JP6311406B2 - Inverse characteristic filter and radio receiver using the same - Google Patents

Description

  The present invention relates to a filter having an inverse characteristic of a Gaussian filter and a radio receiver using the inverse characteristic filter.

  Digitization of commercial land mobile radio such as fire fighting and emergency radio was promoted, and as one of the standards, ARB (Radio Industry Association) STD-T102 was formulated. This STD-T102 is defined for a radio section interface of a narrowband digital communication system (SCPC (Single Channel Per Carrier) / 4-value FSK system), and the modulation procedure on the radio transmitter side is shown in FIG. As shown in FIG. 13, the demodulation procedure on the radio receiver side is defined as shown in FIG.

  That is, in the case of the wireless transmitter of FIG. 13A, a digital (binary) data string is input to the symbol mapping unit 101, and a symbol value (voltage level) corresponding to a frequency deviation of quaternary FSK every two bits. ) Is converted to a pulse signal (baseband signal). After that, the above-mentioned pulse signal is subjected to band limitation by the two-stage transmission filters 102 and 103, and then given to the frequency modulator 104, and is modulated into a quaternary FSK signal in which the transmission carrier is frequency-shifted, It is given to the transmission amplifier. Table 1 shows the relationship between 2-bit data, symbol values, and frequency deviation.

そして、１段目の送信フィルタ１０２は、下記の数１のＨ（ｆ）で表されるルートナイキストフィルタであり、２段目の送信フィルタ１０３は、下記の数１のＰ（ｆ）で表されるガウスフィルタである。表１のシンボルレベルを有する入力インパルス信号の周波数をｆ、シンボル周期をＴ、ナイキストフィルタのロールオフ率をα、シンボル周期をＴ、帯域幅時間積をＢ_bＴとするとき、それぞれの伝達関数Ｈ（ｆ），Ｐ（ｆ）は、下式で定められる。

The first-stage transmission filter 102 is a root Nyquist filter represented by H (f) in the following equation 1, and the second-stage transmission filter 103 is represented by P (f) in the following equation 1. Gaussian filter. When the frequency of the input impulse signal having the symbol level in Table 1 is f, the symbol period is T, the roll-off rate of the Nyquist filter is α, the symbol period is T, and the bandwidth time product is B _b T, each transfer function H (f) and P (f) are defined by the following equations.

これに対応して、図１３（ｂ）の無線受信機の場合、受信信号を周波数変換した中間周波信号は、周波数検波器１１１に入力されて、前記４値ＦＳＫ信号の周波数偏位に対応したシンボル値を有するパルス信号に復調され、１段目の受信フィルタ１１２に入力される。この受信フィルタ１１２は、上述の送信フィルタ１０２と同一の特性で帯域制限を行うルートナイキストフィルタＨ（ｆ）である。こうして、受信フィルタ１１２と送信フィルタ１０２との２つのルートナイキストフィルタＨ（ｆ）は、ナイキストフィルタとして作用し、隣接するシンボル値に影響を与えずに、帯域制限が行われる。前記受信フィルタ１１２からの出力は、２段目の受信フィルタ１１３に入力される。この受信フィルタ１１３は、送信フィルタ１０３に対応して設けられる逆特性フィルタであって、その伝達関数Ｄ（ｆ）は、下式で定められる。

Correspondingly, in the case of the wireless receiver of FIG. 13B, the intermediate frequency signal obtained by frequency-converting the received signal is input to the frequency detector 111 and corresponds to the frequency deviation of the four-value FSK signal. The signal is demodulated into a pulse signal having a symbol value and input to the reception filter 112 at the first stage. The reception filter 112 is a root Nyquist filter H (f) that performs band limitation with the same characteristics as the transmission filter 102 described above. Thus, the two root Nyquist filters H (f) of the reception filter 112 and the transmission filter 102 act as Nyquist filters, and band limitation is performed without affecting adjacent symbol values. The output from the reception filter 112 is input to the reception filter 113 at the second stage. The reception filter 113 is an inverse characteristic filter provided corresponding to the transmission filter 103, and the transfer function D (f) is determined by the following equation.

