CN101476906B - Anti-strong fixed interference digital signal processing system of vortex street flowmeter - Google Patents

Abstract

The invention relates to the field of flow detection and provides an anti-strong fixed interference digital signal processing system of a vortex flowmeter, which comprises a vortex flow sensor, a charge amplifier, a passive low pass filter, a limiting amplifier, an active low pass filter, a voltage follower, a power supply module, a DSP chip, an LCD module, a keyboard module, a communication module, an EEPROM and an active crystal oscillator. The system processes the output signal of the vortex flow sensor by using a method of notch filtering and FFT-based power spectrum analysis and carries out spectral calibration, acceleration algorithm, limiting filtering and average treatment after obtaining the frequency of the vortex signal so as to eliminate the influences of strong fixed interference and obtain the value of the frequency of the vortex signal quickly and accurately.

Description

The anti-strong fixed digital signal processing system of vortex street flowmeter that disturbs

Technical field

The present invention relates to field of flow detection, be a kind of digital signal processing system of vortex street flowmeter, particularly a kind of is the anti-strong fixed digital signal processing system of vortex street flowmeter that disturbs that core, power spectrum analysis combine with notch filter with digital signal processor (DSP).

Background technology

Handle the output signal of vortex flow sensors based on the power spectrum analysis method of FFT, can from the sensor output signal that contains noise, extract flux of vortex street information effectively, compare with the method for amplification, filtering, shaping and the counting of routine, can expanded range than and guarantee the in-site measurement precision.Decide the vortex signal frequency because this method is the principle according to the energy maximum, think promptly that in sensor output signal the pairing frequency of spectrum energy maximum is the frequency of flux of vortex street.As seen, adopting the prerequisite of this method is the energy that the energy of vortex flowmeter signals is greater than noise.In the ordinary course of things, the energy of most of random noise all is the energy less than vortex signal.But some scene exists some strong jammings, and its energy is often greater than the vortex signal energy.According to the field experiment interpretation of result, from Analysis of Pipe Vibrations having the greatest impact to vortex shedding flow meter.When flow hour, the energy of pipe vibration is often greater than the energy of vortex signal, and vibration frequency is in the frequency range of vortex signal.If still judge, will cause very big measuring error according to the principle of energy maximum.Consider after field apparatus and piping erection well, frequency this actual conditions that just have been fixed up of vibration, patent of the present invention proposes following scheme and solves this problem: the signal that (1) gathers vortex street sensor carries out spectrum analysis, according to vortex flowmeter signals be broadband signal and fixedly vibration signal be the characteristics of narrow band signal, and field apparatus has a related parameter, rotating speed of motor etc. is for example determined the fixedly interfering frequency of pipeline; (2), design notch filter, to fall into fixedly undesired signal according to fixing interfering frequency.Usually use notch filter different with us, here She Ji notch filter not some or several want the frequency component of trap to be restricted to zero, but be restricted to by amplitude.If amplitude limitation that will trap frequency because this frequency within the frequency range of flow signal, just may be removed flow signal, also just can't reflect the size of flow to zero.By amplitude is the parameter that digital vortex shedding flowmeter is provided with, when the output voltage of charge amplifier and voltage amplifier ends amplitude less than this, with regard to thinking that the output of sensor is undesired signal, just with its amputation.

Summary of the invention

The problem to be solved in the present invention is: present digital vortex shedding flowmeter all is based on maximum this principle of vortex signal energy and removes to handle the vortex street sensor signal, when there is the strong jamming of some fixed frequencies in measure field, its energy is greater than the energy of vortex flowmeter signals, if go to determine flow signal by the principle of energy maximum, will produce wrong result.

The technical solution adopted in the present invention is: the vortex street sensor output signal is carried out notch filter, fixing interference frequency component is fallen into the amplitude of ending of instrument setting.This is under the normal condition by amplitude, the pairing signal amplitude of flow measurement weight range lower limit.Then, carry out spectrum analysis again, promptly adopt power spectrumanalysis to go out signal spectrum based on FFT, determine the frequency of vortex signal then by the position of seeking the peak power spectrum, again the vortex signal frequency is carried out aftertreatment, comprise frequency spectrum correction, acceleration, limit filtration and average treatment, with frequency values accurately apace.Just can calculate flow according to frequency and instrument coefficient.

Concrete technical solution is as follows:

System of the present invention is by vortex street sensor, charge amplifier, passive low ventilating filter, limiting amplifier, active low-pass filter, voltage follower, power module, dsp chip, LCD module, Keysheet module, communication module, EEPROM, active crystal oscillator;

From the faint charge signal of vortex street sensor output, at first change it into voltage signal through charge amplifier; Because signal amplitude is less, and contains many noises, passes through one-level low-pass filter circuit and one-level amplitude limiting amplifier circuit again; For fear of signal is saturated circuit, signal frequency are calculated and impacted, adopt amplitude limiting amplifier circuit; Signal after the amplification passes through low-pass filter again, the high frequency interference in the filtered signal, and by voltage follower (impact damper) output, ADC sampling and conversion by DSP carries become digital quantity; The method that DSP adopts power spectrumanalysis and the notch filter based on FFT to combine is handled the signal that the ADC sample conversion is come; When signal quality is relatively good, be that interfering energy is smaller, when numerical value is no more than signal amplitude, do not need to use notch filter, employing goes out signal spectrum based on the FFT power spectrumanalysis, determines the vortex signal frequency by seeking the peak power spectral position, again the vortex signal frequency is carried out aftertreatment, comprise frequency spectrum correction, accelerating algorithm, limit filtration and average treatment, obtain vortex signal frequency accurately fast; When there is more intense fixing interference in the scene, open notch filter, the limit interferences noise makes it be no more than vortex flowmeter signals, carries out aforesaid spectrum analysis and aftertreatment again, obtains the vortex signal frequency; The vortex signal frequency is delivered to LCD show, and deliver to outside MCU (single-chip microcomputer) by communication module and handle.

