WO2015136929A1

WO2015136929A1 - Frequency signal generator

Info

Publication number: WO2015136929A1
Application number: PCT/JP2015/001337
Authority: WO
Inventors: 小林　薫
Original assignee: 日本電波工業株式会社
Priority date: 2014-03-12
Filing date: 2015-03-11
Publication date: 2015-09-17
Also published as: JP6463336B2; JPWO2015136929A1

Abstract

[Problem] To reduce spurious output of a DDS. [Solution] Providing an addition unit (5) for adding random noise to a digital signal outputted from a waveform data memory (2). A circuit for generating PN sequential data is used in a random noise generation unit (3), which is provided with a circuit for performing a data process so that digital values that constitute the noise are set to "-1" and "1." This data process prevents an offset (0.5) from being generated in the output value of the noise. A filter (4) is composed, for example, of a plurality of serially connected moving-averaging circuits, the filter (4) performs a process so that the signal level of the frequency spectrum of the random noise is biased toward a frequency band that deviates from the DDS setting frequency. The random noise outputted from the filter (4) is inputted to the addition unit (5).

Description

Frequency signal generator

The present invention relates to the technical field of an apparatus for generating an analog waveform based on waveform data consisting of digital values.

A DDS (Direct Digital Synthesizer) includes a phase accumulator, a waveform data memory, and a D / A (Digital / Analog) converter, and obtains an analog waveform of an arbitrary frequency. Specifically, an address is output from the phase accumulator based on the frequency setting data, and the waveform data (amplitude data) in the waveform data memory is read out according to the address and consists of a digital signal (digital value group), for example. A sine wave is obtained and converted to an analog signal.
By the way, spurious generation is inevitable in DDS. Quantization noise can be cited as the cause, and a periodic error generated from the relationship between the clock frequency and the DDS output frequency is also caused. In the D / A converter, the time interval of the output timing of the analog signal corresponding to the input of each digital signal and the magnitude of the analog value are not strictly constant, which is also a factor of spurious. .

For this reason, conventionally, as described in Patent Document 1, for example, a technique is known in which pseudo-random noise (noise) is added to waveform data read from the waveform data memory to diffuse the spurious. However, this method has a problem that the level of floor noise in the vicinity of the output frequency of the DDS is increased, the S / N ratio is deteriorated, and the effect of improving the spurious is low.

JP-A-8-97744 (paragraphs 0192 to 0195)

The present invention has been made under such circumstances, and an object thereof is to reduce spurious in a frequency signal generator that generates an analog waveform based on waveform data composed of digital value groups.

The frequency signal generator of the present invention includes a waveform output unit that outputs a periodic waveform signal composed of a digital signal,
A noise generator that generates random noise composed of digital signals;
A filter that performs processing for biasing the signal level of the frequency spectrum in the random noise to a frequency band that deviates from the frequency of the waveform signal, with respect to the output signal output from the noise generation unit,
An adder for adding the output signal of the waveform output unit and the output signal of the filter;
And a conversion unit for converting the digital signal obtained by the addition unit into an analog signal.

In the present invention, random noise consisting of a digital signal is added to a periodic waveform signal consisting of a digital signal output from the waveform output unit, so that spurious appearing in the analog output is diffused. The random noise is processed by a filter so that the signal level of the frequency spectrum is biased to a frequency band that deviates from the frequency of the waveform signal. Therefore, the level of floor noise in the vicinity of the target output frequency (set frequency) can be suppressed low, and a high spurious suppression effect can be obtained.

1 is a circuit diagram showing a first embodiment of a frequency signal generator of the present invention. It is a circuit diagram which shows a part of circuit of a random noise production | generation part. It is a circuit diagram which shows the comparative example of a frequency signal generator. It is the frequency spectrum of the output of DDS when random noise is not superimposed. It is the figure which expanded a part of frequency spectrum shown in FIG. It is the frequency spectrum of the random noise output from the random noise generation part. It is a frequency spectrum of the output of DDS when a random noise is superimposed using the circuit of the comparative example shown in FIG. It is the figure which expanded a part of frequency spectrum shown in FIG. It is the frequency spectrum of the random noise after processing the random noise shown in FIG. 6 with a filter. It is a frequency spectrum of the output of the low-pass filter of DDS when random noise is superimposed using the circuit of the embodiment shown in FIG. It is the figure which expanded a part of frequency spectrum shown in FIG. . It is a circuit which shows the specific example of the filter used for this invention. It is a circuit which shows an example of the circuit which performs the moving average of the filter shown in FIG. It is a circuit diagram which shows the 2nd Embodiment of this invention. It is a time chart of the band movement process part shown in FIG. It is a frequency spectrum after performing a band shift process with respect to the random noise after processing by the filter. 15 is a frequency spectrum of an output of a low-pass filter of a DDS when random noise is superimposed using the circuit of the embodiment shown in FIG. It is the figure which expanded a part of frequency spectrum shown in FIG. . It is a circuit diagram which shows the modification of the 2nd Embodiment of this invention.

