JPH0666637B2 - Digital waveform shaping filter - Google Patents

Description

The present invention relates to a digital waveform shaping filter suitable for teletext broadcasting and the like. More specifically, the present invention relates to a digital waveform shaping filter suitable for, for example, generating a character signal transmission spectrum in a character broadcasting transmitter or a character signal generator for measurement with high accuracy, high stability and at low cost. It is a thing.

In recent years, much research has been conducted on teletext. In such a teletext, the text signal is transmitted as a binary NRZ signal by being multiplexed with the VBL of the video signal of the television, so that it is affected by the characteristics of the television transmission system.

The eye opening ratio and the eye opening area ratio are known as the signal quality evaluation parameters of the above-mentioned binary NRZ signal. Here, the eye opening ratio is a level difference between a steady value of "1" level and a steady value of "0" level when a large number of pulse waveforms of a character signal are superimposed in synchronization with the clock frequency of the pulse transmission signal. , The ratio of the level difference between the waveform of the minimum value of "1" level and the waveform of the maximum value of "0" level,
In FIG. 1, it indicates the ratio of the length of the straight line ef to the length of the straight line ab. The eye opening area ratio is the ratio of the eye opening area to the complete rectangular pulse rectangular area, and is the ratio of the hatched area to the rectangular abcd area in FIG.

And, the fact that both the eye opening ratio and the eye opening area ratio are large indicates a desirable transmission system. From this point of view, as a result of considering the characteristics of the television transmission system, the 60% cosine roll-off characteristic that the eye opening ratio is 95% or more and the eye opening area ratio is close to the maximum value as shown in FIG. It is used as a character signal transmission spectrum. In the teletext transmitter or the character signal generator for measurement, the following means have been conventionally adopted to realize such characteristics.

The first means known in the prior art is, as shown in FIG. 3, a character signal generator 1 for generating a character signal 3 by a binary NRZ signal.
And introduce this into LC type waveform shaping filter 2
A character signal 4 having a 60% cosine roll-off characteristic was obtained.

With such means, the 60% cosine roll-off characteristic can only be obtained approximately due to the circuit configuration. Also, the influence of temperature characteristics, the binary NRZ signal of the filter input is exactly 50%.
Considering each point such as not being the duty ratio,
In this direction, it is difficult to obtain good accuracy and stability, and it is unsuitable as a reference spectrum generating circuit for examining the influence of the transmission system on the spectrum.

The second means known in the prior art is to introduce the binary NRZ signal 3 generated from the character signal generator 1 into the shift register 5 as shown in FIG. The gain is adjusted via the attenuator 6, and the result is analogically added by the adder 7 to form the waveform. That is, it has a so-called transversal filter.

The reason why the waveform can be formed by the above-mentioned second means will be described below.

First, the impulse response of the filter will be mentioned. Generally, the impulse response of the filter is defined as the filter output when the unit impulse UP is input to the filter FL as shown in FIG. Here, the input signal x (t) is considered to be a continuous impulse train as shown in FIG.
That is, the impulse at time t is considered to have a magnitude of x (t). Therefore, the output signal y (t) is considered to be the sum of the responses of this continuous impulse train. For example, in FIG. 6, the input signal x (t)
Is input to the filter FL, the output y at time t ₀
(T ₀ ) is the level a at the time t ₀ of the response of the impulse having the magnitude of x (t ₂ ) at the time t ₂ and the response b at the times t ₁ , t ₀ and t _−1. , c, d etc. is the sum of the responses from the past to the future. The output y (t) thus obtained is expressed by the following equation.

Here, h (t) is the impulse response of the filter FL.

It is known that the above equation is expressed by the following equation: Y (ω) = H (ω) · X (ω) (2)

Here, Y (ω) is a representation of y (t) in the frequency domain, and H (ω) is
(Ω) is the representation of h (t) in the frequency domain, and X (ω) is x
It is an expression in the frequency domain of (t).

Now, assuming that the input signal is a binary NRZ signal, the transmission spectrum of the character multiplex signal is 2 as shown in FIG.
Signal y with 60% cosine roll-off spectral characteristics for isolated "1" pulse x (t) of value NRZ signal
(T), and H (ω) = Y (ω) / X (ω) (3) from the equation (2) is calculated as H (ω) (see H (ω) in FIG. 7). All you have to do is configure the filter you have.

Next, the equation (1) is converted into a sampled value sequence x (nT), y (nT), h (nT) for x (t), y (t), h (t) for each time interval T [N], y (n), h (n)]. FIG. 8 shows a sample series x of the signal shown in FIG.
It is expressed using (n), y (n), and h (n).
That is, the output y (n) at the sampling time point n
Is the response of the input value x (n + 8) at the time of n + 8 a
And the response of x (n + 5), x (n) and x (n-7)
It is the sum of the responses from the past to the future such as b, c and d, and is expressed by the following equation.

