Direct current signal measurement data filtering method

Abstract

The invention discloses a DC signal measurement data filtering method, and belongs to the technical field of testing. According to the method, a filtering scheme based on local weighting is provided for overcoming the defects of an existing mean value filtering technology. The method is characterized in that in aperture measurement time of an integral multiple power line period, weighted averaging is carried out on an ADC sampling data sequence by adopting a trigonometric function, and the weight is linearly and progressively decreased from the center of the sequence to the two ends; and carrying out continuous calculation by adopting a data queue half-length overlapping and recursive updating mode. Compared with traditional mean filtering, the method has better periodic noise and high-frequency noise suppression capability, is low in sensitivity to noise frequency fluctuation, can output nearly two times of measurement points under the same aperture time, and remarkably improves the measurement resolution and speed. The method is high in processing efficiency and easy to deploy in an existing test instrument.

Description

Technical Field

[0001] This invention belongs to the field of testing technology, specifically relating to a method for filtering DC signal measurement data. Background Technology

[0002] DC signals are a common form of electrical signal, and the testing and analysis of DC parameters for different materials, devices, and components is one of the fundamental methods for understanding their characteristics, widely used in scientific research, R&D experiments, industrial production, and other fields. With breakthroughs and applications of new technologies, materials, and processes in various fields, higher demands are being placed on the precision of DC parameter testing. Taking semiconductor parameter testing as an example, based on improvements in chip manufacturing processes and breakthroughs in third- and fourth-generation semiconductor materials, the current and voltage fluctuations in semiconductor devices during static DC parameter testing have reached the fA and nV levels. This necessitates the ability to sensitively detect subtle changes in the signal under high voltage and current conditions, placing higher demands on the noise, resolution, and dynamic range of testing instruments.

[0003] Common DC signal testing instruments include active measurement units (SMUs), digital multimeters, and semiconductor parameter analyzers. Their testing principles for DC signals are largely the same. The main principles of DC signal testing are as follows: Figure 1 As shown, the instrument's measurement section mainly comprises four core units: current measurement link, voltage measurement link, data acquisition, and control and processing. It supports the measurement of both voltage and current, two fundamental parameters. The current and voltage measurement links primarily convert and condition the measured current and voltage signals, transforming them into voltage signals suitable for the data acquisition input range to facilitate subsequent digital acquisition and processing. The data acquisition section converts the conditioned analog signals from the measurement link into digital signals for subsequent data processing and storage. The control and processing section performs calculations and filtering on the acquired raw measurement data to obtain the final measurement results, and also controls other functional units to conduct voltage and current acquisition tests in an orderly manner according to the test tasks. The combined performance of these functional units determines the instrument's testing capabilities.

[0004] For DC signal testers, there are two main approaches to reduce measurement noise and improve dynamic range and resolution to achieve measurement capabilities in the fA and nV range. One approach is to improve the performance of the hardware circuitry, enhancing the noise level of analog measurement circuits such as current and voltage measurement links, and the voltage resolution of the data acquisition unit, thereby improving the overall instrument performance. While this method effectively improves performance, it often comes at the cost of significantly increased hardware complexity, instrument size, and overall cost. In scenarios where cost and size are limited, it may not meet application requirements, necessitating performance improvements within constraints. The other approach is to integrate digital filtering algorithms into the control and processing units to reduce measurement noise and improve the ability to distinguish sensitive signals. This method can improve instrument performance without altering the hardware design.

[0005] Figure 2 The image shows the signal waveform at the ADC front end of a digital multimeter, where the horizontal axis represents 10ms / division; the vertical axis represents 1mV / division; the sampling rate is 1.0MS / s; and the bandwidth is 20.0MHz. Figure 2 It can be seen that the signal waveform exhibits obvious periodicity. Further observation of the periodic variation pattern of the signal reveals that its period is 20ms, which coincides with the period of the 220V AC power supply. Since the instrument's measurement channel uses active devices such as power amplifiers, it is inevitably subject to interference from noise sources such as AC power supply. This noise interference is periodic, and its variation period is often the same as the AC power line cycle.