The digital processing method that adopts notch filter to combine with spectrum analysis, at first the vortex street sensor output signal is carried out notch filter, the strong jamming noise of fixed frequency is fallen into the amplitude of ending of instrument setting, adopt the frequency spectrum that goes out signal based on the power spectrumanalysis of FFT again, then, determine the frequency of vortex flowmeter signals by the position of seeking the peak power spectrum.

The trap width and the ρ of the notch filter of design ₂Between the pass be ρ ₂=a+bw+cw ², wherein, w is the trap width, a=1.0001568, b=-3.1566654, c=-15.010120; Trap depth is $d=\frac{1-{\mathrm{\ρ}}_1}{1-{\mathrm{\ρ}}_2};$ The width and the degree of depth according to trap just can be determined ρ ₁And ρ ₂, ρ ₁And ρ ₂Be the parameter in the notch filter transport function, relevant with the width and the degree of depth of notch filter.

Adopt the fft algorithm of decimation in frequency base 2, in DSP, adopt the assembly language programming, adopt fixed-point arithmetic to realize this algorithm; In order to analyze the frequency spectrum of small magnitude signal exactly, before being FFT, data are carried out convergent-divergent, promptly when data are too big, data are dwindled; And when data too hour, data are amplified.

In order to accelerate response speed, the vortex signal frequency values is placed among the Q1 of round-robin queue of 50 length, be provided with a threshold value fre_limit; When the frequency values that newly enters Q1 on same direction continuous 4 then averaging process restarts when surpassing fre_limit, past data does not participate on average.

Adopt the method for limit filtration that the vortex signal frequency is handled, promptly the frequency that enters Q1 is carried out limit filtration

$f_{\mathrm{in}}=\left\{\begin{array}{cc}{f}_c,& |f_c-f_p|\≤|\mathrm{fre}\_\mathrm{limit}|\\ f_p+|\mathrm{fre}\_\mathrm{limit}|,& f_c-f_p\≥|\mathrm{fre}\_\mathrm{limit}|\\ f_p-|\mathrm{fre}\_\mathrm{limit}|,& f_p-f_c\≥|\mathrm{fre}\_\mathrm{limit}|\\ 0,& f_c\≤0\end{array}\right.$

In the formula, f _InBe filtered frequency values, f _cBe current frequency, f _pIt is the frequency of exporting in the past.

Adopted the two-stage average treatment, promptly except the Q1 of round-robin queue, opened up the Q2 of round-robin queue of 50 length again, the result who is obtained by power spectrumanalysis puts into Q1 earlier, the frequency data of Q1 the inside are carried out foregoing processing after, average again; Average result is put into Q2, Q2 is gone minimum and maximum value after, carry out running mean again, obtain final vortex signal frequency.

Digital signal processor adopts the DSPIC33FJ256MC710 type dsp chip of MICROCHIP company; The external capacity of dsp chip is 4KB memory chip EEPROM, to preserve instrument parameter, flow information; Dsp chip the 51st is connected with communication module with 52 pin, communication module and outside MCU communication; DSP the 63rd pin is connected with the active crystal oscillator U11 of high precision.

Advantage of the present invention is: the method that (1) adopts notch filter to combine with power spectrum has solved the influence of strong jamming noise to vortex flowmeter signals effectively; (2) propose accelerating algorithm, accelerated the response speed of instrument; (3) by limit filtration and two-stage average treatment, reduced the fluctuation of measurement result.

Description of drawings

Fig. 1 is the general structure block diagram of system hardware of the present invention.

Fig. 2 is a system of the present invention main algorithm process flow diagram.

Fig. 3 is the circuit theory diagrams of charge amplifier in the system of the present invention.

Fig. 4 is the bias voltage circuit schematic diagram.

Fig. 5 is the circuit theory diagrams of passive low ventilating filter and limiting amplifier in the system of the present invention.

Fig. 6 is the circuit theory diagrams of active low-pass filter and voltage follower in the system of the present invention.

Fig. 7 is the circuit theory diagrams of dsp chip in the system of the present invention.

Fig. 8 is the circuit theory diagrams that 24V transforms to 3.3V in the system of the present invention.

Fig. 9 be in the system of the present invention 3.3V to the circuit theory diagrams of 3V.

Figure 10 is the circuit theory diagrams of EEPROM module in the system of the present invention.

Figure 11 is the circuit theory diagrams of communication module in the system of the present invention.

Figure 12 is the schematic diagram of keyboard circuit in the system of the present invention.

Figure 13 is the circuit theory diagrams of liquid crystal interface in the system of the present invention.

Figure 14 is the circuit theory diagrams of active crystal oscillator in the system of the present invention, system's operate as normal display circuit.

The notch filter amplitude versus frequency characte of Figure 15 ρ=0.99 o'clock.

The notch filter amplitude versus frequency characte of Figure 16 ρ=0.9 o'clock.

Figure 17 ρ ₁=0.998, ρ ₂=0.99 o'clock notch filter amplitude versus frequency characte.

