JPS617713A - Automatic equalizer - Google Patents

Abstract

PURPOSE:To eliminate the necessity of an input waveform delaying line, to reduce the error in correlation calculation accompanying a change with lapse of time, and to remove the ghost distortion of an automatic equalizer, by providing an interpolator between an AD converter and input waveform memory. CONSTITUTION:Video signals from the input terminal 10 of an automatic equalizer are supplied to a transversal filter 20 and, at the same time, to a correlation data generating means 60, timing circuit 49, and clock generating circuit 50 through an AD converter 42. The output of the filter 20 is supplied to an output waveform memory 45 through an AD converter 44. Moreover, a data generating means 60 is controlled by means of a microprocessor 48 and timing corresponding interpolation data are produced and supplied to an input waveform memory 43. The memory 43 and the output waveform memory 45 are controlled at the timing of the circuit 49. Then the gain of the tap of the filter 20 is corrected by using the interpolation data produced by the data generating means 60 by means of the output of the memory 43 and the error resulting from the change with lapse of time of correlation calculation is reduced without using any input waveform delaying line.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention relates to an automatic equalizer, and in particular to a device that uses a predetermined waveform in a transmission signal to correct the tag gain of a transversal filter to counter waveform distortion in a transmission system. This invention relates to an automatic equalizer that performs equalization.

[Technical background of the invention and its problems]

An automatic equalizer is used as a means for correcting distortion of a transmitted waveform, and is known as a ghost removal device in, for example, television receivers.゛Figure 1 is a circuit diagram showing a conventional automatic equalizer, and shows an example where it is applied to a ghost removal device. (Reference: Haka Murakami, "Ghost Clean System," Toshiba Shiview Vol. 38, No. 7, 1981) . In the conventional automatic equalizer shown in FIG. 1, the transversal filter has a so-called feedforward connection, and the video signal applied to the input terminal 10 is supplied to the waveform equalization circuit 20. 20 is composed of a transversal filter 21, a low-pass filter 22, and a subtracter 23.

The transversal filter 21 includes a tapped delay element 211 that delays the original video signal for a predetermined period of time according to the distortion signal.
, a weighting circuit 212 that determines the weighting amount of this tapped delay element 211, an adding circuit 213 that adds the output of this weighting circuit 212, a tap gain memory 214 that stores the amount to be weighted on the weighting circuit 212, and these circuits. It is composed of a clock driver 215 that generates clocks etc. that define operation timing. Then, subtraction between the video signal including the distorted signal and the low-pass filter 22 that outputs the signal corresponding to the distorted signal is performed in the subtracter 23, and the output terminal 3
At 0, a signal with distortion removed is obtained. Note that the tap gain to be stored in the tap gain memory 214 is calculated as follows.

That is, the input waveform data is stored in the input waveform memory 43 via the input waveform delay line 41 and the A/D converter 42,
Data from the output terminal 30 is stored in an output waveform memory 45 via an A/D converter 44.

The data stored in this output waveform memory 45 is stored in the ROM
Difference calculations are performed in the microprocessor 48 according to the program set in the microprocessor 47.

Is this difference calculation result RA? The signal is stored in I446, and then compared with a reference signal, and the comparison result is stored in RAM46VC as an error signal. This error signal is correlated with the difference signal of the input signal according to a predetermined algorithm to form data for the tap gain memory 214.

In the above conventional automatic equalizer, the delay time T between the tags of the tapped delay element 211, which delays the original video signal to generate the signal for removing the distortion signal, is generated by the clock generation circuit 501C. It is controlled according to the system clock pulse and matches the period of the clock driver 215. In this case, the Ω clock frequency is a value smaller than the reciprocal of twice the highest frequency of the input video signal input to the terminal 10, for example, 93nw (1/3 fsc: fsc is
Subcarrier frequency) K is selected. Here, if the number of tags in the transversal filter 21 is N and the delay time between each tap is T, then the delay time range of ghosts that can be removed is (N-1)T, but in reality, the adder 2
13. The low-pass filters 22 each have a delay time τadd
, τI, PF is shifted, the actual removable ghost distortion delay time is T + τadd + τI, PF or more, NT + τ
add+τLPF or less. Within this delay time Q
Ghost distortion can be eliminated by setting the tap gain (Ci) (the series of C1 to CN is expressed as (Ct)) for the load circuit 212 to an appropriate value according to the flowchart shown in FIG. The component is canceled at the output terminal 30.