[0006] Based on the noise characteristics in DC signal testing, mean filtering is widely used in DC parameter testing equipment such as digital multimeters and source measurement units. This method cancels noise by averaging the measured data over integer multiples of the power line period. Specifically, mean filtering refers to averaging M test data points directly in the digital acquisition data queue. The mathematical expression for this method is shown in formula (1): (1); In the formula: y n x is the filtered output value after the nth sampling; n is the unfiltered nth sample value; M is the average number of data points used.

[0007] The above filtering methods are usually formed into filtering algorithms and deployed in the processor of DC parameter testing instruments. Testers can adjust the number of test data to be averaged by changing the measurement aperture time of a single average filter (the integration time required for a single filter calculation), thereby obtaining different filtering effects. The longer the measurement aperture time, the higher the measurement resolution, but the measurement speed will decrease accordingly. How to optimize the noise level and improve the measurement resolution while keeping the measurement speed constant has become a current problem.

[0008] The existing solutions have the following drawbacks: (1) When the measurement aperture time is constant, the mean filtering scheme is not strong enough to suppress this type of noise when the mean filtering aperture time and the noise period are not integer multiples. (2) When the measurement aperture time is constant, the mean filtering scheme is not strong enough to suppress high-frequency noise. (3) The mean filtering scheme is highly sensitive to the frequency fluctuation of noise, which will cause large fluctuations in the noise of the measured signal after filtering. Summary of the Invention

[0009] In view of the above-mentioned technical problems in the prior art, the present invention proposes a DC signal measurement data filtering method, which is reasonably designed, overcomes the shortcomings of the prior art, and has good effect.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: A method for filtering DC signal measurement data, using an analog-to-digital converter (ADC), includes the following steps: Step 1: Obtain sampling data Set the measurement parameters, including the measurement aperture time and the sampling rate of the analog-to-digital converter, and determine the number of sampling points m of the analog-to-digital converter ADC during the measurement aperture time; Step 2: Data extraction; Extract a sampled data sequence of length m from the continuous sampled data stream of the ADC; Step 3: Assign weights; Each data point in the data sequence is assigned a weight based on a trigonometric function to form a weighted sequence, where the weight is maximum at the center of the data sequence and decreases linearly towards both ends of the data sequence; Step 4: Weighted calculation; Calculate the weighted average of the weighted sequences and output a filtered measurement. ; Step 5: Update the data sequence using a sliding window method; Remove the sampling data within the first aperture time / 2 from the current data sequence, and add the newly collected sampling data within the first aperture time / 2 to the end of the current data sequence to form a new data sequence; Step 6: For the new data sequence, repeat steps 3 to 5 to achieve continuous filtering output.

[0011] Preferably, the filtered measurement value It is calculated using the following mathematical formula: (2); Where: m is the number of ADC sampling points during the aperture measurement time; This is for measuring ADC sampling data points within the aperture time. For the data sequence of the first Data points at each location; For the data sequence of the first Data points at each location; This is the output value of the nth filter.

[0012] Preferably, the aperture measurement time is configured to be an integer multiple of the AC power line cycle.

[0013] Preferably, by using a sliding window update method, 2N-1 filtering outputs are achieved within N times the measurement aperture time.

[0014] Preferably, the method is implemented in a field-programmable gate array, a central processing unit, or a digital signal processor.

[0015] Preferably, the aperture measurement time can be configured and adjusted according to actual testing requirements.

[0016] The beneficial technical effects of this invention are as follows: 1. Enhanced noise suppression capability: Due to the adoption of a center-weighted strategy, the amplitude-frequency characteristic curve of the method of this invention has a wider and deeper notch, which has a stronger suppression capability for periodic noise and high-frequency noise components that are not strictly integer multiples of the aperture time, thereby effectively improving the signal-to-noise ratio and effective resolution.

[0017] 2. Insensitive to frequency fluctuations: The wider amplitude-frequency notch makes the performance degradation of the method of the present invention more gradual and the output more stable when facing changes in noise frequency (such as power frequency fluctuations).

[0018] 3. Higher measurement speed: By overlapping data queues and recursive calculations, the number of measurement points can be nearly doubled within the same total measurement time, achieving both improved filtering effect and higher measurement speed.

