JPH0294913A - Automatic waveform equalizer - Google Patents

Abstract

PURPOSE:To prevent the occurrence of unnecessary tap gain caused by a noise power component in an input signal for a correlative operation by using an input signal taken in with the same timing as past signal. CONSTITUTION:The input signal and an output signal are respectively taken in an input waveform memory 208 and an output waveform memory 209, a differential signal is obtained by executing differential calculation, and next, the correlative operation is executed. At this time, since waveform integral is executed after the correlative operation, a waveform integral result obtained in signals prior to the signals taken in this time is to be used for the correlative operation. On the other hand, a current value is used for the differential signal. Thus, it can be prevented that noise components in the two signals, for which the correlative operation is to be performed, have correlation, the occurrence of the tap gain, which must not appear essentially, can be prevented, and an AGC function essentially provided for a transversal filter 220 can be sufficiently shown.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to an automatic waveform equalization device used in a ghost removal device for a television receiver.

(Prior Art) As a conventional automatic waveform equalization device, there is a device as shown in FIG. This device is known, for example, from Japanese Patent Application Laid-Open No. 59-21131.
It is disclosed in Publication No. 5. This system has already been put into practical use, as described in the literature, R Junz.

Hurakani, Hiroyuki Iaaan
d 5hioeyukiTakehara GHO3
T CLEAN 5YSTEH: IEEE
CE-29° No. 3, August 1983J.

Hereinafter, in FIG. 0 for explaining the apparatus of FIG. 2, a portion 220 surrounded by a broken line is a transversal filter. An input video signal supplied to an input terminal 201 is converted into a digital signal by an analog-to-digital (hereinafter referred to as A/D) converter 202 and then input to a transversal filter 220 . The signal is then subjected to waveform equalization processing in the transversal filter 220 and is output from the output terminal 206. Transversal filter 22
0 is a delay element group 203, a multiplier group 204 provided at each tap of the delay element group 203, and this multiplier group 204.
It has an adder 205 that combines the outputs of.

The output digital signal of A/D converter 202 is further supplied to input waveform memory 208 and timing circuit 213.

The input waveform memory 208 is a circuit that captures and holds an input signal using a waveform capture timing signal 200, which will be described later.

Timing circuit 213 has input waveform memory 208 . Timing signals required by output waveform memory 209 and microprocessor 212.

It generates a clock CK, a waveform capture timing signal 200, and the like.

The output waveform memory 209 stores the waveform capture timing signal 2.
00 captures and holds the output of the transversal filter 220. The contents of this output waveform memory 209 are used for error signal calculation, which will be described later.

The entire operating procedure of the above system is controlled by a macroprocessor 112. Macro processor 112
A program is stored in the ROM 211, and the microprocessor 112 reads the contents of the ROM 211 and performs various processes according to the read program.The RAM 210 is used for the above-mentioned calculations.

In order for the transversal filter 220 to perform a waveform equalization operation, it is necessary to control the tap gain. The tap gains are calculated and modified by microprocessor 212 according to a tap gain modification algorithm written in ROM 211, stored in tap gain memory 207, and provided to each group of tap multipliers 204.

When the input waveform memory 208 captures the input signal and the output waveform memory 209 captures the output signal according to the waveform capture timing signal 200 from the timing circuit 213,
Microprocessor 212 uses these data to perform arithmetic processing according to a tap gain modification algorithm to calculate a new tap gain. The calculated new tap gain value is transferred to the tap gain memory 207 and stored there to be used in the multiplier 204 of the transversal filter 220.
It becomes the count value of the group. By repeating this series of operations, the tap gain value converges to a certain value.

Next, the tap gain correction algorithm stored in the ROM 211 will be explained.

First, each timing necessary for explaining tap gain control will be defined as follows with reference to FIG.

A ghost canceling (OCR) signal having a constant waveform is applied as an input signal to the input terminal 201 in FIG. 2 at a field or frame period as shown in FIG. 3. O
Regarding the CR signal, see Document 2 "Shigeo Matsuura: Broadcasting by inserting a reference signal for ghost canceller control of TV signals - Nikkei Electronics 1987.10.19 (No. 4)
23) p231-p225".

Here, the signal at the input terminal 201 is assumed to be X([).

In FIG. 3, the GCR signal has a pulse waveform, but even a rectangular waveform can be used in the same manner as in this example by performing predetermined processing.

This also applies to the embodiments described below.

The input signal x (t) is analog-to-digital converted before being taken into the input waveform memory 208 .

The tarok cycle of analog-to-digital conversion is T, as shown in (■) in FIG. As shown in the conceptual diagram of A/D conversion (■) in FIG. 3, the sampled discrete signal of the continuous signal X ([) is x (kT), which will be denoted as Xk. Also, the series of Xl will be expressed as (Xk).