In explaining the algorithm for determining the tap gain with reference to the flowchart shown in FIG. 2, for convenience of explanation, it is assumed that the adder 213 and the low-pass filter have no delay time, and the delay time of the input waveform delay line 41 is also ignored. It shall be. Based on this assumption, the process of determining the tap gain using the flowchart shown in FIG. 2 will be explained. First, in step 61, the power is turned on or the receiving channel is switched. In step 62, the input waveform memory 431. The input waveforms ('; c
k), the output waveform (9k) is captured. (here.

xk and yk indicate that they are not differential data. ) i.e.
From the input video signal applied to the input terminal 10, under the control of the timing circuit 49, the leading edge of the vertical synchronizing pulse, which is used as a reference signal for ghost distortion signal detection, is extracted and applied to the input waveform delay line. 41 and an input A/D converter 42, and is stored in an input waveform memory 43 as an input waveform series (9k) of discrete values. Also, a predetermined amount of the video signal at the output terminal 30 at the same time is extracted, and this is output A.
The signal is passed through the /D converter 44 and stored in the output wave memory 45 as an output waveform series (9k) of discrete values.

Next, in step 63, based on the data in the input waveform memory 43 and output waveform memory 45, the microprocessor 4
8 follows the program stored in the ROM 47 and follows the formula below.
...(1)yk=yk+1-yk...
... Perform (2). (In x', yk indicates the superposition result.) And the input difference waveform rxk as the difference result
), the output difference waveform (yk) is stored in the RAM 46.

Thereafter, in step 64, an error calculation is performed between the reference waveform 1rk stored in the ROM 47 and the output difference waveform. This error calculation is a calculation for detecting the amount of ghost distortion as transmission distortion, and is based on the calculation shown in the following equation.

1k=yk-rk ・・・・・・・・・・・・(3)
The error waveform data string (lk), which is the result of the error calculation using equation (3), is stored in the RAM 46. Here, the reference waveform (rk) is for removing only rear ghost distortion,
For example, when removing ghost distortion with a delay of more than 0.5 μs, rk20 may be used.

Next, in step 65, the peak of the input differential waveform data string (xk) performed in step 63 is detected by reading it from the RAM 46. This is a process performed to determine the data range of the data to be subjected to the correlation calculation in the next step 66. If the peak position is −p, the tap gain correction performed in step 66 The correlation calculation for is shown by the following equation.

(i=1..2.....N) In equation (4), ν is the number of tap gain corrections, A.

B indicates p-Ath data to p+Bth data for the pth data. That is, person, B, and p are integer parameters used to indicate the data width. Also α
is the correction coefficient.

In other words, as shown in equation (4), the (v+1) times in the first tap of the tap gain memory 214 include =p at the peak position of the differential data, and are from p- to p+B.

In this way, the i-th tag gain of the (v+1)th time is successively determined. (Reference: 8hinich
i Makino etal °'A Novel
Au＊omatic Ghost Canceller”
IE3Trans(E-26,43p629.

Aug, 1980) The time required for the equalization loop shown in FIG. 2 depends on the parameters A and B, but ghost distortion is usually removed by executing the equalization loop for several fields.

In the above explanation, for convenience of explanation, the delay time τadd of the adder 213, the delay time 22 of the low-pass filter 22, etc. were not taken into account, but in reality, the first tap gain Ci is i XT + ( rad + TLPF
), the above (4) is for removing ghost distortion having a delay time of .

That is, the error calculation data (lk
)-71! (t) and the input differential waveform data (xk)-x(t) is (k+i)T-(kT-(radd+rl, pF)
))=iT+(radd+rbpp)・・・・・・
...It is shown in (6). Thus, the tag gain Ct matches the delay time of the ghost to be removed.

Here, the delay time τadd in the low-pass filter 22 connected to the output side of the adder circuit 213 is the clock cycle normally generated by the clock driver 215.

That is, it is longer than the delay amount T of the unit delay element constituting the transversal filter 211.

Therefore, the sum of the delay time τadd in the adder circuit 213 and the delay time τLpF in the o-bus filter 22 can be expressed as follows.