The above equation corresponds to the equation (1) representing the filter output of the continuous signal. Then, the band of Y (ω) is set to ω ₁ (the seventh
It is known as a sampling theorem that the continuous signal y (t) can be restored from the sample value series y (n) if the condition of T ≦ π / ω ₁ is satisfied. Therefore, according to the time interval T satisfying this condition, x (t) [binary NRZ signal] shown in FIG.
And h (t) [impulse response of the filter] are sampled (see FIG. 9) and equation (4) is applied, a sampled value sequence y (n) of the 60% cosine roll-off spectral signal y (t) is obtained. ) (See FIG. 9). Further, since the value of h (n) decreases as n increases, the equation (4) is approximated by the following equation using this property.

The above equation is realized by the circuit shown in FIG. 10, that is, a circuit obtained by digitizing a part of the circuit shown in FIG. Here, 10 is a 1-clock delay element, 11 is a read clock, 12
Is a multiplier and 13 is a D / A converter. Here, since the input x (n) is a binary NRZ signal, it can take only a value of "1" or "0". Therefore, the 1-clock delay element 10 can be replaced with a shift register. Further, it is a circuit for restoring the continuous signal y (t) from the sample value series y (n) of the outputs of the D / A converter 13 and the low-pass filter 8. Thus, in FIG. 10, the multiplier 12,
Although the adder 7 and the D / A converter 13 are used, the circuit in which these parts are analog circuits is the same as the above-mentioned fourth circuit.
Corresponds to the figure. In other words, the attenuator 6 shown in FIG.
The function of distributing the output of the shift register 5 by resistance is fulfilled, and it functions as the multiplier 12 in FIG. The value of the coefficient h (n) shown in FIG. 10 corresponds to the value given by adjusting the resistance value of the attenuator 6 shown in FIG.

However, the second means described above (see FIG. 4)
Has a drawback that an error occurs due to the temperature characteristic of the attenuator and the analog addition accuracy, and further, the attenuator cannot be easily adjusted to obtain the 60% cosine roll-off characteristic, and therefore the reproducibility and maintainability are poor. Seen.

The third means known from the past is configured by all digital circuits of the circuit shown in FIG. 10, and as shown in FIG. 11, each tap output 15 [x
(N + N) to x (n-N)] and each tap coefficient signal 14 [h
(-N) to h (N)] are respectively multiplied by the digital multiplier 16, and the results are digitally added by the digital adder 17, and then the D / A converter 13 and the low pass filter 8 are used. By doing so, a 60% cosine roll-off spectrum signal 4 is obtained.

The third means can obtain a highly accurate and highly stable output as compared with the second means, but it requires a large number of multipliers and adders, and thus it is difficult to achieve cost reduction and downsizing. It was

In view of the above points, an object of the present invention is to provide a digital waveform shaping filter that obtains good characteristics despite a simple configuration.

Next, as a prerequisite technique for explaining one embodiment of the present invention, the digital waveform shaping filter shown in FIG. 12 will be described first. 18 is a read-only memory ROM
, And the other numbers 1, 3, 5, 11, 15, 13, 8, 8 are as described above.

In FIG. 11 described as the conventionally known third means, the portion surrounded by the broken line is, as described above, the expression (5). Is a concrete circuit. Where input x (n
+ N) to x (n−N) take a value of “1” or “0” (because the input signal x (t) is a binary NRZ signal), and impulse responses h (−N) to n (N) ) Represents H (ω) calculated by the equation (3) on the time axis (see h (t) in FIG. 7),
Sampled at T intervals (see Fig. 9 h (n))
Therefore, it is possible to carry out the calculation in advance to obtain the respective predetermined constant values. Considering this, the expression (5) represents that y (n) is uniquely determined by the bit pattern of x (n + N), x (n + N−1), ..., x (n−N). Will be there. Therefore, using the ROM 18 as shown in FIG. 12, the address shown by the bit pattern of x (n + N), x (n + N−1), ..., X (n−N) is calculated based on the equation (5). By previously writing the calculated value of y (n), a circuit having a function equivalent to that of FIG. 11 can be obtained. Moreover, this brings about an effect that the number of required elements is drastically reduced. By the way, regarding the part surrounded by the broken line in the 11th figure, 2N + 1 multipliers and 2N adders
A total of 4N + 1 elements are required, but in the case of the circuit configuration of this embodiment, only one ROM is required.