Furthermore, let the data string captured into the waveform memory 208 at the νth time, which is the end of one capturing period, be (ν)
The waveform is captured at the timing of the second waveform capture,
Let (X) be the data string stored in the RAM 210 (see 1).

The output signal y (t) appearing at the output terminal 206 is similarly expressed.

Next, FIG. 4 shows a tap gain correction algorithm.

The tag gain correction algorithm will be explained below with reference to the above signal definition and FIG. First, in step 401, one of the input signals is stored in the input waveform memory 209.
The output waveform memory 209 also stores one period of the output signal.
The period (ν) is accommodated in each case. This (X).

(ν) (y l) In step 402, the error meter (ν
) (ν) The error (e) is obtained by subtracting the reference signal (r,) from the output signal (y(). Here, if (ν) is not appended to (r, l) is the reference signal,
This is because the signal has a predetermined shape that does not change over time.

Next, in step 403, waveform integration of the input signal is performed. A method of synchronous addition with leakage is used for this waveform integration. The captured input signal (ν) data string (X l is captured into the input waveform memory at the timing of the previous convergence evening, and is divided into waveform product, (see 1) (X l −K) and are added together, resulting in new waveform integral data, (ν) (X,, l. The waveform integrated data string contains '
(dash) to distinguish it from the data imported into the waveform memory.

The waveform integration used here is not essentially necessary.If the calculation speed in step 404 described below is fast and the correlation calculation range can be sufficiently wide, this waveform integration is not necessary. Since the calculation speed is finite, it is necessary to narrow the correlation calculation range to near the peak of the main signal as an input signal.At this time, if the input signal contains a lot of noise, the main signal of the GCR signal It becomes impossible to accurately detect the peak position of the signal, the position information where distortion exists in the input signal, which is the result of correlation calculation, cannot be accurately determined, and the position indicated by the position information is also unstable. Become something.

For this reason, the tap gain of the transversal filter 220 tends to become unstable. To prevent this, the input signals are synchronously added to reduce the noise level in the input signals, so that the peak position of the main signal can be accurately determined. I have to. Regarding this technology, Japanese Patent Application Laid-Open No. 59-198087
It is described in. At step 404, waveform integrated data is performed. The tap gain is then corrected in step 405 using the obtained correlation result dj. In step 406, the following (x l =”k
l + (nk1 (n,) is a noise component.

Further, the output signal (y,) can be expressed as follows.

(y 1=a(xkl (a is a constant).

Also, the noise is ideal white Gaussian noise. pay.

The above is the tap gain correction algorithm.

This algorithm runs from step 403 to step 40.
There is a problem with the processing method in Section 4.

Let us now examine the correlation calculation in step 404 in detail. To simplify the discussion, assuming that the input signal is distortion-free (ghost-free), the output signal can be written as:

It is assumed that there is no correlation with the reference signal X, that is, k k −〇, where the bar means the average. Therefore, the error signal eH is (ek) = (3'l, l (rkl-(a-
1) Correlation amount d, which is (r l +a(nH))
, is (ν) (ν) dj=η(x H) (ek, jl −'f,, Ca+1)(r, l (ek, 1, (ν) +η・(・k)゛(ν) ( n, ) la+J (i5(a-1) (rk+jl (n k)-〇i
a (rkl (nk, l = o) 2 (ν) From the definition formula of waveform integral, the second term (n) is (ν), (c1) (ν) (n ) = (n ) + (1-K) (n
k) k k (0<K<1).

Substituting this into the second term, we get d, = (a-1) (rk) (rk+j) The above equation becomes d, -〇 when y=o, but when j=0, that is, d. is d = (a-1) 6 moa (to 1-) 6°Or (Σ(r Hr,) ='5+-k
k+J (ν) (ν) Σ(n)(n,)=6 2)k
When d = 0, the tap gain becomes a balanced value and ghosts are removed, so the above formula % formula % (+5.6 2 is the power of the reference signal and noise signal) n Therefore, the output The original AG, which reduces the amplitude of the signal and outputs it in accordance with the magnitude of the reference signal rk.
The C function will no longer function effectively.

Note that since the noise is band-limited, the auto-autocorrelation function of the noise is not an ideal impulse but has a shape of 5inX/X, and the taps near the main tap are also affected by the above.

(Problem to be solved by the invention) As mentioned above, the noise components contained in the two signals used to calculate the correlation have a high autocorrelation, so they should not be multiplied by the main tap. A gain factor appears, which causes the signal to be attenuated at and near the main tap even if the input signal is undistorted. Moreover, this effect causes the disadvantage that the AGC function inherent in the transversal filter is impaired.

Therefore, this invention can prevent unnecessary tap gain from occurring due to noise power components in the input signal when the input signal is in a ghost-free state, and has the AGC function, which is the original function of a transversal filter. An object of the present invention is to provide an automatic waveform equalization device that can prevent deterioration of the waveform.