[First embodiment]
FIG. 1 shows an embodiment in which the frequency signal generator of the present invention is applied to a DDS. Reference numeral 1 is a phase accumulator, and 2 is a waveform data memory. The phase accumulator 1 includes an adder 11 and a latch circuit 12. The adder 11 receives frequency setting data for setting a target frequency for the frequency signal output from the DDS. If the frequency setting data is N, for example, N is output from the latch circuit 12 by the first clock signal, and 2N, 3N, 4N... Are sequentially output from the latch circuit 12 by the subsequent clock signal.

The waveform data memory 2 stores, for example, waveform data (a digital value group corresponding to the amplitude value of the waveform) for generating a sine wave, and the waveform data is read using the output signal of the latch circuit 12 as an address. Accordingly, the frequency of the frequency signal specified by the waveform data is determined according to the value of N.
The phase accumulator 1 and the waveform data memory 2 correspond to a waveform output unit that outputs a periodic waveform signal.

3 is a random noise generation unit, which is constituted by, for example, a pseudo random noise generation circuit. In the present application, random noise includes pseudo-random noise. Examples of the pseudo random noise generation circuit include a circuit that outputs a PN (Pseudo Noise) series signal in synchronization with a clock signal input to the DDS, for example, 200 MHz. For example, a shift register and Ex-OR (exclusive logic) And a circuit that generates a Gold sequence using an M sequence of a prefeed pair. In this example, the logic “0” is converted to “1” and the logic “1” is changed to “−1” with respect to the time series signal including the logic “1” and the logic “0” which are PN sequence signals. A logic value processing circuit for converting to a random noise generation unit 3 is provided. Therefore, the random noise generation unit 3 includes a PN sequence generation circuit and a signal processing circuit provided on the output side of the PN sequence generation circuit. FIG. 2 shows an example of the signal processing circuit.

The reason why the logic “0” and “1”, which are PN series signals, are converted to “1” and “−1”, respectively, is to the filter 4 described later in the case of logic “0” and “1” data. This is because an offset of 0.5 occurs in the input value to be input. When an offset is generated in this way, there is a problem that a DC component corresponding to the offset is also superimposed on the output of the frequency signal generator. Note that the logic “1” which is a PN sequence signal may be output as it is, and the process of converting the logic “0” into “−1” may be performed.
FIG. 2 shows an example of the signal processing circuit described above. In FIG. 2A, the signal processing circuit 31 performs a process of adding a 1-bit logic “1” to the 1-bit input signal and outputs a 2-bit signal represented by 2's complement. To do. In FIG. 2B, the signal processing circuit 32 is composed of a multiplexer, and 2-bit signals “01” and “11” expressed by two's complement numbers are input to one input end and the other input end of the multiplexer, respectively. Yes. The multiplexer receives an input signal “11” at the other input terminal when “0”, which is a 1-bit signal, is input as a selection signal, and inputs one when “1” is input as a selection signal. The input signal “01” at the end is output.

A noise processing filter 4 is provided on the output side of the random noise generator 3. The filter 4 is configured to perform processing for biasing the signal level (power intensity) of the frequency spectrum of random noise output from the random noise generating unit 3 to a frequency band that deviates from the set frequency (output frequency) of the DDS. Has been. A specific configuration example of the filter 4 will be described later.