こうして２段の受信フィルタ１１２，１１３によって濾波されたパルス信号は、ビット変換部１１４に入力され、そのシンボル値に応じて、前記表１の関係から、２ビットの２値データ列に復号されて出力される。

The pulse signals thus filtered by the two-stage reception filters 112 and 113 are input to the bit conversion unit 114, and are decoded into a 2-bit binary data string from the relationship of Table 1 according to the symbol value. Is output.

ARIB STD-T102

  As shown in FIG. 14, the gain of the Gaussian filter P (f) approaches 0 as the frequency increases, so that the filter D (f) having the inverse characteristic has a frequency as shown in FIG. As it increases, the gain increases exponentially and approaches infinity. Therefore, if the reception filter 113 is simply configured as an inverse characteristic filter of the transmission filter 103, the convergence of the impulse response to 0 becomes very slow. This is shown in FIG.

  FIG. 16 is a waveform diagram simulating impulse responses of the Gaussian filter P (f) and the filter D (f) having the inverse characteristic and denoted by reference numerals A1 and A2, respectively. The same applies to the graphs in the following description, but the input impulse (output pulse from the frequency detector 111) is oversampled and simulated at 16 times the frequency. It can be understood that the impulse response of the Gaussian filter P (f) converges to 0 in a few samples before and after, as indicated by reference symbol A1 in FIG. Therefore, the Gaussian filter can be realized relatively easily with an FIR filter of 64 stages (tap) or less. However, as indicated by reference symbol A2 in FIG. 16, the inverse filter D (f) has a slow convergence of the impulse response, and even a 64-stage (tap) FIR filter leaves a large error, which is practically wireless. It is difficult to mount on the integrated circuit chip of the receiver. Alternatively, if the number of stages (taps) is limited at a stage where the impulse response cannot be converged to 0, an error remains and the signal is distorted.

  An object of the present invention is to provide an inverse characteristic filter of a Gaussian filter capable of suppressing the number of stages (tap) and a radio receiver using the same.

  The inverse characteristic filter of the present invention is an FIR filter having an inverse characteristic of a Gaussian filter, applied in a first band from a lower limit frequency to a pass frequency band, and having a desired inverse characteristic of the Gaussian filter. In the transfer function and the transfer function graph, the upper limit frequency of the pass frequency band is set as a symmetric point, and the point is symmetrical with the first transfer function, and the point symmetric characteristic is twice the upper limit frequency. The second transfer function applied in the second band up to the frequency of the second frequency, and the second transfer function having a line symmetric characteristic with the double frequency as the axis of symmetry, And a third transfer function applied in a third band up to a frequency that is three times the upper limit frequency, and a characteristic that is point-symmetric with the third transfer function with the three times the frequency as a symmetric point. ,That In symmetric properties, and a fourth transfer function to be applied in the fourth band of up to four times the frequency of the upper limit frequency, characterized by.

  Therefore, the first band to the fourth band have a continuous and smooth characteristic, the impulse response of the FIR filter converges faster, and the gain fluctuation in the pass frequency band can be suppressed, and the inverse characteristic filter is This can be realized with an FIR filter having a small number of stages (tap).

  In addition, the inverse characteristic filter of the present invention multiplies the second to fourth transfer functions by a window function to converge the gain to 0 in the fifth band having a frequency higher than the fourth band. It is characterized by.

  According to the above configuration, by multiplying the second to fourth transfer functions by a window function, the gain of the transfer function is smoothly and finally converged to a value of 0, and noise removal capability at high frequencies is achieved. Can be improved.

  Furthermore, in the inverse characteristic filter of the present invention, the input signal is sampled over a predetermined period, a sampling unit for holding the sampling value, and a predetermined coefficient for the sampling value held by the sampling unit, respectively. A multiplication unit that multiplies, and an addition unit that adds each multiplication value obtained by the multiplication unit to obtain a filter output, and values corresponding to the first to fourth transfer functions as coefficients in the multiplication unit Are set respectively.

  According to said structure, the inverse characteristic filter of the said Gaussian filter is realizable by the FIR filter which consists of a multistage (tap) product-sum calculator.

  In the inverse characteristic filter of the present invention, an analog / digital converter is provided on the input side, and a digital / analog converter is provided on the output side. The analog / digital converter has a period sufficiently shorter than the symbol period of the input signal. It is a digital filter that performs oversampling.