τadd+rLpr-=y'll'+Δf 7=Q,
l, z−=””””(7)0≦△τkuT
・・・・・・・・・・・・(8) Furthermore, the above equation (5) has i=1.2. ...Can be rewritten to N. As can be seen from equation (9), a time difference of △τ occurs between the input difference data and the error data that are the targets of the correlation calculation, and this causes the problem of generating incorrect data in the correction value of the tap gain. It causes Note that the time difference between the input difference data and the error data is the delay 1 of the unit delay element constituting the transversal filter 211.
: When it is an integral multiple of T, there is no substantial effect, and the tap gain correction amount according to the above equation (9) converges to a predetermined value. However, the time base Δτ, which is smaller than the delay amount T of the unit delay element in the above equation (9), not only causes an offset in the time direction in the tap gain memory 214 and increases residual ghosts. , which causes instability in the automatic equalization according to the algorithm shown in FIG. 2 above. Therefore, in order to eliminate the relative time difference Δτ between the input difference data and the error data that are the targets of the correlation calculation, it is necessary to give the input data a delay time of τd by the input waveform delay line 41. That is, the delay time τd of the input waveform delay line 41 must be adjusted so that τd=Δτ.

As mentioned above, in conventional automatic equalizers that use a correlation method to correct tap gains, when acquiring data to be subjected to correlation calculations, timing is required to eliminate relative time errors between the data. An analog delay line for adjustment is essential. Additionally, this delay line must also pass the video signal. Furthermore, a buffer for matching impedances after the analog delay line is also required. In this way, conventional automatic equalizers require an analog delay line for timing adjustment, and
There is a problem in that the amount of delay changes depending on the change in the analog delay line over time, making the automatic equalization function unstable.

[Purpose of the invention]

In view of the above circumstances, it is an object of the present invention to provide an automatic equalizer that can remove ghost distortion without using an analog delay line for timing adjustment that is subject to changes over time.

[Summary of the invention]

In this invention, in an automatic equalizer that controls the tag gain of a transversal filter by performing a correlation calculation between input difference data and error data, a delay line is used to reduce the timing error between the two data items that are subject to the correlation calculation. First, by performing a linear combination calculation of discrete value input waveforms, interpolated data is generated, and this interpolated data is used to perform a correlation calculation, thereby preventing the influence of timing errors between the data subject to the correlation calculation. to obtain a stable automatic equalization effect.

[Embodiments of the invention]

FIG. 3 is a circuit diagram showing an embodiment of the automatic equalizer according to the present invention, which differs from the conventional automatic equalizer shown in FIG. 1 in that the input waveform delay line 41 is deleted; A feature of the configuration is that an interpolator 60 is newly provided.

In this invention, when performing the correlation calculation for tap gain correction, the timing between the data to be subjected to the nine correlation calculations is not directly adjusted, but interpolated data corresponding to the timing deviation is used. is generated and a correlation calculation is performed using the interpolated data to correct the tap gain of the transversal filter.

In the embodiment shown in FIG. 3, the tap gain is modified to be substantially inverse to the flowchart shown in FIG. 2, but the embodiment is characterized in that interpolation data is used during the mutual calculation.

FIG. 4 is a circuit diagram showing details of the interpolator 6o in FIG. 3, and the interpolator 6o is constituted by a so-called transversal type filter. Unit delay device 6 connected in series in 3 stages
The respective outputs of 02a, 602b, and 602c are added to adders 603a, 603b, and 603c.

The adder/subtractors 604a, 604b, 6
Each output of 04c is multiplied by a predetermined coefficient by a multiplier 60.
The outputs of 4a, 604b, and 604c are added by an adder 605, and the adder 605 receives interpolated data x (nT)
occurs. Furthermore, the data Ω(nT) obtained at the output of the unit delay device 602b and the interpolated data '? (nT) is applied to the selector 606, which is controlled by a control signal from the microprocessor 48 and outputs the data applied to the input terminal 601 to the output terminal 607.K interpolated data Q(nT), non-interpolated data x (nT
). That is, output 5 of output terminal 607? c(n, T) becomes interpolated data as necessary.

Next, the interpolation calculation by the interpolator 60 will be explained.