In FIG. 1, reference numeral 5 denotes an adder. The adder 5 is a digital signal (amplitude value data) read from the waveform data memory 2, for example, a 32-bit digital signal, and a 32-bit output from the filter 4. The digital signal (output signal) is added. On the output side of the adder 5, a D / A converter 6 that converts the added value, which is the output value of the adder 5, into an analog signal is provided. On the output side of the D / A converter 6, a low-pass filter 7 is provided for removing harmonics contained in the analog waveform signal obtained from the D / A converter 6 to obtain a waveform with less distortion. Yes.

Furthermore, a high pass filter 8 is provided on the output side of the low pass filter 7. The high-pass filter 8 is for removing noise existing in a frequency band away from the set frequency based on the filtering process performed by the filter 4.

In order to facilitate understanding of the operation of the present invention, when random noise is not used, random noise is used, but when the filter 4 is not provided (see FIG. 3), and when the circuit shown in FIG. The operation will be described in order.

4 is obtained from the low-pass filter 7 when random noise is not added to the digital signal read from the waveform data memory 2 (a circuit obtained by removing the random noise generator 3 and the adder 5 from the circuit of FIG. 3). This is the frequency spectrum of the frequency signal (DDS output signal). The clock signal is 200 MHz. In this example, the set frequency of the DDS output signal is 42 MHz. FIG. 5 is an enlarged view of the vicinity of the set frequency in the frequency spectrum shown in FIG. In the series of frequency spectrum diagrams including FIG. 4, the signal level corresponding to the set frequency is 0 dB. As can be seen from FIGS. 4 and 5, spurious appears in the vicinity of the set frequency.

Next, the frequency spectrum when random noise is added to the digital signal read from the waveform data memory 2 will be described. FIG. 6 is a frequency spectrum of random noise generated by the random noise generator 3. FIG. 7 is a frequency spectrum of a frequency signal (DDS output signal) obtained from the low-pass filter 7 using the circuit of FIG. FIG. 8 is an enlarged view of the vicinity of the set frequency (42 MHz) in the frequency spectrum shown in FIG. As can be seen from FIGS. 7 and 8, the floor noise level is high (floor noise is degraded) and the spurious improvement effect is low when only random noise is used.

Subsequently, a frequency spectrum when the circuit shown in FIG. 1 which is an embodiment of the present invention is used will be described. In this example, a circuit in which a random noise generating unit 3, a filter 4, and an adding unit 5 are added to the circuit when the frequency spectrum of FIG. 4 is collected is used. The frequency spectrum of the random noise output from the random noise generator 3 is the same as that shown in FIG. FIG. 9 is a frequency spectrum of random noise output from the filter 4. In this example, the filter 4 performs processing for biasing the signal level of the frequency spectrum shown in FIG. 3 to a frequency band lower than 42 MHz. FIG. 10 shows the frequency spectrum of the frequency signal obtained from the low-pass filter 7 of the circuit shown in FIG. FIG. 11 is an enlarged view of the vicinity of the set frequency (42 MHz) in the frequency spectrum shown in FIG.

As can be seen by comparing FIG. 4 with FIG. 10 and FIG. 11, according to the circuit of FIG. 1, the spurious is improved by about 10 to 15 dB, and the floor noise near the set frequency of DDS is also reduced. Has been. The floor noise that is increased in the low frequency band by biasing the frequency spectrum of the random noise by the filter 4 is far away from the set frequency, so that a simple filter such as an LC that combines an inductor and a capacitor is used. It can be removed by a filter or the like. In FIG. 1, as such a filter, a high-pass filter is provided on the output side of the low-pass filter 7.

Here, an example of a specific circuit of the filter 4 is shown in FIGS. In this example, the filter 4 is configured by connecting 10 stages of moving average circuits in series as shown in FIG. The moving average circuit in each stage is 9-n (n is an integer from 1 to 10), and averages a group of digital values consisting of 8 consecutive digital values as shown in FIG. The first digital value of the digital value group is shifted to the next digital value. For example, when digital values are sequentially input as d0, d1, d2,... D7, d8, d9, d10... By a clock signal, a digital value group consisting of eight digital values from d0 to d7 by a certain clock signal. Is averaged for the digital value group from d1 to d8 by the next clock signal, and further averaged for the digital value group from d2 to d9 by the next clock signal.