[Structure of the Invention] (Means for Solving the Problems) In this invention, when there is a ghost-free input signal, the correlation calculation result do is 0 so that no gain is applied to the main tap.
Make it so that To this end, it is sufficient to ensure that the noise component in the signal used in the correlation calculation does not have autocorrelation. Therefore, as the input signal used for the correlation calculation, the input signal taken in at the same time as the error signal is no longer used, but the input signal taken in at the past signal acquisition timing is used.

This can prevent the noise components in the two acquired signals for correlation calculation from having a correlation. Correlation calculation should originally use signal sequences acquired at the same acquisition timing, but in the case of a signal that has a periodic waveform of a predetermined shape, which is the object of this invention, it is necessary to change the input signal. What is given is mainly distortion components such as ghosts, and since the temporal changes of these distortion components are relatively slow, there is no practical problem even if signals with different input signal acquisition timings are used for correlation calculations. No, but it was taken as close to the present as possible (Xk
) is preferable.

(Operation) The above means can prevent noise components in two signals for which correlation calculation is performed from having a correlation.

Therefore, it is possible to prevent a tap gain that should not appear from occurring, and the AGCJll function inherent in the transversal filter can be fully utilized.

(Example) Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows an embodiment of the present invention. The overall block configuration of the device of the present invention is the same as the configuration shown in FIG. 2, and the only difference from the conventional device is the correlation calculation function section.

Therefore, the algorithm for this correlation calculation is shown in FIG. 1 and will be explained using FIG. 2 as well.

In step 101, input signals (x l and output (ν)
) Take in the signal (yl, ). Next, in step 102, (ν) error calculation is performed to obtain (ek) in the first step 1.
In step 03, a correlation calculation is performed. At this time, the waveform product, (see 1
)・min is performed after the correlation calculation, so that the waveform integration result obtained from the signal before the current signal is used for the correlation calculation.

On the other hand, the error signal e converges (ν) in step 101. The current (eh) is used.

The correlation amount dj can be found as follows. Using this, the following equation can be obtained by performing the calculation as explained in FIG. 4 above.

ctj = (a-1) Ax (rk) (rk, ) (C1
) (ν) Here, since there is no correlation between (n) and (nk), the second term becomes 0, and dj=(a-1)ax(rh)(rk+j1.d,=O. Now, the tap gain becomes a balanced value and there is no ghost residual component in the transversal filter output, so if we set d5-0 in the above equation, a=1, that is, the output signal will have the same magnitude as the input signal. , the gain of the main tap is not affected by the noise power in the input signal, so the output signal is not attenuated as in the conventional case.
The problem that the AGC effect inherent in the transversal filter disappears is solved.

After performing the above correlation calculation, the waveform integration of the captured input signal is performed in step 104, and (ν) (X) is obtained.Next, in step 105, the tap gain is modified by d obtained above. ■ will be carried out.

In step 106, the value of ν is counted up by one in preparation for signal acquisition in the next cycle.

In this embodiment, waveform integration is performed for 1xk), but if the speed of correlation calculation is fast, waveform integration is unnecessary. In this case, (C1)
(, -1) (x ) instead of (x k )
will be used for correlation.

In taking the correlation, not only the past input signals close to the input signal xk but also the past input signals are called over-addition.

Alternatively, an average of some of these may be used.

FIG. 5 shows an algorithm using synchronous addition as a second embodiment. This will be explained with reference to FIG.

Step 501 is a signal acquisition step. Here, the past n input signals are stored in the input waveform memory 208.

In step 502, error calculation is performed, and the error (ν) signal (ek) is determined.
(scene n) The average (
The result (X') is obtained. At this time, (X,) taken in step (v) 501 is not used for average calculation. When this is used, as described in the conventional example, the tap gain is affected by the noise component in the input signal.

(v) A correlation is found in step 504 between (X') and (ek) found as described above. Then, in step 505, the tap gain is corrected using this correlation calculation result. In step 506, the past (scene n) input signal data string (n) stored in the input waveform memory 208 is prepared for averaging the newly captured data.

The above is the operation according to the algorithm of the embodiment shown in FIG. In this algorithm, there is no correlation, so as described in the first embodiment, the tap gain is not unnecessarily influenced by noise components in the signal.

Actual noise components are band-limited and their autocorrelation functions have a spread. In this case, there is a possibility that the correlation of the noise will affect taps other than the main taps as well.

Therefore, only when modifying the tap gains of the main tap and its vicinity, the processing method in the above embodiment is used, and when modifying other tap gains, the newly acquired data (ν) (xh) is used. You may use it to perform correction processing.

That is, FIG. 6 further shows the algorithm in the third embodiment.