  According to the above configuration, the inverse characteristic filter as described above can be realized only by setting the coefficient in the product-sum calculator of the digital filter, and can be realized with a small number of stages (tap). For example, in the case of oversampling at 1/16 of one symbol period, that is, 16 times the sampling frequency, it can be accommodated in about 32 stages (tap).

  Furthermore, a radio receiver according to the present invention includes the above-described inverse characteristic filter.

  According to the above configuration, the inverse characteristic filter for the Gaussian filter as the transmission-side band limiting filter can be implemented by the FIR filter having a small number of stages (tap) in the radio receiver.

  The inverse characteristic filter of the present invention creates a second transfer function by making the first transfer function of the pass frequency band point-symmetric in a frequency band higher than the required pass frequency band, and the first and second By creating the third and fourth transfer functions by making the transfer function line symmetric, the pass gain is smoothly reduced.

  Therefore, the impulse response can be quickly converged to 0, and the inverse characteristic filter can be realized by an FIR filter having a small number of stages (tap).

It is a graph which shows typically the transfer function of the inverse characteristic filter concerning one embodiment of the present invention. It is a graph which shows the transfer function of the inverse characteristic filter which concerns on one Embodiment of this invention. FIG. 15 is a waveform diagram showing impulse responses of the Gaussian filter shown in FIG. 14 and the inverse characteristic filter shown in FIG. 2. It is a graph which shows the passage characteristic at the time of combining the Gaussian filter shown in FIG. 14, and the inverse characteristic filter shown in FIG. It is a graph which shows typically a window function used for a preferred embodiment. It is a graph which shows the window function used for preferable embodiment. 3 is a graph showing a transfer function of an inverse characteristic filter obtained by multiplying the inverse characteristic filter of FIG. 2 by the window function. FIG. 15 is a waveform diagram showing impulse responses of the Gaussian filter shown in FIG. 14 and the inverse characteristic filter shown in FIG. 7. It is a graph which shows the passage characteristic at the time of combining the Gaussian filter shown in FIG. 14, and the inverse characteristic filter shown in FIG. FIG. 15 is a waveform diagram showing impulse responses of the Gaussian filter shown in FIG. 14 and a modified example of the inverse characteristic filter shown in FIG. 7. It is a block diagram which shows the digital filter which is one structural example of the inverse characteristic filter which concerns on one Embodiment of this invention. It is a block diagram which shows an example of a radio receiver provided with the inverse characteristic filter which concerns on one Embodiment of this invention. It is a block diagram which shows the modulation procedure by the side of the radio | wireless transmitter for the band restriction prescribed | regulated by ARIB STD-T102, and the demodulation procedure by the side of a radio | wireless receiver. It is a graph which shows the transfer function of a Gaussian filter. It is a graph which shows the transfer function of the filter of the reverse characteristic of the said Gaussian filter. FIG. 16 is a waveform diagram showing impulse responses of the Gaussian filter shown in FIG. 14 and the inverse characteristic filter shown in FIG. 15.

  FIG. 1 is a graph schematically showing a transfer function of an inverse characteristic filter according to an embodiment of the present invention. This inverse characteristic filter is an FIR filter applied as the second-stage reception filter 113 in the wireless receiver shown in FIG. The transfer function of the inverse characteristic filter of the present invention is shown in FIG. 1B, and FIG. 1A shows the transfer function of the Gaussian filter shown in FIG. The transfer function of a filter is shown typically.

  As described above, the gain of the Gaussian filter approaches 0 as the frequency increases. Therefore, as shown in FIG. 1A, a simple inverse characteristic filter has a gain that increases exponentially with increasing frequency and approaches infinity. For this reason, the convergence of the impulse response is very slow as it is. In FIG. 1, the upper limit frequency of the pass frequency band (the band through which the signal should pass) is β, and the transfer function in that band is D (f).

Therefore, the present inventor has found a transfer function shown in FIG. It should be noted that the first transfer function D ₁ (f) applied in the first band from the lower limit frequency to the upper limit frequency β (the band through which the signal passes) is the desired Gaussian filter transfer function. A transfer function D (f) having P (f) as an inverse characteristic as it is, and a transfer function after the upper limit frequency β (high) is represented by the first transfer function D ₁ (f ) Is symmetrically drawn, and the rate of increase of the pass gain is smoothly reduced.