FIG. 13 shows a first-stage moving average circuit, to which a 2-bit digital value output from the random noise generator 3 is input. d0 to d7 indicate digital values output from the random noise generating unit 3 by the clock signal. D is a delay circuit that delays and outputs an input value for each clock signal. The digital signal input to the first-stage delay circuit D includes seven clock signals input to the moving averaging circuit. Sometimes, it is output from the delay circuit D at the final stage. A is an adder circuit.

In a general moving average circuit, a circuit that multiplies a digital value by 1/8, that is, a rounding processing circuit that performs rounding processing to divide the digital value by “8” is provided after the final stage adder A. Yes. By dividing eight continuous time series data by “8”, an average value of the eight digital values is obtained. On the other hand, the moving average circuit as a filter used in the present embodiment is for biasing the signal level of the frequency spectrum of random noise to a low frequency band. The target signal can be obtained without dividing the added value of the obtained eight data by “8”. Then, by dividing the added value of the eight data by “8”, the signal level as noise can be made a necessary level. In the specification and claims of this application, “moving averaging circuit” Includes a circuit that does not perform rounding processing, and also includes a circuit that performs rounding processing. When rounding processing is not performed (when the added value of 8 data is not divided by “8”), the digital value is increased by 3 bits in the moving averaging circuit. Are input, and a 5-bit digital value is output.

The 10-stage moving averaging circuit 9-n shown in FIG. 12 uses the same circuit as that shown in FIG. 13 except for the number of bits of the digital value, and the number of bits of the digital value is 1. Since each stage increases by 3 bits, a 32-bit digital value is output from the 10th stage moving average circuit 9-10. 12 and 13, the numbers in parentheses represent the number of bits.

The number of samples subject to moving average in the moving average circuit is not limited to 8, and the number of stages in the moving average circuit is not limited to 10. The filter 4 used when obtaining the frequency spectrum of FIG. 9 is configured by connecting a moving average circuit that does not perform rounding processing in a plurality of stages in series. The number of connection stages of the averaging circuit is different from the examples of FIGS.

The frequency spectrum of the noise generated by the random noise generating unit 3 is distributed over a wide band, and therefore noise is generated even in a region where the set frequency of the DDS exists, and the floor noise is deteriorated. The role of the filter 4 is to reduce the energy of noise in a region where the set frequency of the DDS exists and in the vicinity of the frequency spectrum of noise distributed over a wide band. Therefore, the filter 4 may be an FIR (Finite impulse response) filter or an IIR (Infinite Impulse Response) filter.

When an FIR filter or a moving averaging circuit is used, as the number of data values of one group to be averaged increases, the low-pass filter characteristic is biased toward a lower frequency. The number of data values may be selected. When using an FIR filter, the frequency characteristic may be adjusted by adjusting the weighting coefficient multiplied by each tap (each stage).

When the entire operation of the DDS shown in FIG. 1 is summarized, an address is output from the phase accumulator 1 according to the frequency setting data, and waveform data (waveform data such as sine wave data) is output from the waveform data memory 2 by this address. A digital value group corresponding to the amplitude value) is output. On the other hand, a PN sequence signal that is pseudo-random noise is output from the random noise generation unit 3, and the signal processing unit included in the random noise generation unit 3 converts the logic “0” of the signal into “1”. Logic “1” is converted to “−1”. The filter 4 performs processing for biasing the signal level (power intensity) of the frequency spectrum of random noise to a frequency band that deviates from the set frequency of the DDS, and the processed random noise and the waveform data are added. . Next, the addition value is converted into an analog signal by the D / A converter 6 and the analog signal is passed through the low-pass filter 7 and the high-pass filter 8 to obtain a frequency signal that is an output of the DDS.

According to the above-described embodiment, random noise is added to the digital signal output from the waveform data memory 2. The random noise is processed by the filter 4 so that the signal level of the frequency spectrum is biased to a frequency band that deviates from the set frequency of the DDS. Accordingly, the floor noise level in the vicinity of the set frequency (output frequency) in the DDS can be suppressed to a low level, and a high spurious suppression effect can be obtained.
[Second Embodiment]
The DDS set frequency in the first embodiment is 42 MHz. Even when the set frequency is set to 21 MHz, for example, if the filter 4 used to obtain the frequency spectrum of FIG. There is concern that the level of floor noise will be high. Therefore, in the second embodiment, a process is performed in which a band that is biased in the frequency spectrum of the noise signal output from the filter 4 is moved to a band that is higher than the set frequency and sufficiently far away.

