JPS59114674A - Discrete fourier conversion analyzer - Google Patents

Abstract

PURPOSE:To follow an external sampling clock frequency with high precision to the frequency of an analog input signal by always detecting the ratio between high and low spectra for an observed spectrum. CONSTITUTION:A relative shift is detected by the calculation of a control circuit 18 between the observed frequency in an input signal and the sampling frequency. This calculated value is latched by a latch circuit 25, and the output of the circuit 25 is converted into analog signals by a D/A converter 26. Then the bias voltage supplied from a DC power supply 27 is added to the analog signal when necessary by an adder circuit 28. The output of this addition is applied to a clock signal generator 17 as a signal that controls the clock frequency of the generator 17. As a result, the sample clock frequency of an external sample clock input terminal 16 is corrected to obtain a specific relation with the frequency of the observed spectrum of the input signal of a system 11 to be measured.

Description

[Detailed Description of the Invention] This invention relates to a discrete Fourier transform analyzer that digitally analyzes an input signal frequency, and in particular, uses a frequency that has a specific relationship with the human input signal as a reference, and analyzes its harmonics or subharmonics. So-called order analysis (order ratio analysis)
This paper relates to a discrete Fourier transform analyzer that performs.

<Prior Art> Frequency analysis is performed in which a frequency spectrum, which is a frequency component of an input signal, is digitally determined by performing discrete Fourier transform, for example, fast Fourier transform, on the input signal. In this case, in order to correctly analyze up to high frequencies, it is necessary to make the sampling frequency sufficiently high. However, increasing the sampling frequency increases the scale of arithmetic processing and increases processing time.

From this point of view, for example, when analyzing turbine vibration and measuring the relationship between the generated vibration and the rotational frequency of the turbine, if you measure the higher-order frequency or lower-order frequency with respect to the rotational frequency of the turbine, the rotation It is very convenient to be able to analyze well the relationship between vibration and vibration, and for this reason order ratio analysis (order analysis) is carried out. Conventionally, this type of order ratio analysis has been carried out as shown in FIG. 1, for example. That is, the system to be measured 11 is, for example, a turbine, the vibrations occurring in the system to be measured 11 are detected by a vibration detector, and the detected vibration signal is sent to a discrete Fourier analyzer, for example, a fast Fourier analyzer 12. is supplied to the signal input terminal 13 of.

- The reference in the strain relief measurement system 11, in this example the rotational speed of the turbine, is detected by a rotation detector that generates one pulse per rotation, for example, and this detected clock pulse' is transmitted from the terminal 14 to the rotation speed. It is supplied to a frequency converter 15. The frequency converter 15 doubles the frequency of the input clock pulse, for example, and is a frequency multiplier. This is generally a so-called phase-locked loop (P.

L, L) is used, and a frequency that is multiplied by the input signal frequency (K is not necessarily an integer) is output.

The output signal of the frequency converter 15 is sent to the fast Fourier transformer 12
is applied to the external sample clock input terminal 16 of . Therefore, if K is selected to be an integer, in this example, the vibration signal from the turbine thread 11 input to the signal input terminal 13 is sampled at a sampling frequency that is an integral multiple of the rotational frequency of the turbine, and is digitalized by fast Fourier transform. frequency analysis is performed. In general, a vibration system often generates vibrations that are higher harmonics or subharmonics relative to the fundamental frequency of the vibration, that is, the turbine rotation frequency. Therefore, by sampling the input signal at a frequency twice as high as the frequency of the basic clock of the system under test 110, it is possible to perform accurate analysis with a relatively small number of samples.

In this way, conventional order hole analysis uses a PLL as the frequency converter 15, and the PLL detects deviations from the input signal by phase detection, so it can track changes in the input signal frequency of an octave or more. I can't let it happen. Therefore, if the input clock signal frequency f changes significantly, it may be necessary to prepare a different PLL depending on the frequency range of the input clock and use it by switching. If these conditions are not satisfied, the lock state will be lost and the device will no longer function as a frequency converter.

Furthermore, conventionally, it has been necessary to extract a signal that is considered to be related to the signal to be analyzed from the system under test 11 as a clock signal separately from the input signal. It was necessary to provide means for detecting frequency in addition to means for detecting vibration. Moreover, there is a drawback that the circuit configuration of the frequency converter 150 is complicated as a whole.

<Summary of the Invention> An object of the present invention is to provide a discrete Fourier transform analyzer that has a relatively simple configuration and enables order hole analysis without requiring special means for detecting a clock signal. It is in.

According to this invention, a discrete Fourier transform analyzer scatters spectra higher and lower by a certain frequency with respect to the observed spectrum in the analysis spectrum, and detects the level ratio of these higher and lower spectra. From the level ratio, it is detected that the ratio between the frequency of the observed spectrum and the frequency of the sampling clock deviates from a predetermined value, and the sampling clock frequency is corrected in accordance with this detection. In this way, the sampling clock frequency can always be made to follow the frequency of a specific component of the system under test by using the observed spectrum obtained from the analysis spectrum itself.