Now, if data Q((n+2)T) is input to the input terminal 601, the unit delay unit 602a +602
Q((n+1
)T).

Obtain data x(nT), Q((n-1)T). Here, multipliers 604a, 604b, 604
The multiplication coefficients at c are K/2°1/2. /2, the output of the adder 605; (nT) is expressed by the following equation.

Interpolated data zcnT) sThe Fourier transform of the original data R'(n) is respectively S? (jw) and x(jw), and when the above equation C11 is Fourier transformed, x(
jw)=-(x(jw)-1-Q(jw)lljWT)
A(Q(jw)II jWT+x(jw))−(xcj
w)ljW2'rtenQ(jw)l! −jWT ) =H
(jw) ・x(jw)・・・・・・・・・・・・・ aυ If we pay attention to the phase characteristic of the above formula (12), it shows the additive property, and the amplitude characteristic I H(jw) l is shown as. Therefore, the transfer function of the interpolator 60 changes the phase of the input data according to the (l:1 equation), and changes the absolute value according to the above first equation. 0 above
This is a characteristic diagram showing the frequency and amplitude characteristics when = 174 in equation 4), and the amplitude changes as shown in the figure, but the change is 0 up to around 3.58 MHz, which is the color carrier frequency. It is between .75 dB and 1.1 dB, which poses no practical problem.

In order to successively correct the tap gain stored in the tag gain memory, when performing correlation calculation in the algorithm according to equation (4) above, Δτ
By performing the correlation calculation using the interpolated data when there is a timing shift, it is possible to prevent a tap gain correction malfunction due to a malfunction of the correlation calculation operation. In this way, Δτ
When there is a timing difference between the correlation calculation data by As can be seen from FIG. 4, it is delayed by a time 2T compared to the output signal 9 of the output terminal 30 at the same time, which is the target data. That is, when performing a correlation calculation using interpolated data, the correlation calculation may be performed using error data corresponding to data obtained by delaying the output waveform data from the output terminal 30 at the same time by two sample periods (2T). In other words, when modifying the tap gain depending on the correlation calculation using interpolated data, 1(k-2) is used instead of the error waveform data lk in equation (4) above.
Let be correlation data, i=:1 , 2 ,...N
(Qk indicates interpolated data) Correlation calculation is performed using interpolated data.

Furthermore, the correlation calculation at the time of data sampling at time nT, which does not require the use of interpolated data, may be performed to correct the tap gain by performing the correlation calculation according to equation 00 without using the interpolated data.

Using the above aa and ae formulas, input waveform data
Tag gain correction is performed using interpolated data obtained when advancing by /2. On the contrary, when input waveform data is delayed from the output waveform by T/2 relative to the error signal lk that is the target of correlation calculation, input interpolated data x
(nT') and an error signal xk-i delayed by one sample period with respect to this.

i=1.2. ...N All you have to do is perform the following calculation.

In this way, with respect to input data, data whose timing is shifted by Δτ with respect to nT at the time of sampling is generated by interpolation calculation, and a correlation calculation is performed using the input interpolation data, thereby correcting the tag gain. The maximum timing deviation in tag gain correction according to this embodiment is:

It is indicated by. This timing deviation of the degree shown by equation (181) does not pose any practical problem.

In addition, in order to effectively reduce the maximum value of the timing deviation shown in the above equation I by half, it is necessary to use both the interpolated data by the interpolator 60 according to the above equations (15 to 19) and the data before the interpolation calculation. A similar interpolation calculation using .

As mentioned above, in this embodiment, in order to correct the tag gain by calculating the correlation between the input waveform data and the error data, the timing difference between the two data is corrected as in a conventional automatic isograph. , by performing an interpolation operation that generates interpolated data for input waveform data without using an input waveform delay line, which is an analog delay line for timing adjustment, the timing difference between the data to be computed in the correlation operation is corrected and tapped. Optimize the amount of gain modification. Therefore, malfunctions in tap gain correction due to changes over time in the analog delay line used for timing adjustment, which has been a problem with conventional automatic equalizers, can be improved. Furthermore, as shown in FIG. 4, the interpolator 60 used in this embodiment can be integrated on one chip with a simple configuration consisting of a unit delay element using a CCD, an adder, and a multiplier. Since the interpolator itself can be realized by an adder using a bit shift operation when the multiplication coefficient is a power of 2, an all-digital interpolator can also be integrated on one chip.

