a relative manner the co-canai interference filter and the noise filter so that the co-channel interference and the noise of the coded data are filtered out relatively.