[0019] 4. High processing efficiency and low resource consumption: The method of this invention only performs weighted averaging operations in the time domain, without the need for complex frequency domain transformations (such as FFT) or long-tap FIR filtering calculations. It has a fast processing speed and low computational resource requirements for processors such as FPGA and CPU, which helps to reduce system cost and power consumption. Attached Figure Description

[0020] Figure 1 This is a block diagram of a DC signal testing instrument.

[0021] Figure 2 This is a schematic diagram of the noise waveform of the signal at the ADC input terminal of a digital multimeter.

[0022] Figure 3 This is a schematic diagram illustrating the principle of the filtering method described in this invention, showing the data queue and the trigonometric function weight distribution.

[0023] Figure 4 This is a comparison of the amplitude-frequency response curves of the traditional mean filtering algorithm and the method of this invention.

[0024] Figure 5 This is a flowchart of the process of the method of the present invention. Detailed Implementation

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: In view of the shortcomings and drawbacks of the current technical solutions, this invention proposes a DC signal measurement data filtering method, which is based on a local weighted filtering method to further improve the suppression capability of DC signal noise under the condition of a fixed measurement aperture time. This invention provides a DC signal measurement data filtering method that, based on mean filtering, applies trigonometric function weighting to the sampled data within the measurement period, and assigns the highest weight to the center point data within the aperture time interval, thereby enhancing the filtering effect. The main implementation principle of this method is as follows: Figure 3 As shown, this method also utilizes the periodic characteristics of noise, setting the measurement period (i.e., aperture time) to an integer multiple of the power line period (NPLC) so that periodic noise interference cancels each other out, thereby achieving the purpose of reducing noise.

[0026] Unlike the traditional mean filtering method (where all data points have the same weight), this method assigns different weights to the ADC sampling data according to trigonometric functions. ADC sampling data with m sampling points within the measurement aperture time is selected, and this data segment is weighted using trigonometric functions based on formula (2) to finally obtain the corresponding weighted sampling value y(n).

[0027] (2); In the formula: m is the number of ADC sampling points during the aperture measurement time; x is the ADC sampling data points during the aperture measurement time; y(n) is the nth filter output value.

[0028] like Figure 3 It can be seen that during a single measurement, the data points at the very beginning and the very end of the sampling data queue have the lowest weights, while the data points at the center have the highest weights. Therefore, the closer the sampling data is to the center, the greater its influence on the corresponding sampled value y(n) after weighting. The influence weights of the data points near the very beginning and the very end of the data queue are close to 0. To reflect the changes in the measured voltage or current signal in a timely and effective manner, this method adopts an overlapping approach for adjacent weighted calculation data queues. That is, the data in the latter half of the aperture time of the previous weighted filtering data queue will be used as the first half of the data in the subsequent weighted filtering data queue. In the subsequent weighted filtering calculation, the data points with the lowest weights in the previous weighted filtering data queue will be assigned the highest weights. This approach avoids the problem that the data points near the very beginning and the very end of the weighted filtering data queue have low weights and cannot reflect the influence of this part of the voltage and current data on the measurement results. On the other hand, this method can output 2N-1 measurement sampling points within N times the measurement aperture time, which is nearly twice as fast as the traditional mean filtering method.

[0029] The amplitude-frequency response curves of the traditional mean filtering algorithm and the locally weighted NPLC filtering algorithm used in this method are as follows: Figure 4 As shown, the horizontal axis represents frequency in units of "N×1 / aperture time", and the vertical axis represents the corresponding amplitude response. From the amplitude-frequency characteristic curves, it can be seen that this method has two main advantages over the traditional mean filtering method in terms of noise suppression: First, when the measurement aperture time is constant, and the aperture time and noise period are not integer multiples of each other, this method has a stronger ability to suppress such noise, especially high-frequency noise components; second, the amplitude-frequency curve of this method has a wider notch gap, exhibiting lower sensitivity when the noise frequency fluctuates.