Figure 18 ρ ₁=0.98, ρ ₂=0.9 o'clock notch filter amplitude versus frequency characte.

Figure 19 ρ ₁=0.99, ρ ₂=0.998 o'clock notch filter amplitude versus frequency characte.

Figure 20 ρ ₁=0.9, ρ ₂=0.98 o'clock notch filter amplitude versus frequency characte.

Figure 21 ρ ₁=0.994, ρ ₂Trap depth under=0.99 o'clock different trap frequency.

Figure 22 ρ ₁=0.998, ρ ₂Trap depth under=0.99 o'clock different trap frequency.

Figure 23 trap width and ρ ₂Between relation curve.

Figure 24 50Hz and 55Hz superposed signal.

The amplitude spectrum of Figure 25 50Hz and 55Hz superposed signal.

Signal waveform behind Figure 26 trap.

The frequency spectrum of signal behind Figure 27 trap.

Figure 28 frequency computation part result's fluctuation situation.

Figure 29 frequency jitter situation after treatment.

Figure 30 aftertreatment process flow diagram.

Figure 31 whole software structure block diagram.

Figure 32 watchdog routine process flow diagram.

Embodiment

The present invention is further illustrated below in conjunction with accompanying drawing.

The general structure block diagram of system hardware of the present invention as shown in Figure 1.System of the present invention is by vortex street sensor, charge amplifier, passive low ventilating filter, limiting amplifier, active low-pass filter, voltage follower, power module, dsp chip, LCD module, Keysheet module, communication module, EEPROM, active crystal oscillator.

The course of work of system of the present invention is: from the faint charge signal of vortex street sensor output, at first change it into voltage signal through charge amplifier.Because signal amplitude is less, and contains many noises, so, one-level low-pass filter circuit and one-level amplitude limiting amplifier circuit passed through again.For fear of signal is saturated circuit, signal frequency are calculated and impacted, adopted amplitude limiting amplifier circuit here.Signal after the amplification passes through low-pass filter again, the high frequency interference in the filtered signal, and by voltage follower (impact damper) output, ADC sampling and conversion by DSP carries become digital quantity.The method that DSP adopts FFT power spectrumanalysis and notch filter to combine is handled the signal that the ADC sample conversion is come.When signal quality was relatively good, promptly interfering energy was smaller, when numerical value is no more than signal amplitude, did not need to use notch filter.Employing goes out signal spectrum based on the power spectrumanalysis of FFT, determine the vortex signal frequency by seeking the peak power spectral position then, again the vortex signal frequency is carried out aftertreatment, comprise frequency spectrum correction, accelerating algorithm, limit filtration and average treatment, to obtain vortex signal frequency more accurately fast.When there is more intense fixing interference in the scene, open notch filter, the limit interferences noise, make it be no more than vortex flowmeter signals, thereby do not influence measurement result, carry out aforesaid spectrum analysis and aftertreatment again,, obtain vortex signal frequency more accurately fast to eliminate the strong fixed influence of disturbing.The vortex signal frequency is delivered to LCD show, and deliver to outside MCU (single-chip microcomputer) by communication module and handle.The signal processing algorithm flow process as shown in Figure 2.

As shown in Figure 3, vortex street sensor, be of the charge amplifier amplification of the charge signal of piezoelectric sensor output via amplifier U2A, capacitor C 1, C2, C4, C5, C6, C7, resistance R 1, R2, R3, composition, wherein, C4, C5, C6, C7 are feedback capacities, have determined the enlargement factor of charge amplifier.Bias voltage Vref among Fig. 3 and other figure is provided by biasing circuit shown in Figure 4.Biasing circuit is made up of operational amplifier U2B, resistance R 6, R7 and capacitor C 14, and 1.5V is provided DC voltage.

Pass through capacitance C11 through the signal after amplifying, after removing flip-flop, deliver to the passive low ventilating filter of forming by resistance R 8, R9 and capacitor C 15, enter the limiting amplifier of forming by amplifier U5A, capacitor C 17, resistance R 10, R11, diode D3, D4 again, as shown in Figure 5.Adopting limiting amplifier is the relation that the amplitude of considering the vortex flow sensors output signal becomes with frequency square, and the frequency range broad of vortex signal, so the signal amplitude variation range is very big.And during the flow of same sensor measurement different medium, the difference of signal amplitude is also very big.In order to guarantee that the amplitude of vortex street sensor signal saturated situation can not occur under different frequency and the different medium situation, with the expanded range ratio, and make the voltage magnitude of delivering to Chip Microcomputer A C input end near full scale, guaranteeing enough resolution, so, limiting amplifier adopted.Utilized the fixing characteristics of forward conduction voltage of diode, cost is low, and is simple in structure.

Deliver to low-pass filter and voltage follower through the signal behind the limiting amplifier, as shown in Figure 6.Low-pass filter is made of the multiple feedback-type Butterworth low-pass filter of a second order amplifier U5B, capacitor C 22, C23, resistance R 12, R15, R16, R17.Voltage follower is made up of amplifier U5C.The output signal of voltage follower is delivered to the 33rd pin AN9 of DSP, as shown in Figure 7, and for the laggard line period atlas analysis of ADC sample conversion of DSP.Because the analog signal processing circuit of system adopts the 3V supply voltage, we have adopted the amplifier of rail-to-rail input, output here.

System of the present invention adopts the dsPIC33FJ256MC710DSP chip of MICROCHIP company for handling core, as shown in Figure 7.System uses the FFT computing when signal Processing, need jumbo SRAM, but outer extension memory can increase system cost, power consumption, volume greatly, and dsPIC33FJ256MC710 has the SRAM of 30KB.