In the embodiment described above, an example was shown in which the interpolation operation for input data is performed using the interpolator 60, but in the embodiment shown in FIG. You can do it. That is, the interpolation calculation may be performed under the control of the microprocessor 48.

FIG. 6 shows a flowchart in this case. In the same figure, steps 61 to 65 are similar to the operations in the tag gain correction shown in FIG. 2 described above, so their explanation will be omitted.

After detecting the peak value of the input waveform difference data string, which is the result of performing a difference calculation on the input waveform data, in Stella '765 in FIG. 6, an interpolation calculation is performed in step 68. The interpolation calculation in this case follows the above-mentioned Fuebe formula, and the opening formula is set to =-. In addition, the amount of interpolated data string to be subjected to correlation calculation required for one tap gain correction for a given tap, that is, the number of calculations for interpolation calculation is P
Since −A≦n≦P+B, ())+B>
-(P-A)=(A+B) times. Practically speaking, this number of times may be about 15 times. In this way, step 68
When interpolated data is generated as a result of a predetermined number of interpolation calculations, tap gain correction is performed in step 69 using this interpolated data. When correcting the tap gain in step 69, error data to be used in correlation calculation is selected and correlation calculation is performed according to the timing deviation of ±T/2 with respect to the unit delay time TK, which is the tap interval.

This is due to the unit delay time T when performing the correlation calculation using the interpolated data
0.6) depending on whether the interpolated data
The microprocessor 48 selectively controls whether to perform the correlation calculation using the formula or the correlation calculation using the formula (17). Note that if the timing difference between the input difference data string and the error signal string is an integral multiple of the unit delay time T of the transversal filter, the tap gain is corrected by the correlation calculation according to the above-mentioned formula (15i). .

Considering the time required for interpolation calculations, the main instructions required for interpolation calculations are: a load instruction to transfer data from the RAM 46 to the microprocessor 48;
Addition/subtraction instructions, bit shift instructions to halve data,
These include store instructions that transfer data from microprocessor 48 to RAM 46. In the case of an 8-bit microprocessor, each of the above instructions takes an average of several microseconds, so for convenience, we assume that each instruction requires 3 microseconds to execute. The instructions required to execute the first term are load instruction x 2, addition instruction x 1,
There are 4 instructions in total (bit shift instruction x 1), and the instructions required to execute the second term are 6 instructions in total (load instruction x 2, addition instruction x 1, and bit shift instruction x 3), and the instructions required to execute the third term are: The number of instructions required is the same as the number of instructions required to execute the second term. Therefore, the number of instructions required to perform the above-mentioned interpolation operation is 20 instructions, which is the addition of the store instruction to the three instructions for adding and subtracting each term of the +11th equation. If the number of calculations A+B of the above interpolation calculation is 15, the time t required to perform the interpolation calculation is t = 20xl 5x3 = 0.9 (ms)
--It becomes OI.

The time required for the interpolation calculation shown by this equation 01 is:

Considering that the time required for one equalization loose operation is several fields (one field is 16.7 m5), this is a sufficiently practical time.

In this way, the interpolation calculation may be performed by the operation of the microprocessor 48 instead of by the interpolator 60 shown in FIG.

As described above, in the automatic equalizer according to the present invention, when performing correlation calculation and correcting the tag gain of the transversal filter to perform waveform equalization, For a difference in delay time relative to Instead of dealing with this by using interpolated data, we deal with it by generating interpolated data. In other words, for the circuit that performs the correlation calculation, when transmitting the data to be calculated, there is a timing shift during the correlation calculation due to the delay element between the data transmission paths, and sampling every unit delay time T of the transversal filter. By generating interpolated data from the data and performing a correlation calculation using this interpolation data, the error in the correlation calculation result is reduced and automatic equalization is carried out automatically.

Therefore, in the present invention, the circuit configuration including the transvaal filter as shown in FIGS. 3 and 6 is not limited to a cyclic type, but has an original signal path as shown in FIG. Even if the configuration does not include an output side adder/subtractor, the microprocessor 48 may generate interpolated data.