Specifically, in the graph of the transfer function, in the second band from the upper limit frequency β to the doubled frequency 2β, the upper limit frequency β is a symmetric point V1, and the point is symmetrical with the first transfer function D ₁ (f). The characteristic second transfer function D ₂ (f) is applied. Next, the second transfer function D with the double frequency 2β as the symmetry axis V2 in the third band from the double frequency 2β to the upper limit gain and beyond that band to the triple frequency 3β. _A third transfer function D ₃ (f) having a characteristic symmetrical with ₂ (f) is applied. Further, in the fourth band up to four times the frequency 4β, the third transfer function D ₃ (f) and the fourth transfer function D ₄ (f ) Apply. Further, the fifth transfer function D ₅ (f) in the fifth band having a frequency above the fourth band is set to 1.

The above transfer functions will be described in detail as follows. The transfer function P (f) of the Gaussian filter is
P (f) = exp {-(2ln2) (fT / B _b T) ² }
(F: filter input signal frequency, T: symbol period, B _b T: bandwidth time product)
In this case, the transfer function D (f) = of the inverse characteristic filter is
D (f) = 1 / P (f)
It is. When the lower limit frequency is 0 and the pass frequency band is β, the transfer functions D ₁ (f) to D ₅ (f) are
D ₁ (f) = D (f) (0 ≦ f ≦ β)
D ₂ (f) = 2D (β) −D (2β−f) (β <f ≦ 2β)
D ₃ (f) = 2D (β) −D (f−2β) (2β <f ≦ 3β)
D ₄ (f) = D (4β−f) (3β <f ≦ 4β)
D ₅ (f) = D (0) = 1 (f> 4β)
It becomes.

When the roll-off rate of the Nyquist filter is α and the symbol period is T, the upper limit frequency β of the pass frequency band is
β = (1 + α) / 2T
It becomes.

With this configuration, in the first band, the first band input to the inverse characteristic filter by the inverse transfer function D ₁ (f) of the Gaussian filter transfer function P (f). Can be restored to a shape close to the input signal to the Gaussian filter. Next, in the second band, the gain increase rate is gradually reduced by the second transfer function D ₂ (f) that is point-symmetric with the _first transfer function D ₁ (f). Further, the transfer functions D ₃ (f) and D ₄ (f) in the third band and the fourth band continuously reduce the gain to 1 smoothly, and the transfer function D ₅ in the fifth band. It is maintained at 1 by (f).

  Accordingly, the first band to the fourth band have a continuous and smooth characteristic, the impulse response of the FIR filter converges faster, and the gain fluctuation in the first band up to the upper limit frequency β can be suppressed. The inverse characteristic filter can be realized by an FIR filter having a small number of stages (tap).

  With this configuration, the transfer function of a filter that is simply the inverse characteristic of a Gaussian filter is FIG. 15, whereas in this embodiment, the pass gain is reduced smoothly as shown in FIG. I can go. Accordingly, the impulse response that has not converged for a long time as indicated by reference symbol A2 in FIG. 16 can be quickly converged to 0 as indicated by reference symbol A2 in FIG. In FIG. 3, as in FIG. 16, the impulse responses of the Gaussian filter and the inverse filter are simulated and indicated by reference numerals A1 and A2, respectively. As described above, also in FIG. 3, the input impulse (the output pulse from the frequency detector 111) is oversampled at 16 times the frequency and simulated. Therefore, as is apparent from FIG. 3, the impulse response of the inverse characteristic filter converges to 0 in about 30 samples, and the FIR filter composed of about 30 stages (tap) product-sum operation units performs the Gaussian filter. It is understood that an inverse characteristic filter can be realized. 2 to 3 can be obtained by inverse fast Fourier transform, and FIGS. 3 to 2 can be obtained by fast Fourier transform.

FIG. 4 shows the pass characteristics when the Gaussian filter P (f) on the transmission side and the filters D ₁ (f) to D ₅ (f) having the reverse characteristics of this embodiment are combined. 4A shows gain characteristics, and FIG. 4B shows phase characteristics. As is apparent from FIG. 4, there is almost no gain fluctuation in the pass band shown in FIG. 4A, and the phase characteristic shown in FIG. 4B is also a continuous straight line.