<Embodiment> Next, an embodiment of the discrete Fourier transform analyzer according to the present invention will be described with reference to FIG. 2 and the following drawings. In FIG. 2, for example, vibration to be measured from a system under test 11 is detected by a vibration meter such as an acceleration detector, and this analog signal to be measured is sent to a fast Fourier transform analyzer 12, which is a type of discrete Fourier transformer. The signal is supplied to the signal input terminal 13. A sampling clock from a clock signal generator 17 is applied to an external sample clock input terminal 16 of the fast Fourier transform analyzer 12, and the analog signal at the signal input terminal 13 is sampled by this sampling clock and processed by fast Fourier transform. Frequency analysis is performed digitally.

In this invention, we have an observed spectrum at angular frequency ω in the frequency spectrum obtained by fast Fourier transform, a spectrum with +□ at angular frequency ω that is higher by a certain frequency, and a spectrum with -1 at lower angular frequency ω. A deviation of the ratio between the external sampling clock frequency f8 and the observed spectrum frequency ω by a predetermined value is detected from the levels of these observed spectra, high and low spectra. This detection is performed by the control circuit 18.

In this way, the present invention extracts three spectra: the observed spectrum, higher frequencies, and lower frequencies. Generally, in such a fast Fourier transform, an input signal is sampled at a fixed period, and the samples are cut out and analyzed at fixed time intervals, and a cutting error occurs due to this cutting. In order to reduce this error, frequency analysis is performed by multiplying the sample data of the input analog signal by a window function, for example, a vanning window function (1+singing ω+1). Here, ω1 is 10, T is the frame time, ωl is the lowest angular frequency obtained by fast Fourier transform analysis, and ω=×ωl.

When performing frequency analysis by multiplying the vanning window function in this way, one pure spectrum ω has a high spectrum ωl (+1) and a low spectrum ω (−1) on both sides, as shown in Figure 3A. It is observed separately into three spectra such as 1.The input analog signal is expressed as vl(t
)=sIIIωkt, then multiplying this by the non-nning window function (1+houseω11) yields +Fω−−1t. The observed spectrum ωk is obtained by multiplying the window function conventionally used in fast Fourier transform.
A higher spectrum ωl (+1') and a lower spectrum ω (-+) are obtained.By multiplying the analog signal by a window function, etc., that is, by modulating the required three The three spectra obtained in this way have a higher spectrum ωl(
+4 and −1 levels in the lower spectrum ω are respectively i
It becomes.

Now, assuming that the external sampling clock signal frequency f8 is constant, the angular frequency of the input analog signal v1(t) is −
When changing from ωl to +ωl, that is, v 1(t):
5L11 (10 dω in ω), spectrum ωk at -ω1≦dω≦ωl.

+3 to ω. Each level of −+ for ω and ω is represented by curve 21 in Figure 4.
, 22° and 23, respectively. This horizontal axis is d
ω/ωl. In the state where dω is 01, that is, the sampling clock frequency f8 is a predetermined value for ωk, ω
The spectrum of is 1 as shown in curve 21, and the maximum is ! The level of this spectrum decreases whether 1lldω increases or decreases. On the other hand, +1 for a high spectrum ω means that dω is O as shown in the curve 23, and 1 as described above, the higher the dω, the higher the level, and the lower the dω, the lower the level. Conversely, when the spectrum ω is -1 at a low level, as shown by a curve 23, dω is 1 when dω is O, and when it becomes lower than this, the level increases, and when it becomes higher, the level decreases. Therefore, by analyzing the level relationship between these three spectra, it is possible to detect a deviation in the relationship between the external clock and the frequency of the input signal.

For example, as shown in Figure 3B, in the lower spectrum ω -
If the level of + is larger than that of +1 in the higher spectrum ω, it is a case where the input signal frequency is shifted to a relatively lower side with respect to the sampling clock frequency.
These three spectra ωk.

+1 to ω. Each level of −1 in ω is A+ r A
2 + Aa, the song in Figure 4 has 21,
The levels at 22 and 23 are respectively As at the same time,
The frequency positions at which A2 and A3 occur are the positions indicated by the dotted line 24, and from this it is possible to know the frequency deviation dω.

Conversely, when the frequency of the input signal v 1 (t) shifts to a higher side, there are three spectra ω, and ω +1. Each level of -1 in ω is +1 in the higher spectrum ω, as shown in Figure 3C.
The level of ω increases, and the level of □ in the lower spectrum ω decreases. The deviation can be detected from these relationships and FIG. 4.

In this way, the relative deviation between the observed frequency in the input signal and the sampling pulse frequency is detected by the control circuit 18 by calculation, and this calculated value is latched in the latch circuit 25. The output of the latch circuit 25 is converted into an analog signal by a DA converter 26, and a bias voltage from a DC power supply 27 is applied to this analog signal as necessary to an adder circuit 28.
The output of this addition is given to the clock signal generator 17 as a signal for controlling the clock cycle mantissa.