Next, consider the storage capacity required by the ROM 18 in the circuit configuration shown in FIG. Since the address bus is 2N + 1 lines, this ROM 18 requires a storage capacity of 2 ^{2N + 1} × M bytes when M bytes are configured per address. Therefore, if the number of taps must be increased to shape a desired waveform, a ROM with a very large capacity is required.

Therefore, an improvement for reducing the storage capacity, that is, one embodiment of the present invention will be described next.

The above-mentioned equation (5) can be decomposed as follows, for example.

According to the above equation, y (n) is x (n + N), and x (n + N-1) , x y 1 , which is uniquely determined by the (n) becomes the bit pattern (n), x (n- 1) , x (n−2), ..., X (n−N) means a sum with y ₂ (n) uniquely determined by the bit pattern. Therefore, as shown in FIG. 13, the ROM is divided into two, and for one of the ROMs, for the address indicated by the bit pattern of x (n + N), x (n + N-1), ..., x (n) y ₁ (n)
For the other ROM, y ₂ (n) is stored for the address indicated by x (n-1), x (n-2), ..., x (nN), By adding these read outputs with the adder 17,
Functions equivalent to those in Figure 12 are realized. Thus, it can be said that FIG. 13 shows the first reference example of the present invention.

The circuit shown in FIG. 13 has two ROMs as compared with the circuit shown in FIG.
Although it doubles, similarly, if it is assumed that one address consists of M bytes, the storage capacity required is ROM that outputs y ₁ ; address bus is N + 1, so 2 ^{N + 1}
ROM that outputs × M bytes y ₂ : 2 ^N × M bytes because there are N address buses.

Therefore, a total of 2 ^{N + 1} × M + 2 ^N × M = 2 ^N (2 + 1) × M = 2 ^N
× 3 × M bytes will suffice.

Furthermore, the formula (5) is y (n) = h (-N) x (n + N) + h (-N + 1) x (n + N-1) + ... + h (o) x (n) + ... + h (N) x (n -N) means that if each term is arbitrarily combined and y (n) is expressed as L partial sums, Becomes Here, mi represents the number of terms forming each partial sum.

Therefore, if each partial sum is converted into a ROM and the circuit configuration shown in FIG. 14 is provided, it becomes equivalent to the circuits shown in FIGS. 12 and 13. The second reference example shown in FIG. 14 has ROM L pieces and adders (L−1) pieces which are required, and is compared with the configuration shown in FIG. 2L-2) requires a large number of elements, but the storage capacity required for ROM is Becomes Therefore, by properly selecting the values of L and m _{1 to} m _L , it is possible to configure a circuit in which the storage capacity of the ROM can be extremely small without increasing the number of elements excessively.

Next, the relationship between the binary NRZ signal and the sampling frequency (1 / T in FIG. 9) will be described. The sampling frequency (1 / T) shown in FIG. 9 is an example displayed on the assumption that it is three times the clock frequency of the binary NRZ signal in order to satisfy the above sampling theorem. . Therefore,
When a binary NRZ signal is sampled, "1" of the binary NRZ signal
X (n) corresponds to 3 and 1 corresponds to 3 and 0 respectively.
Therefore, "0" has three consecutive values (see x (n) in FIG. 9). In other words, in the shift register 5 in each of the reference examples shown in FIGS. 13 and 14,
It means that "1" or "0" is shifted in groups of three. FIG. 15 is a waveform diagram showing this situation.

FIG. 15 is an explanatory diagram showing the signal waveform of the filter output when a binary NRZ signal is input to the filter, based on the format of FIG. That is, in order to shape the pulse train of a binary NRZ signal, a signal with a 60% cosine roll-off spectrum, which is the filter response of the isolated pulse "1" (see 19 to 23 in Fig. 15), is binary. This figure shows that the NRZ signals may be superposed according to the pulse trains “1” and “0” and output at time intervals satisfying the sampling theorem. For example, consider the superposition of five binary NRZ signals, as shown in FIG. Then, as is clear from this figure, the time
Each output at t ₀ , t ₁ and t ₂ is It is expressed by Here, y ₀ , y ₁ and y ₂ are waveform shaping outputs at times t ₀ , t ₁ and t ₂ , respectively, and S1 to S5 correspond to “1” and “0” of the binary NRZ signal. Takes a value of "1" or "0". a _{-2 to} a ₂ , b _{-2 to} b ₂ , c _{-2 to} c ₂ are signals 19 to ₂ shown in FIG. 15 at times t ₀ , t ₁ and t ₂ , respectively.
It has a value of 3.