By the way, in the above-described embodiment, as shown in FIG. 2, the fifth transfer function D ₅ (f) is converged to 1. Therefore, since the gain does not approach 0 at a high frequency as it is, the noise removal capability by this inverse characteristic filter is low. Therefore, the second to fourth transfer functions D ₂ (f) to D ₄ (f) are multiplied by the window function, and the gain is converged to 0 in the fifth band having a frequency higher than the fourth band. It is preferable. This is schematically shown in FIG. Thereby, it is possible to improve the noise removal capability at high frequencies.

  Specifically, the window function W (f) uses a cosine curve to converge the gain to 0. As shown in FIG. 5B, the range in which the gain is decreased is defined as γ. In the example of FIG. 5, γ = 3β. In that case, the window function W (f) is as follows.

W (f) = 1 (0 ≦ f ≦ β)
W (f) = [cos {(f−β) π / γ} +1] / 2 (β <f ≦ β + γ)
W (f) = 0 (4β <f)
It is.

The window function W (f) in the above equation is as shown in FIG. A total transfer function obtained by multiplying the window function W (f) by the second to fourth transfer functions D ₂ (f) to D ₄ (f) is as shown in FIG. Due to the difference between the transfer function shown in FIG. 7 and the transfer function shown in FIG. 2, the impulse response corresponding to the reference symbol A2 in FIG. 3 changes as indicated by the reference symbol A2 in FIG. FIG. 9 shows pass characteristics when combined with a Gaussian filter P (f) on the transmission side. FIG. 9A shows gain characteristics, and FIG. 9B shows phase characteristics. As is clear from the reference symbol A2 in FIG. 8, there is almost no influence as the impulse response converges slightly to zero. Further, from FIG. 9, the gain characteristic and the phase characteristic have almost no influence.

  The range γ is a range of 3β or less, and as the range γ is set larger, the gain in the range exceeding the first band as the pass band is gradually reduced to zero. On the other hand, as the range γ is set smaller and the gain starts to drop faster, the noise removal capability in the high band can be improved. However, it takes time (vibrates) to converge the impulse response. Therefore, in the above example, γ = 3β, but a lower value such as γ = 2β or γ = 1.5β may be used. FIG. 10 shows an impulse response when γ = β. As shown in FIG. 10, it takes time (vibrates) to converge the response.

  FIG. 11 is a block diagram showing a digital filter 120, which is a configuration example of the inverse characteristic filter according to the embodiment of the present invention. The digital filter 120 is an FIR filter realized by DSP software. As described above, data oversampled at a frequency 16 times the symbol frequency is input from the frequency detector 111 to the input terminal IN, and sequentially transferred to and stored in 31 cascaded buffers B1 to B31. It will be done. Therefore, the buffers B1 to B31 constitute a sampling unit, and can hold sampling values over approximately two symbol periods.

The data of the input terminal IN and the data of the buffers B1 to B31 are supplied to the corresponding multipliers X0 to X31, and multiplied by the coefficients set in the corresponding coefficient units C0 to C31. The outputs of the multipliers X0 to X31 are sequentially added by the adders Z1 to Z31, and output from the output terminal OUT as a filter output. Thus, the digital filter 120 can be realized by a 32-stage (tap) FIR filter. Each coefficient unit C0 to C31 receives the inverse Fourier transform of the value of the data point in the impulse response indicated by the reference A2 in FIG. 3, that is, the transfer functions D ₁ (f) to D ₅ (f). The value of the impulse response is set as a coefficient. For example, the reference symbol A2 in FIG. 3 and the reference symbol A2 in FIG. 8 are coefficients.

  In this manner, the inverse characteristic filter 113 of the Gaussian filter can be realized by the digital filter 120 including a multi-stage (tap) product-sum calculator. Then, by using such an inverse characteristic filter 113 as an inverse characteristic filter for the Gaussian filter 103 as a transmission-side band limiting filter shown in FIG. 13A in the wireless receiver shown in FIG. 13B, The inverse characteristic filter can be implemented by an FIR filter having a small number of stages (tap). In addition, the buffer B1-B31 is a ring buffer, so that the load of the transfer process can be reduced.