The clock signal generator 17 is composed of, for example, a voltage controlled oscillator. As a result, external sample clock input terminal 16
The sample clock frequency of the system under test 11 is corrected.
always has a certain relationship to the frequency of the observed spectrum of the input signal.

In this case, as described above, the value of the frequency deviation is detected from the relationship shown in FIG. 4 by observing the three spectra, and the deviation is not limited to the case where the correction data is latched in the latch circuit 25. You can do it like this. Alternatively, by detecting the relative level of the spectrum ωl (+1 and -1 at ω), for example, if -s is larger at ω, data that lowers the sample clock frequency by a certain value is added to or subtracted from the data in the latch circuit 25. , observe the subsequent state and gradually change the data of the launch circuit 25 to add + to the spectrum ω.
The control may be performed until the levels of I and ω match -+ Furthermore, there is no need to provide the control circuit 18, since the fast Fourier transform (FFT') analyzer 12 is an arithmetic device. The calculations performed by the control circuit 18 may be placed within the FFT analyzer 2, and the results of the calculations may be latched by the latch circuit 25. In addition, the DC power supply 27 is used to apply a bias voltage, and its value is fixed, so it is used as a fixed digital value in the control circuit 18 or in the FFT analysis analyzer 2. You can also give it as

Furthermore, the frequency of the clock signal generator 170 may be controlled by digital processing, so that each time a digital value or pulse is applied, the output clock is raised or lowered by a certain frequency depending on the pulse and polarity. The latch circuit 25 and ADf converter 26 can also be omitted. For example, as shown in FIG. 5, the clock signal generator 17 counts the clock of the basic clock generator 31 with a down counter 32, and when the count value of the down counter 32 reaches O, it is detected with a zero detection circuit 33. , the frequency setting value obtained by the FFT analyzer 12 or the control circuit 18 at that time is preset in the down counter 32, and the output of the zero detection circuit 33 is sent to the flip-flop 35. It is also possible to obtain an external sample clock signal with a 50% tutity by dividing the frequency into 50%.From a good point of view, it is possible to obtain three spectra: the observed spectrum, a certain number of higher spectra, and a fixed number of lower spectra. To obtain only these, there is no need to perform fast Fourier transform.
Discrete Fourier transform may also be used. Furthermore, +2 degrees for ω and -2 for ω may also be detected as low spectra and low spectra with respect to the observed spectrum ωk.

<Effects> As described above, according to the present invention, the ratio of high and low spectra to the observed spectrum observed by the Fourier analyzer is always detected, and the external sampling clock is set so that the ratios of the high and low spectra are at the same level. By controlling the frequency, it is possible to make the external sampling clock frequency follow the frequency of the analog input signal with higher precision than the frequency resolution provided by ordinary fast Fourier transform. Moreover, this makes it possible to reduce the spectral level error due to the cut-off error peculiar to the discrete Fourier transform.

Furthermore, in order to utilize the spectral analysis output of the input signal in this way, a detector for detecting the basic clock is provided separately in the system under test, such as a rotation detector as described in the case shown in Figure 1. There is no need to provide Of course, if a clock can be obtained from the system under test 11, it is also possible to use it to make the external sampling clock frequency follow this frequency. In particular, in order to perform arithmetic processing on the action of the control circuit 18 within the fast Fourier transform analyzer 12, order analysis can be performed by simply providing simple components such as a launch circuit, a DA converter, and an external clock signal generator. , does not require a complicated frequency converter that required a conventional multi-stage PLL.Also, even if the system under test has multiple vibration sources, it can easily follow the frequency of any vibration source. Furthermore, even when the input analog signal contains noise, analysis with considerably high accuracy is possible.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing a conventional discrete Fourier transform analyzer, Fig. 2 is a block diagram showing an example of a discrete Fourier transform analyzer according to the present invention, and Fig. 3 shows its observed spectrum and high and low spectra. A diagram showing the level relationship of
FIG. 4 is a diagram showing the level relationship of three spectra with respect to input signal deviation, and FIG. 5 is a block diagram showing a modification of the clock signal generator. 11: System under test, 12: Discrete Fourier transform analyzer, 1
3: Signal input terminal, 16: External sample clock input terminal, 17: Clock signal generator, 18: Hrli control circuit, 25: Launch circuit, 26: DA converter, 28: Adder. Patent Applicant Takeda Riken Kogyo Co., Ltd. Agent Takuoki Kusano 1 Figure'yr72 Figure 3 Square 5 Circle

Claims (1)

[Claims] (1) In a discrete 7-lier transform analyzer, a means for obtaining spectra higher and lower by a certain frequency with respect to the observed spectrum in the analysis spectrum, and a means for obtaining the observed spectrum by detecting the level ratio of these higher and lower spectra. A discrete Fourier transform analyzer comprising means for detecting a deviation of the ratio between the frequency of the sampling clock and the frequency of the sampling clock from a predetermined value, and correcting the frequency of the sampling clock based on the detected value.