FIG. 16 shows a third reference example of the present invention, which is configured to realize the above matrix operation.

In the figure, as already mentioned, 1 is a character signal generator,
Reference numeral 5 is a shift register, S1 to S5 are bit outputs of the shift register, 18 is a ROM, 13 is a D / A converter, 8 is a low pass filter, and 4 is a 60% cosine roll-off spectrum signal.

Further, 24 is a 1/3 frequency dividing counter, the 1/3 frequency dividing pulse 25 is a shift pulse for shifting the shift register 5, and the two counter outputs A _{0 to} A ₁ are ROM 18
Used to specify the address of. When the generation period of the binary NRZ signal transmitted from the character generator 1 is 1 / fc seconds, the frequency of the input pulse 27 introduced into the 1/3 counter 24 becomes 3fc and both of these signals are generated. Is preliminarily configured to be transmitted in synchronization. Therefore, the addressing to the ROM 18 is performed with the frequency 3fc, while the fetching of the binary NRZ signal into the shift register 5 is performed with the frequency fc.

Here, in the ROM 18, the accumulated data (S
For each address corresponding to the bit pattern formed by 1～S5) and 1/3 counter output A _0, A _1, are stored respectively the values of y _0, y ₁ and y ₂ of (6), wherein Keep it. Therefore, y ₀ , y ₁ and y ₂ are sequentially read from the ROM 18 according to the address change of the 1/3 counter outputs A ₀ and A ₁ (read cycle is 1/3 fc). afterwards,
Only one bit of the new binary NRZ signal is taken into the shift register 5, and new y ₀ , y ₁ and y ₂ are sequentially read again. Hereinafter, the same operation is repeated. When this operation is compared with the operation shown in FIG. 12 described as the prerequisite technique, y ₀ , y ₁ and y ₂ are changed to the 12th time at times t ₀ , t ₁ and t ₂ .
It corresponds to reading from the illustrated ROM 18, respectively.

FIG. 17 shows an embodiment of the present invention. In the present embodiment, the ROM 18 is divided into three, and the outputs y ₀ , y ₁ and y ₂ are stored in separate ROMs for the addresses corresponding to the bit patterns of S1 to S5, and the data selector 26 To sequentially output y ₀ , y ₁ and y ₂ . Here, the output switching is performed in synchronization with the read clock 11.

That is, in FIG. 17, in order to output y ₀ , y ₁ , and y ₂ from each ROM, for each address shown in the bit pattern of S1 ... S5 in each ROM, the following formula (6) is used. The calculation result of is stored.

y ₀ = a _-2・ S1 ＋ a _-1・ S2 ＋ a ₀・ S3 ＋ a ₁・ S4 ・ a ₂・ S5 y ₁ ＝ b _-2・ S1 ＋ b _-1・ S2 ＋ b ₀・ S3 + b ₁・ S4 ・ b ₂・ S5 y ₂ ＝ c _-2・ S1 ＋ c ₋₁・ S2 ＋ c ₀・ S3 ＋ c ₁・ S4 ・ c ₂・ S5 Moreover, if the frequency of the read clock of each ROM is fc [Hz], the frequency of the read clock 11 applied to the data selector 26 is 3・ It becomes fc [Hz].

Next, the RO required in this embodiment shown in FIG.
The M capacity is compared with the ROM capacity required in FIG. 12 (base technology) and FIG. 13 (first reference example). For convenience of explanation, assuming that the total number of taps shown in FIG. 12 (that is, the number of shift register outputs) uses M bytes of ROM per J, 1 address, the present embodiment (FIG. 17): The number of taps of the shift register required to obtain the same degree of accuracy as the circuit of FIG. 12 is 1/3 of the number of taps of the shift register in FIG. 12 (see FIG. 15). Therefore J /
Three ROMs with three address buses are required, and a total of 3 ROMs requires 2 ^{J / 3} x 3 x M bytes.

In the case of Fig. 12 shown as a prerequisite technique: J address buses are required and ROM storage capacity is 2 ^J ×
It becomes M bytes.

In the case of the first reference example (Fig. 13): Since two address bus J / 2 ROMs are required, 2
A total of 2 ^{J / 2} x 2 x M bytes are required.

As is clear from the above, it is understood that the ROM capacity required in this embodiment (FIG. 17) is remarkably small as J increases, as compared with FIGS. 12 and 13. Thirdly, the same applies to the reference example (see FIG. 16). In other words, in consideration of the number of elements, the cost, the required ROM capacity, etc., the third reference example (FIG. 16) and this embodiment (FIG. 17) are better than other embodiments. It has an excellent circuit.