DsPIC33FJ256MC710 is the high-end product in the Microship 16 bit DSP series, and it has the procedure stores capacity of 256KB and the RAM capacity of 30KB.It seamlessly integrates the computing power of digital signal processor (Digital Signal Processor is abbreviated as DSP) and the control characteristic of single-chip microcomputer (MCU).This integrated function is very desirable for the application that needs high speed, double counting and control.In addition, peripheral hardware in it the is integrated very abundant sheet mainly contains 10/12 ADC, direct memory access (DMA) (DMA), port PO RTA～port PO RTG and the Electric Machine Control PWM module of house dog, 9 16 bit timing devices (can be paired into 4 32 bit timing devices), input capture module, output comparison module, SPI serial ports, I2C bus, UART module, enhancement mode CAN module, QEI module, 32 tunnel input channels.Though dsPIC33FJ256MC710 does not have integrated liquid crystal driver module,, add a slice liquid crystal driver module, can accomplish convenient control and low-power consumption equally.

The dsPIC33FJ256MC710DSP chip provides the function of management power consumption, and this function is to realize by the clock source of CPU management and peripheral hardware selectively.The dsPIC33FJ256MC710 control device is with following four kinds of management by methods power consumptions: 1. clock frequency; 2. based on the park mode and the idle pulley that instruct; 3. the pattern of dozing off of software control; 4. carry out peripheral hardware control selectively with software.Use by certain or multiple power managed pattern can drop to the power consumption of system minimum.

In order to satisfy the requirement of the explosion-proof and static immunity experiment of vortex shedding flow meter, we have selected high-level efficiency, isolated form DC/DC module, and have designed peripheral holding circuit, filtering circuit, as shown in Figure 8.Owing to used the DC/DC of isolated form, we can with the simulation of system connect watchcase, provide system best shielding.The 3.3V voltage of DC/DC output through after the filtering as the power supply of numerical portion; Contain big ripple among the DC/DC output 3.3V, through becoming the power supply of 3V by LDO after the LC filtering as the simulation part, simulation ground and digitally between connect by magnetic bead, as shown in Figure 9.

Figure 10 shows that the EEPROM module.In order to preserve instrument parameter, flow information, it is the EEPROM of 4Kb that system has extended out a slice capacity.It has programmable WatchDog Timer, low-voltage measuring ability, and has the characteristics of low power consumption.100,000 times erasable number of times has guaranteed its reliability service.

Figure 11 shows that communication module.Utilize UART (51 pin of DSP and 52 pin) interface and the outside MCU of DSP to carry out communication, obtain the instrument parameter information or export corresponding flow information.Added the one-level light-coupled isolation between DSP and the outside MCU, data had been carried out the one-level shaping pulse again before entering optocoupler, guaranteed reliability of data transmission and security.

The man-machine interface circuit of system of the present invention is made up of keyboard circuit and LCD display circuit, as Figure 12 and shown in Figure 13.Keyboard circuit adopts simple nonencoded keyboard design, has arranged quadruple linkage altogether: set, be shifted, increase progressively and confirm, as shown in figure 12.Key assignments is by the universaling I/O port timing scan of DSP.A corresponding mouthful line is a high level when key is pressed.The LCD circuit of system of the present invention as shown in figure 13.Select the LCD of customization for use, be exclusively used in instrument such as flow, temperature, pressure, adopted the display packing of segmentation, be convenient to programming Control, power consumption is low.

Figure 14 shows that system active crystal oscillator, operate as normal display circuit.Because software algorithm to the time base accuracy requirement higher, high-precision active crystal oscillator U11 has selected in system, C37 is a power supply coupling capacitor, and the clock of time, temperature long-term stability can be provided, and guarantees the reliability service of system.After system's operate as normal, 64 pin of dsp chip, as shown in Figure 7, with the signal of output system clock, by resistance R 23 driven for emitting lights diode DS1, display system operate as normal.

The algorithm that system of the present invention adopts FFT power spectrumanalysis and notch filter to combine.When the signal quality of vortex flow sensors output was relatively good, promptly the energy of undesired signal did not need to use trapper than vortex flowmeter signals energy hour.At this moment, at first adopt the FFT power spectrumanalysis to go out signal spectrum, determine the frequency of vortex signal then by the position of seeking the peak power spectrum, the vortex street frequency is proofreaied and correct obtain accurate frequency again, the frequency after proofreading and correct is averaged obtain final frequency measurement again.Just can calculate flow according to frequency and instrument coefficient.Often there is more intense fixedly interference in some industry spot, for example, the pipeline that vortex shedding flow meter is housed links to each other with the stand that motor is installed, in this case, if flow hour, disturbing the signal that is produced just to be far longer than vortex flowmeter signals, show from instrument, is exactly a changeless frequency; If flow is not very little, undesired signal also often greater than vortex flowmeter signals, shows from instrument, be exactly frequency in saltus step, be the frequency of flux of vortex street for a moment, be the frequency of interference noise for a moment.For this reason, the vortex street sensor output signal is carried out spectrum analysis, and in conjunction with the related parameter that has of field apparatus, for example, the frequency of rotating speed of motor or valve opening/close etc. are judged the frequency that fixed in place is disturbed.With this frequency (may be one, also may be a plurality of),, enter system of the present invention by the panel setting as the trap frequency of notch filter.System of the present invention at first carries out notch filter to the vortex street sensor output signal, the interference components of fixed frequency is fallen into the amplitude of ending of instrument setting.This is under the normal condition by amplitude, the pairing signal amplitude of flow measurement weight range lower limit.It is to be noted, we do not fall into the amplitude of this interference frequency component in zero, may within the frequency range of flux of vortex street,, just can't reflect the pairing flow size of this frequency because consider this interfering frequency if this frequency component is fallen into zero.Then, carry out spectrum analysis again, promptly adopt power spectrumanalysis to go out signal spectrum based on FFT, determine the frequency of vortex signal then by the position of seeking the peak power spectrum, again the vortex street frequency is proofreaied and correct and obtain accurate frequency, again to the frequency after proofreading and correct quicken, limit filtration and average treatment, obtain final frequency measurement.Just can calculate flow according to frequency and instrument coefficient.The signal processing algorithm flow process as shown in Figure 2.