  FIG. 12 is a block diagram illustrating an example of a wireless receiver including the inverse characteristic filter 113 as described above. This radio receiver 1 is configured by a double superheterodyne system, and the signal received by the antenna 3 is filtered by the FSK high-frequency signal component of, for example, 440 MHz through the band-pass filter 4 and amplified by the amplifier 5. Then, it is input to the mixer 6 at the first stage. In the mixer 6, for example, an intermediate frequency signal (first intermediate frequency signal) of 46.35 MHz obtained by mixing with an oscillation signal of 486.35 MHz, for example, from the local oscillator 7 is obtained by the bandpass filter 8. The intermediate frequency component is filtered and amplified by the amplifier 9 and then input to the second-stage mixer 10.

  In the mixer 10, the intermediate frequency signal is mixed with an oscillation signal of 45.9 MHz, for example, from the local oscillator 11, and the obtained intermediate frequency signal (second intermediate frequency signal) of 450 kHz, for example, is bandpass. The intermediate frequency component is filtered by the filter 12, amplified by an amplifier (intermediate frequency amplifier) 13, and then input to the analog / digital converter 14. In the analog / digital converter 14, the input signal is converted into a digital signal having a rate of 38.4 ksps (sample per second), for example, and input to the demodulation circuit 21.

  The demodulating circuit 21 has the same configuration as that shown in FIG. 13B, and is configured by a DSP (digital signal processor) that incorporates the digital filter 120 (implemented by software). Then, the audio signal is demodulated, converted into an analog signal by the digital / analog converter 15, and sonicated from the speaker 16. By using such a demodulating circuit 21, the radio receiver 1 including the inverse characteristic filter of the present invention can be realized.

DESCRIPTION OF SYMBOLS 1 Wireless receiver 3 Antenna 14 Analog / digital converter 15 Digital / analog converter 16 Speaker 21 Demodulation circuit (DSP)
101 Symbol mapping unit 102 Transmission filter (Root Nyquist filter)
103 Transmission filter (Gaussian filter)
104 Frequency modulator 111 Frequency detector 112 Reception filter (Root Nyquist filter)
113 Reception filter (reverse characteristic filter)
114 bit conversion unit 120 digital filter B1 to B31 buffer X0 to X31 multiplier C0 to C31 coefficient unit Z1 to Z31 adder

Claims (5)

In an FIR filter having the inverse characteristics of a Gaussian filter,
A first transfer function applied in a first band from a lower limit frequency to a pass frequency band and having an inverse characteristic of the desired Gaussian filter;
In the graph of the transfer function, the upper limit frequency of the pass frequency band is a symmetric point and has a point symmetric characteristic with the first transfer function, and the point symmetric characteristic has a frequency up to twice the upper limit frequency. A second transfer function applied in the second band;
It has a characteristic that is line-symmetric with the second transfer function with the double frequency as the axis of symmetry, and is applied in the third band up to three times the upper limit frequency with the line-symmetric characteristic. A third transfer function;
It has a characteristic that is point-symmetric with the third transfer function with the triple frequency as a symmetric point, and is applied in the fourth band up to four times the upper limit frequency with the point-symmetric characteristic. Having a fourth transfer function;
An inverse characteristic filter characterized by   The inverse of claim 1, wherein the second to fourth transfer functions are multiplied by a window function so that the gain is converged to 0 in a fifth band having a frequency higher than that of the fourth band. Characteristic filter. A sampling unit that samples the input signal over a predetermined period and holds the sampling value;
A multiplier for multiplying the sampling value held in the sampling unit by a predetermined coefficient,
An addition unit that adds each multiplication value obtained by the multiplication unit to obtain a filter output;
3. The inverse characteristic filter according to claim 1, wherein values corresponding to the first to fourth transfer functions are respectively set as coefficients in the multiplication unit. Equipped with an analog / digital converter on the input side, a digital / analog converter on the output side,
4. The inverse characteristic filter according to claim 3, wherein the analog / digital converter is a digital filter that performs oversampling with a period sufficiently shorter than a symbol period of an input signal.   A radio receiver comprising the inverse characteristic filter according to any one of claims 1 to 4.

Images