[0030] The locally weighted NPLC filtering algorithm used in this invention can be deployed in processors such as FPGAs and CPUs in DC signal testing instruments. Its workflow is as follows: Figure 5As shown, after the measured voltage or current signal is converted into a digital signal through external signal conditioning and data acquisition, the original sampling data is truncated according to the set measurement aperture time. The truncated m original data points are then subjected to weighted filtering, and the final filtered voltage or current measurement result y(n) is output. After one weighted filtering operation, the original sampling data is processed recursively. The sampling data within the previous aperture time / 2 in the current filtering operation is removed from the data queue, and new sampling data within the newly acquired aperture time / 2 is input to begin the next data filtering and measurement value output.

[0031] Based on the above principles, the advantages and features of this method are summarized as follows: (1) Based on its superior periodic noise and high-frequency noise suppression characteristics compared with existing technologies, it can more effectively improve the signal-to-noise ratio and effective resolution; (2) Compared with existing technologies, this method has lower sensitivity to noise frequency fluctuations and can better cope with the problem of noise signal frequency fluctuations; (3) This method can achieve 2N-1 samplings within N times the aperture time, which is close to twice the number of sampling points compared with existing technologies. That is to say, while improving the filtering effect, it also achieves higher measurement speed; (4) This patented filtering method does not involve calculation and feature extraction in the frequency domain of the signal, and does not require complex frequency domain transformation (such as FFT) or long-tap FIR filtering calculation. It has a fast processing speed and low computational resource requirements for FPGA, CPU and other processors, which is conducive to reducing system cost and power consumption; (5) This method can be deployed without changing the hardware design of the test instrument, solving the problem of improving testing capabilities under limited hardware conditions.

[0032] Key points and protection points of this invention: (1) A trigonometric function weighted data filtering algorithm is adopted. The original data points collected during the measurement aperture time are weighted by trigonometric functions. The center point data during the aperture time is given the highest weight and the weight is linearly reduced towards the head and tail data points, thereby improving the filtering effect. (2) This method uses 2N-1 sampling times within N times the aperture time, which can effectively reflect the changes in voltage and current signals and greatly improve the measurement rate. (3) The aperture measurement time can be configured and adjusted according to actual test needs. The higher the aperture time, the better the filtering effect on DC signal noise.

[0033] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for filtering DC signal measurement data, characterized in that, The method for implementing the filter is as follows: Step 1: obtaining sampling data Setting measurement parameters, including measurement aperture time and sampling rate of the analog-to-digital converter, and determining the number of sampling points m of the analog-to-digital converter ADC within the measurement aperture time; Step 2: data interception Intercepting a sampling data sequence with a length of m from the continuous sampling data stream of the ADC; Step 3: weight distribution Distributing a weight based on a trigonometric function to each data point in the data sequence to form a weighted sequence, wherein the weight is maximum at the center point of the data sequence and linearly decreases towards both ends of the data sequence; Step 4: weighted calculation computing a weighted average of the weighted sequence, outputting a filtered measurement value ; Step 5: updating the data sequence in a sliding window manner Removing the sampling data within the first aperture time / 2 in the current data sequence and supplementing the sampling data within the aperture time / 2 newly collected at the tail of the current data sequence to form a new data sequence; Step 6: repeating steps 3 to 5 for the new data sequence to realize continuous filtering output.

2. The direct current signal measurement data filtering method of claim 1, wherein, filtered measurement value is calculated by the following mathematical formula: （2）； In the formula, m is the number of sampling points of the ADC in the measurement aperture time; is the ADC sampling data point in the measurement aperture time; is the data point in the mth position in the data sequence; is the data point in the mth position in the data sequence; is the data point in the mth position in the data sequence; is the data point in the mth position in the data sequence; is the nth filtering output value.

3. The direct current signal measurement data filtering method of claim 1, wherein, The measurement aperture time is configured as an integer multiple of the cycle of the alternating current power line.

4. The direct current signal measurement data filtering method of claim 1, wherein, Through the sliding window updating manner, 2N-1 filtering outputs are realized within N times of the measurement aperture time.

5. The direct current signal measurement data filtering method of claim 1, wherein, The method is deployed in a field programmable gate array, a central processing unit or a digital signal processor to execute.

6. The direct current signal measurement data filtering method of claim 1, wherein, The measurement aperture time can be configured and adjusted according to actual test requirements.

Images