Compare with rejection filter, the benefit of trapper is that its bandwidth can be done very narrowly, simultaneously, can realize the quick decay of stopband with lower exponent number, can to two very close to signal carry out trap.Have the constraint of zero limit notch filter (zero point on unit circle, limit in unit circle, and have inphase angle mutually with limit zero point) transport function be

$H\left(z^{-1}\right)=\frac{1+\mathrm{\α}{z}^{-1}+z^{-2}}{1+\mathrm{\ρ\α}{z}^{-1}+{\mathrm{\ρ}}^2{z}^{-2}}---\left(1\right)$

In the formula, α=-2cos ω ₀, ω ₀Be the trap frequency, ρ (ρ＜1) is the parameter in the notch filter transport function, decision trap width.Amplitude-frequency response when Figure 15 is respectively different ρ with Figure 16, the horizontal ordinate among the figure is a normalized frequency, i.e. ω/ω _s, ω _sIt is the sampling angular frequency.

As seen, the width of ρ decision trap, it is just narrow more that ρ approaches 1 trap more.At its amplitude attenuation to 0 of trap frequency place, this is because at trap frequency place, and molecule is zero, and denominator is non-vanishing.Be the zero point of obtaining notch filter again

$z^{-1}=\frac{-\mathrm{\&alpha;}\&PlusMinus;\sqrt{4-{\mathrm{\&alpha;}}^2\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">i</figure-callout>}}}{2}$

Limit is

$p^{-1}=\frac{-\mathrm{\&alpha;}\&PlusMinus;\sqrt{4-{\mathrm{\&alpha;}}^2\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">i</figure-callout>}}}{2\mathrm{\&rho;}}$

As seen, | z|=1, | p|＜1, promptly zero point, limit was in unit circle on unit circle.

If the signal of interference frequency component carries out trap to containing fixedly with such notch filter, can be with the undesired signal vanishing.But if this frequency within the scope of flow frequency, just might also be removed flow information when eliminating interference, this is that we are undesirable.Non-vanishing for the amplitude versus frequency characte that makes notch filter at the trap frequency place, must make the molecule of formula (1) non-vanishing.In order to make molecule non-vanishing, simultaneously, make molecule and denominator unequal again, formula (1) is rewritten as

$H\left(z^{-1}\right)=\frac{1+{\mathrm{\&rho;}}_1\mathrm{\&alpha;}{z}^{-1}+{{\mathrm{\&rho;}}_1}^2{z}^{-2}}{1+{\mathrm{\&rho;}}_2\mathrm{\&alpha;}{z}^{-1}+{{\mathrm{\&rho;}}_2}^2{z}^{-2}}---\left(2\right)$

In the formula, α=-2cos ω ₀, ρ ₁(ρ ₁＜1), ρ ₂(ρ ₂＜1), ρ ₁And ρ ₂Be the parameter in the notch filter transport function, relevant with the width and the degree of depth of notch filter.Figure 17 to Figure 22 is the amplitude-frequency response characteristic curve of formula (2) when getting different parameters.

The zero limit of obtaining formula (2) is

$z^{-1}=\frac{-\mathrm{\&alpha;}\&PlusMinus;\sqrt{4-{\mathrm{\&alpha;}}^2\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">i</figure-callout>}}}{2{\mathrm{\&rho;}}_1}$ With $p^{-1}=\frac{-\mathrm{\&alpha;}\&PlusMinus;\sqrt{4-{\mathrm{\&alpha;}}^2\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">i</figure-callout>}}}{2{\mathrm{\&rho;}}_2}$

As seen, | z|＜1, | p|＜1, so zero limit is all in unit circle.Draw thus to draw a conclusion:

(1) when zero point is in unit circle, the gain of the trap place of notch filter no longer is 0;

(2) ρ ₁And ρ ₂The trap width of value decision notch filter, their value approaches 1 more, the trap width is just narrow more;

(3) work as ρ ₁＞ρ ₂The time, the trap place is decay; Work as ρ ₁＜ρ ₂The time, the trap place is for amplifying.

Because system of the present invention will be notch filter, so, only analyze ρ ₁＞ρ ₂Situation.Simultaneously, in order to obtain trap depth (gain at trap frequency place) and ρ ₁And ρ ₂Relation, carry out following analysis.