In addition, in the above-described embodiment, the transversal filter is configured to remove rear ghost distortion, but as shown in FIG. Even in a configuration in which an operating delay element 216 is provided, by calculating interpolated data with the microprocessor 48, the influence of relative timing deviation between data that is the target of correlation calculation can be reduced without using an analog delay line. can.

Further, the correlation calculation in modifying the tap gain of the transversal filter according to the present invention is other than the correlation calculation listed in the above equation (4).

A type Σxke 5g that treats the error signal as code data and performs a correlation calculation between this and the difference data for the input waveform.
n1k+1 ----・t2Bk Treat the difference data for the input waveform as code data,
Format % formula % that performs a correlation calculation between this and an error signal treated as code data Format % formula % that uses data whose sampling timing is shifted by 1 as input waveform data and performs a correlation calculation between this and error data % () Correlation calculation format that shifts the sampling timing of the input waveform data by the jth amount and treats it as error data % () 1. Shifts the sampling timing of the input waveform data by the jth amount and treats the error data as code data as well. It is also effective for modifying the tap gain using any correlation format such as the correlation calculation format % expression %.

Furthermore, the tap gain correction may be an incremental correction in which the amount of correction at one time is a fixed value Δ.

Furthermore, it goes without saying that the automatic equalizer according to the present invention may be configured to generate interpolated data by a logic circuit without using a microprocessor to reduce the influence of timing deviation between correlated data.

〔Effect of the invention〕

As described above, according to the automatic equalizer according to the present invention, when performing a correlation calculation between input signal data and error data to be equalized when performing waveform equalization using a transversal filter, the difference between both phases is Since timing errors between data are generated as interpolated data without using a delay line and used as correlation data, errors in correlation calculation data can be reduced without accompanying changes in red time.

In other words, a timing error, which is a relative data transmission time error, occurs between each correlated data transmission path, and an analog delay for timing adjustment is generated by generating interpolated data based on the input discrete data of the transversal filter. The line is considered incondylar.

Therefore, according to the present invention, it is possible to eliminate the need for an analog delay line, thereby eliminating the need for a buffer circuit, etc., thereby providing an automatic equalizer that is a simple circuit and exhibits stable operation. It is.

[Brief explanation of drawings]

@Figure 1 is a circuit diagram showing a conventional automatic equalizer, and Figure 2 is a circuit diagram showing a conventional automatic equalizer.
FIG. 3, FIG. 4, FIG. 7, and FIG. 8 are circuit diagrams showing embodiments of the automatic equalizer according to the present invention, and FIG. FIG. 6, a characteristic diagram for explaining the operation of the automatic equalizer according to the present invention, is a flowchart for explaining the operation of the automatic equalizer according to the present invention. lO...Input 71 child, 20...Transversal filter % 30...Output terminal, 214...Tag gain memory, 43...Input waveform storage means, 45...Output waveform storage means, 48 ...Tap gain correction means, 30.44.45...Data transmission line, 48.60...
- Interpolation data generation means. Agent Patent Attorney Kensuke Chika Figure 2 Figure 3B Year 4wI n Figure 7 103° Figure 8 20

Claims (1)

[Claims] An input terminal into which an input transmission signal to be waveform-equalized is input; a transversal filter that is delayed by an integral multiple of a delay time and has a tap gain memory that stores a tap gain for the tap and contributes to removing a distortion signal included in the transmission signal; input waveform storage means for converting the equalized output waveform signal obtained at the output side of the transversal filter into discrete values to generate input discrete data and storing the same; output waveform storage means for storing, and tap gain correction for calculating a tap gain correction amount for the tap of the transversal filter using the output data of the input waveform storage means and the output data of the output waveform storage means. means, an input data transmission line for applying data corresponding to the input transmission signal to the tap gain correction means, and a relative signal time difference between the output discrete data and the input discrete data is a unit of the transversal filter. Integer multiple (nT) and non-integer multiple (△
and interpolated data generating means for generating interpolated data corresponding to the non-integer multiple ΔT using the input discrete data, when the input discrete data is expressed as the sum (nT+ΔT) of the input discrete data The error in the tap gain correction due to the non-integer multiple (ΔT) in the relative signal time difference between the data and the output discrete data is corrected by interpolation data to prevent malfunction of the waveform equalization effect. Automatic equalizer.