FIG. 14 shows a circuit according to the second embodiment, and a band shift processing unit 40 is provided on the output side of the filter 4. The band shift processing unit 40 includes a multiplexer 402 including one input terminal to which input data is input and the other input terminal to which input data is input through a multiplication unit 401 that multiplies by −1. A value input to one input terminal and a value input to the other input terminal are alternately output.
FIG. 15 is a time chart of the band shift processing unit 40, where d1, d2,... That are input data (output data of the filter 4) are multiplied by -1, and in other words, every other. The polarity of the digital value is reversed. In other words, a frequency signal having “−1” and “1” amplitudes at a frequency ½ of the frequency fs of the clock signal is mixed (mixed) with the input data. Since the frequency of the clock signal is, for example, 200 MHz, to put it roughly, a portion biased by the filter 4 in the noise frequency spectrum moves to 100 MHz.

FIG. 16 is a frequency spectrum of random noise output from the band shift processing unit 40. The frequency spectrum of the noise output from the filter 4 before being input to the band shift processing unit 40 is the same as that shown in FIG. FIG. 17 shows the frequency spectrum of the frequency signal obtained from the low-pass filter 7 of the circuit shown in FIG. FIG. 18 is an enlarged view of the vicinity of the set frequency (21 MHz) in the frequency spectrum shown in FIG. As can be seen from this result, even if there is a concern that the floor noise in the vicinity of the set frequency of the DDS is increased only by using the filter 4, such a concern is caused by using the band shift processing unit 40. Wiped out.

By the way, if the size of the filter 4 is increased, the noise band can be distributed only in a limited region, for example, 10 MHz. However, if the band shift processing unit 40 is used, the configuration of the filter 4 is large. There is an advantage that a simple configuration is sufficient.

For example, based on the timing of a clock signal of 200 MHz, if a data string in which the arrangement of “0”, “1”, “0”, “−1” is repeated and mixed with the data of the filter 4, The part where the frequency spectrum is biased moves to 50 MHz. The frequency signal to be mixed with the output of the filter 4 may be set arbitrarily according to the set frequency of the DDS shown in FIG. 1 by using another DDS for mixing processing. Furthermore, as the filter 4, in the frequency spectrum of random noise, energy is biased to a band higher than the set frequency (biased to a lower band in FIG. 9), and energy is biased (energy is other than that). For example, the band shift processing unit 40 may shift the band of the portion that is higher than the band to a lower level.

By combining the filter 4 and the band shift processing unit 40 in this way, it is possible to obtain random noise having an appropriate frequency spectrum according to the set frequency of the DDS and to superimpose it on the waveform signal.
Furthermore, as shown in FIG. 19, a switching unit 50 is provided, and a mode in which the output of the filter 4 is input to the adding unit 5 through the band shift processing unit 40 and an input to the adding unit 5 without passing through the band shift processing unit 40. The mode to be performed may be selectable. According to such a configuration, spurious can be reduced with respect to a wider setting frequency of DDS.

Claims

A waveform output unit for outputting a periodic waveform signal composed of a digital signal;
A noise generator that generates random noise composed of digital signals;
A filter that performs processing for biasing the signal level of the frequency spectrum in the random noise to a frequency band that deviates from the frequency of the waveform signal, with respect to the output signal output from the noise generation unit,
An adder for adding the output signal of the waveform output unit and the output signal of the filter;
A frequency signal generator comprising: a converter for converting the digital signal obtained by the adder into an analog signal.
The frequency signal generator according to claim 1, wherein the filter is a circuit unit that performs a moving average process on the digital value output from the noise generator.
In order to move the frequency band in the output of the filter between the filter and the adding unit in order to move the frequency band in which the signal level is biased to a position further away from the frequency of the waveform signal. The frequency signal generator according to claim 1, wherein a band shift processing unit is interposed.
The frequency signal generator according to claim 3, wherein the band shift processing unit is configured to mix a frequency signal for moving a frequency band to the filter.
Switchable between a signal path for inputting the output of the filter to the adder through the band shift processor and a signal path for inputting the output of the filter to the adder without passing through the band shift processor The frequency signal generator according to claim 3, further comprising a switching unit.