If ω ₀Be the trap frequency, will $z=e^{j{\mathrm{\&omega;}}_0}=\cos {\mathrm{\&omega;}}_0+j\sin {\mathrm{\&omega;}}_0$ And α=-2cos ω ₀Substitution formula (2), the gain that can obtain at trap frequency place is

$\left|H\left(z^{-1}\right)\right|=\sqrt{\left|\frac{1+{\mathrm{\&rho;}}_1\mathrm{\&alpha;}{z}^{-1}+{{\mathrm{\&rho;}}_1}^2{z}^{-2}}{1+{\mathrm{\&rho;}}_2\mathrm{\&alpha;}{z}^{-1}+{{\mathrm{\&rho;}}_2}^2{z}^{-2}}\right|}$

$=\sqrt{\frac{(1-2{{\mathrm{\&rho;}}_1}^2+{{\mathrm{\&rho;}}_1}^4)-4{\mathrm{\&rho;}}_1(1-2{\mathrm{\&rho;}}_1+{{\mathrm{\&rho;}}_1}^2){\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2{\mathrm{\&omega;}}_0}{(1-2{{\mathrm{\&rho;}}_2}^2+{{\mathrm{\&rho;}}_2}^4)-4{\mathrm{\&rho;}}_2(1-2{\mathrm{\&rho;}}_2+{{\mathrm{\&rho;}}_2}^2){\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2{\mathrm{\&omega;}}_0}}---\left(3\right)$

$=\sqrt{\frac{{(1-{\mathrm{\&rho;}}_1)}^2[{(1+{\mathrm{\&rho;}}_1)}^2-4{\mathrm{\&rho;}}_1{\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2{\mathrm{\&omega;}}_0]}{{(1-{\mathrm{\&rho;}}_2)}^2[{(1+{\mathrm{\&rho;}}_2)}^2-4{\mathrm{\&rho;}}_2\mathrm{<figure-callout\; id="2"\; label="co\; s"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">co</figure-callout>}{s}^{}{}^2{\mathrm{\&omega;}}_0]}}$

If change the trap frequency, the trap depth curve of notch filter under different trap frequencies that just can draw, extremely shown in Figure 22 as Figure 21.Trap depth is the gain of notch filter at trap frequency place.Notice that the implication of the notch filter amplitude-versus-frequency curve figure on Figure 21 to Figure 22 and Figure 15 to Figure 20 is different.

As seen, if ρ ₁And ρ ₂Fixing, when trap frequency during not in normalized frequency 0,0.5 and 1, promptly not 0, near fs/2 and the fs time (this point is easy to be met in practice), trap depth is constant basically.Work as ρ ₁And ρ ₂

Be in close proximity to 1, and ω ₀During not 0, near π, 2 π, formula (3) can be reduced to

$\begin{array}{c}\left|H\left(z^{-1}\right)\right|=\sqrt{\frac{{(1-{\mathrm{\&rho;}}_1)}^2[{(1+{\mathrm{\&rho;}}_1)}^2-4{\mathrm{\&rho;}}_1{\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2{\mathrm{\&omega;}}_0]}{{(1-{\mathrm{\&rho;}}_2)}^2[{(1+{\mathrm{\&rho;}}_2)}^2-4{\mathrm{\&rho;}}_2{\mathrm{<figure-callout\; id="2"\; label="cos"\; filenames="H2009101161071E00022.png,H2009101161071E00041.png"\; state="\{\{state\}\}">cos</figure-callout>}}^2{\mathrm{\&omega;}}_0]}}\\ \&ap;\frac{1-{\mathrm{\&rho;}}_1}{1-{\mathrm{\&rho;}}_2}=d\end{array}---\left(4\right)$

As seen, trap depth d is by ρ ₁And ρ ₂Decision, and be not subjected to ω ₀Influence.

Thus, can design and have the fixedly notch filter of trap depth.The trap width of notch filter is by ρ ₁And ρ ₂Decision, ρ ₁And ρ ₂Approach 1 more, trap is narrow more; The degree of depth of trap is determined by the relation of formula (4).

If need to fall into 2 fixing frequencies, only need design a notch filter again, this notch filter and previous notch filter cascade are got final product.The rest may be inferred, can fall into the interference of n fixed frequency.When specifically realizing, computing after the transport function cascade of several notch filters also can be carried out independent computing to a plurality of notch filters with program, which kind of mode no matter, its calculated amount is identical.

The front points out that the ρ in formula (1) denominator is relevant with the trap width.If ρ ₁=1, ρ in the denominator in the analysis mode (2) ₂Relation with the trap width.The trap width is defined as signal attenuation to the difference between the upper and lower limit frequency at 70.7% place of former amplitude.In order to draw the conclusion that all is suitable for for any sample frequency, at this frequency is carried out normalized, like this trap width has also been carried out normalization.Adopt the Matlab language,,, draw different ρ at different trap frequencies according to formula (2) ₂Under amplitude-frequency response, calculate the trap width according to these curves, obtain under different normalization trap frequencies trap width and ρ thus ₂Between relation, as shown in table 1.

Table 1 trap width and ρ ₂Between relation

Line direction from table 1 is worked as ρ ₂When identical, different trap frequencies, the trap width is substantially the same; From the direction of row, trap width and ρ ₂There is certain relation.For this reason, adopt least square fitting, set up out trap width and ρ ₂Between relation, as the formula (5).

ρ ₂＝a+bw+cw ² (5)

In the formula, w is the trap width, a=1.0001568, and b=-3.1566654, c=-15.010120, shown in Figure 23 is trap width and ρ ₂Between relation curve.