Furthermore, by replacing the above-mentioned ROM with RAM and rewriting the stored data as needed, it is possible to easily form waveforms having various spectra.

Finally, the effect of the present invention will be described.

The first reference example (see FIG. 13) and the second reference example (see FIG. 14)
In the drawing), by dividing the ROM, a configuration using a ROM with a smaller capacity is possible while maintaining the configuration of the simple waveform shaping section.

Furthermore, focusing on the fact that waveform shaping can be performed by superimposing the 60% cosine roll-off spectrum signal that is the output of the isolated pulse “1”, the third reference example (see FIG. 16) and this implementation An example (see Figure 17) was constructed. This embodiment is superior to the first to third reference examples in terms of the number of elements, cost, ROM capacity, and the like from a comprehensive viewpoint.

As described above, according to the present invention, the effect that the configuration of the digital waveform shaping filter can be further simplified can be obtained.

Further, according to the digital waveform shaping filter of the present invention, the characteristics such as stability and accuracy of the character signal transmission spectrum can be significantly improved as compared with the LC type waveform shaping filter and the analog type transversal filter. it can.

Note that the present invention can be applied to various fields of digital data transmission by utilizing the characteristic of the present invention that digital data can be shaped into a waveform having an arbitrary spectrum with a simple circuit.

[Brief description of drawings]

FIG. 1 is a diagram illustrating the eye opening ratio, FIG. 2 is a diagram showing the relationship between the descending coefficient and the eye opening ratio and the eye opening area ratio, and FIGS. 3 and 4 are waveform shaping means according to the prior art. 5 is a block diagram showing the principle of waveform shaping, FIGS. 10 and 11 are block diagrams showing another waveform shaping means according to the prior art, and FIG. 12 is the present invention. FIG. 13 is a block diagram showing a prerequisite technique required for explaining one embodiment of the present invention, FIG. 13 is a block diagram showing a first reference example of the present invention, and FIG. 14 is a second reference example of the present invention. FIG. 15 is a block diagram showing the principle of the present invention, FIG. 16 is a block diagram showing a third reference example of the present invention, and FIG. 17 is a block diagram showing an embodiment of the present invention. is there. 1 ... Character signal generator, 2 ... Waveform shaping filter, 3 ... Binary NRZ signal, 4 ... 60% cosine roll-off spectrum signal, 5 ... Shift register, 6 ... Attenuator, 7 ... … Adder, 8 …… Low pass filter, 9A-9D …… Impulse response waveform, 10 …… 1 clock delay element, 11 …… Read clock, 12 …… Multiplier, 13 …… D / A converter, 14 …… Tap coefficient signal [from left h (-N), h (-N +
1), ..., h (N)], 15 ... Tap output [from the left x (n + N), x (n + N-
1), ..., x (n-N)], 16 ... Digital multiplier, 17 ... Digital adder, 18 ... ROM, 19-23 ... 60% cosine roll-off spectrum signal, 24 ... 1/3 frequency divider counter, 25 ... 1/3 frequency divider pulse, 26 ... Data selector.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takehiko Yoshino 1-10-11 Kinuta, Setagaya-ku, Tokyo Inside the Japan Institute of Transport Technology (56) Reference JP-A-55-46695 (JP, A) Kai 53-145449 (JP, A) JP-A 55-1009025 (JP, A)

Claims (1)

[Claims] 1. A binary NRZ signal x having a period of 1 / fc seconds.
In a digital waveform shaping filter which inputs (t) and outputs a waveform shaped signal y (t), the binary NRZ signal is sequentially inputted in synchronization with a first clock whose frequency is fc hertz, J Books: and (J positive integer) delay means for parallel outputting a delayed bit data S ₁ to S _J from the bit output lines, simultaneously input the delayed bit data S ₁ to S _J as each address, the data The following matrix calculation results using the prescribed impulse response coefficient [A] corresponding to each of S _{1 to} S _J Based on the above, when outputting the waveform shaping data y _{1 to} y _L respectively, y ₁ = A ₁₁ · S ₁ + ... + A _1J · S _J is stored in the first storage means.
, ..., y _L = A _L1 · S ₁ + ... + A _LJ · S _J in the Lth storage means
L (L: positive integer) storing means respectively storing the waveform shaping data y _{1 to} y _L read simultaneously from each of the storing means in synchronization with the first clock, Data selection means for sequentially outputting in synchronization with a second clock having a frequency of L · fc hertz, and processing means for performing low-pass processing after converting the digital output from the data selection means into analog data. The digital waveform shaping filter is characterized by performing waveform shaping based on the principle of superposition of the isolated pulse "1" impulse response signals.