601 is a waveform acquisition step, in which the input signal and output signal of the transversal filter are input (ν) (
ν) Taking it as (Xk) and (yk), the 0th order, S(
ν) In step 602, the error (ek) is determined.

Next, by branching to step 603, it is determined whether the current calculation concerns the main tap and the taps before and after it. For gain calculations that include the main tap and its surroundings, j = -1
.

However, if j is a tough/
If the calculation is related to a correlation value, a correlation amount step 608 is performed in step 607 after step calculation is performed to check whether the above calculation has been performed for all of the specified j.

When all correlation calculations are completed, proceed to step 609,
Tap gain is actually modified.

Thereafter, rewriting is performed such that shi=shi+1, and the data is on standby in preparation for the next cycle of error detection, correlation calculation, and tap gain correction.

As mentioned in the conventional example, when j=0, (ν)! Since the noise components of xi, ) have no autocorrelation, the tap gain is not affected by the noise power even if the processing is performed as in steps 606 and 607.

(ν) For Yota J=-1, O, +1, (ek) and (sea 1) +e) is used, so the input signal is noisy. It is not affected by sound power.

The gist of the present invention is to use input data in which noise components contained therein have no correlation with the error signal as data for correlation calculation used in tap gain correction. Therefore, the means for executing the algorithms in each of the embodiments described above may be hardware including a microcomputer, or may be entirely hardware that does not include a microcomputer.

[Effects of the Invention] As explained above, the present invention prevents the noise power component in the input signal and its fluctuation from unnecessarily reducing the tap gain of the main tap and its vicinity when there is no distortion of the trough tap gain signal. It can be prevented.

This eliminates the need for an AGC circuit, which was conventionally necessary to correct the transversal filter output that was attenuated due to noise power by the main tap and its neighboring taps.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the algorithm of the main part of the device in one embodiment of the present invention, FIG. 2 is a block diagram showing the waveform equalization device, and FIG. 3 is the signal acquisition operation of the waveform equalization device of FIG. 2. FIG. 4 is an explanatory diagram of an algorithm for tap gain correction in a conventional device.
FIGS. 5 and 6 are explanatory diagrams showing algorithms in other embodiments of the invention, respectively. 202... Analog-to-digital converter, 203... Delay element group, 204... Multiplier group, 205... Adder, 207... Tap gain memory, 208... Input waveform memory, 209... ...Output waveform memory, 210...
RAM, 211... ROM, 212... Microprocessor, 213... Timing circuit, 220...
transversal filter. Applicant's representative Patent attorney Takehiko Suzue Figure 1 C, Figure 3 Figure 3

Claims (5)

[Claims] (1) An input signal containing a reference signal periodically having a predetermined shape is input to a variable tap gain transversal filter, and the input reference signal included in the input signal and the output signal of the transversal filter are combined. A waveform equalization device that takes in an output reference signal included in the transversal filter and performs waveform equalization by sequentially correcting the tap gain of the transversal filter using the result of correlation calculation of both signals by a correlation calculation means, The phase-to-phase calculation means performs a correlation calculation between the output reference signal and an input reference signal taken at least one number of times before the time of taking in the output reference signal, or at least one time before the time of taking in the input reference signal. An automatic waveform equalization device characterized in that it is configured to perform a correlation calculation between an output reference signal captured at least one number of captures ago and the input reference signal. (2) The automatic waveform equalization device according to claim 1, wherein the interphase calculation means takes in the input reference signals a plurality of times and creates a signal obtained by linearly combining the signals to perform the correlation calculation. . (3) For the above linear combination, the input and output reference signals of the current k-th sample (the ν-th acquisition cycle) are expressed as X^(^ν^)_k and y^(^ν^)_k, respectively. Time, X^(^ν^-^1^)_k=Kx^(^ν^-
^2^)_k+(1-K)x^(^ν^-^1^)_k
3. The automatic waveform equalization device according to claim 2, wherein the automatic waveform equalization device performs synchronous addition with leakage as shown in FIG. (4) As for the above linear combination, the input and output reference signals of the current k-th sample (the ν-th acquisition cycle) are expressed as x^(^ν^)_k and y^(^ν^)_k, respectively. Time, x^(^ν^-^1^)_k={x^(^ν^-
^1^)_k+...+x^(^ν^-^n^)_k}
3. The automatic waveform equalization device according to claim 2, wherein the automatic waveform equalization device performs synchronous addition represented by /n. (5) The correlation calculation means performs the correlation calculation only for the main tap of the transversal filter or the tap gain correction in the vicinity including the main tap, and for the remaining taps, the error signal calculated from the output reference signal is 2. The automatic waveform equalization device according to claim 1, wherein the automatic waveform equalization device is configured to perform a correlation calculation with an input signal captured at the same time as the error signal is captured.