So we can be according to trap depth and trap width, through type (5) and formula (4) are determined ρ ₁And ρ ₂

We test notch filter.Figure 24 and Figure 25 are signal and the frequency spectrum of the signal of two identical amplitudes of 50Hz and 55Hz when being superimposed.Signal and corresponding frequency spectrum such as Figure 26 and shown in Figure 27 after the notch filter that employing is 0.3 at 50Hz place trap depth, width is 10Hz carries out notch filter to this signal.As seen, the signal amplitude of 50Hz and 55Hz is very approaching before the notch filter; Through behind the notch filter, the signal amplitude of 50Hz approximately decays to original about 30%, and the not decay basically of the signal of 55Hz.The good results are evident for notch filter.

Because the fixed-point algorithm accuracy limitations, when notch filter trap frequency at f＜0.02f _sOr f＞0.48f _sThe time, the notch filter of realizing with fixed-point algorithm had just lost efficacy, because a number is represented with 16 bits in fixed-point algorithm, adopt 1.15 forms, i.e. 1 bit sign position, 15 decimal places, and the employing floating-point arithmetic, one number represents that with 32 bits wherein mantissa is 23, and precision height, dynamic range are big, so, just do not have this problem.Thereby accelerate computing velocity in order to reduce calculated amount as far as possible, in system of the present invention, when notch filter trap frequency at 0.02fs＜f＜0.48f _sAdopt the assembly language fixed-point algorithm to realize notch filter in the time of in the scope, and when the trap frequency is not in this scope, just adopt C language floating-point arithmetic to realize notch filter.

System of the present invention adopts the signal of handling the process notch filter based on the power spectrum analysis method of FFT.FFT is the fast algorithm of discrete Fourier transformation (DFT).Adopt the fft algorithm of decimation in frequency base 2, in DSP, adopt fixed-point arithmetic to realize this algorithm, the precision height, the time is few.In program, adopt assembly language to realize.In order to analyze the frequency spectrum of small magnitude signal exactly, before being FFT, data are carried out convergent-divergent, promptly when data are too big, data are dwindled; And when data too hour, data are amplified, after handling like this, the computational accuracy of FFT is greatly improved.

System of the present invention determines the frequency of vortex signal by seeking the peak power spectral position after signal is carried out power spectrumanalysis.In order to realize accurate, measurement fast, result of calculation some effective aftertreatments have been carried out to flow.

(1) adopt the center of gravity correction method that FFT power spectrumanalysis result is proofreaied and correct.Updating formula as the formula (6).

$k^{\&prime;}=\left\{\begin{array}{cc}\frac{k*y\left(k\right)+(k+1)*y(k+1)}{y\left(k\right)+y(k+1)},& y(k-1)<y(k+1)\\ y\left(k\right),& y(k-1)=y(k+1)\\ \frac{k*y\left(k\right)+(k-1)*y(k-1)}{y\left(k\right)+y(k-1)},& y(k-1)>y(k+1)\end{array}\right.---\left(6\right)$

In the formula, k ' is the power spectrum position after proofreading and correct, and k is the position of peak power spectrum, and y (k) is the amplitude at peak power spectrum place.After overcorrect, just can calculate the vortex signal frequency, as the formula (7).

$f=\frac{k^{\&prime;}*f_s}{N}---\left(7\right)$

In the formula, f _sBe sample frequency, N is a sampling number.

(2) accelerate response speed.Owing to will average processing, can cause the delay of flowmeter output to the result.If flow has bigger variation at short notice, this delay will cause bigger error.In order to accelerate response speed, the vortex signal frequency values is placed among the Q1 of round-robin queue of 50 length, be provided with a threshold value fre_limit.When the frequency values that newly enters Q1 on same direction continuous 4 then averaging process restarts when surpassing fre_limit, past data does not participate on average.

(3) limit filtration.Vortex shedding flow meter is weak output signal in low discharge, and signal to noise ratio (S/N ratio) reduces, and signal quality is poor, and the vortex street frequency jitter that spectrum analysis obtains is bigger.Particularly small-bore vortex shedding flow meter, owing to reasons such as sensor installations, this phenomenon is more obvious.Among Figure 28, curve 1 be 15mm bore vortex shedding flow meter near lower frequency limit, the vortex street frequency that each FFT calculates, curve 2 is the 40mm bore vortex street frequency that each FFT obtains after calculating when lower frequency limit is distant.As seen, curve 1 fluctuation is big, and numerical value is about 30Hz.At this moment, even if take average measure, also can't make the output frequency value stabilization.For this reason, system of the present invention adopts the method for limit filtration that the vortex signal frequency is handled, and promptly the frequency that enters Q1 is carried out limit filtration, as the formula (8).

$f_{\mathrm{in}}=\left\{\begin{array}{cc}{f}_c,& |f_c-f_p|\&le;|\mathrm{fre}\_\mathrm{limit}|\\ f_p+|\mathrm{fre}\_\mathrm{limit}|,& f_c-f_p\&GreaterEqual;|\mathrm{fre}\_\mathrm{limit}|\\ f_p-|\mathrm{fre}\_\mathrm{limit}|,& f_p-f_c\&GreaterEqual;|\mathrm{fre}\_\mathrm{limit}|\\ 0,& f_c\&le;0\end{array}\right.---\left(8\right)$

In the formula, f _InBe filtered frequency values, f _cBe current frequency, f _pIt is the frequency of exporting in the past.

(4) average treatment.Because the vortex street sensor output signal is a stationary stochastic process, and the digital signal processed data limited length, must employing on average could reflect the frequency of vortex signal more accurately.For this reason, adopted the two-stage average treatment, promptly except the Q1 of round-robin queue, opened up the Q2 of round-robin queue of 50 length again, the result who is obtained by power spectrumanalysis puts into Q1 earlier, the frequency data of Q1 the inside are carried out foregoing processing after, average again; Average result is put into Q2, Q2 is gone minimum and maximum value after, carry out running mean again, obtain final vortex signal frequency.

Figure 29 is through the above final curve of output in back of handling.Figure 30 is the process flow diagram of above-mentioned aftertreatment.

The whole software structure block diagram of system of the present invention as shown in figure 31.The watchdog routine process flow diagram shown in figure 32.

Claims (5)

1. the anti-strong fixed digital signal processing system of vortex street flowmeter that disturbs, comprise vortex street sensor, charge amplifier, passive low ventilating filter, limiting amplifier, active low-pass filter, voltage follower, power module, dsp chip, LCD module, Keysheet module, communication module, EEPROM, active crystal oscillator, it is characterized in that:

From the faint charge signal of vortex street sensor output, at first change it into voltage signal through charge amplifier; Because signal amplitude is less, and contains many noises, passes through one-level low-pass filter circuit and one-level amplitude limiting amplifier circuit again; For fear of signal is saturated circuit, signal frequency are calculated and impacted, adopt amplitude limiting amplifier circuit; Signal after the amplification passes through active low-pass filter again, the high frequency interference in the filtered signal, and by voltage follower output, ADC sampling and conversion by dsp chip carries become digital quantity; The method that dsp chip adopts power spectrumanalysis and the notch filter based on FFT to combine is handled the signal that the ADC sample conversion is come; When signal quality is relatively good, be that interfering energy is smaller, when numerical value is no more than signal amplitude, do not need to use notch filter, employing goes out signal spectrum based on the FFT power spectrumanalysis, determines the vortex signal frequency by seeking the peak power spectral position, again the vortex signal frequency is carried out aftertreatment, comprise frequency spectrum correction, accelerating algorithm, limit filtration and average treatment, obtain vortex signal frequency accurately fast; When there is more intense fixed frequency interference noise in the scene, open notch filter, the limit interferences noise makes it be no more than vortex flowmeter signals, carries out aforesaid power spectrumanalysis and aftertreatment again, obtains the vortex signal frequency; The vortex signal frequency is delivered to the LCD module show, and deliver to outside MCU by communication module and handle;

The digital processing method that adopts notch filter to combine with power spectrumanalysis based on FFT, at first the vortex street sensor output signal is carried out notch filter, the strong jamming noise of fixed frequency is fallen into the amplitude of ending of instrument setting, adopt the frequency spectrum that goes out signal based on the power spectrumanalysis of FFT again, then, determine the frequency of vortex flowmeter signals by the position of seeking the peak power spectrum;

The trap width and the ρ of notch filter ₂Between the pass be ρ ₂=a+bw+cw ², wherein, w is the trap width, a=1.0001568, b=-3.1566654, c=-15.010120; Trap depth is

According to trap

The width and the degree of depth just can be determined ρ ₁And ρ ₂, ρ ₁And ρ ₂Be the parameter in the notch filter transport function, relevant with the width and the degree of depth of notch filter.

2. the anti-strong fixed digital signal processing system of vortex street flowmeter that disturbs as claimed in claim 1 is characterized in that: adopt the fft algorithm of decimation in frequency base 2, adopt the assembly language programming in dsp chip, adopt fixed-point arithmetic to realize this algorithm; In order to analyze the frequency spectrum of small magnitude signal exactly, before being FFT, data are carried out convergent-divergent, promptly when data are too big, data are dwindled; And when data too hour, data are amplified.

3. the anti-strong fixed digital signal processing system of vortex street flowmeter that disturbs as claimed in claim 1, it is characterized in that: in order to accelerate response speed, the vortex signal frequency values is placed among the Q1 of round-robin queue of 50 length, is provided with a threshold value fre_limit; When the frequency values that newly enters the Q1 of round-robin queue on same direction continuous 4 then averaging process restarts when surpassing threshold value fre_limit, past data does not participate on average.

4. the anti-strong fixed digital signal processing system of vortex street flowmeter that disturbs as claimed in claim 3 is characterized in that: adopts the method for limit filtration that the vortex signal frequency is handled, promptly the frequency that enters the Q1 of round-robin queue carried out limit filtration,

$f_{\mathrm{in}}=\left\{\begin{array}{cc}{f}_c,& |f_c-f_p|\&le;|\mathrm{fre}\_\mathrm{limit}|\\ f_p+|\mathrm{fre}\_\mathrm{limit}|,& f_c-f_p\&GreaterEqual;|\mathrm{fre}\_\mathrm{limit}|\\ f_p-|\mathrm{fre}\_\mathrm{limit}|,& f_p-f_c\&GreaterEqual;|\mathrm{fre}\_\mathrm{limit}|\\ 0,& f_c\&le;0\end{array}\right.$

In the formula, fre_limit is a threshold value, f _InBe filtered frequency values, f _cBe current frequency, f _pIt is the frequency of exporting in the past.

5. the anti-strong fixed digital signal processing system of vortex street flowmeter that disturbs as claimed in claim 4, it is characterized in that: adopted the two-stage average treatment, promptly except the Q1 of round-robin queue, open up the Q2 of round-robin queue of 50 length again, the result who is obtained by power spectrumanalysis puts into the Q1 of round-robin queue earlier, the frequency data of the Q1 of round-robin queue the inside are carried out averaging after limit filtration handles again; Average result is put into the Q2 of round-robin queue, the Q2 of round-robin queue is gone minimum and maximum value after, carry out running mean again, obtain final vortex signal frequency.

Images