Application of DSP in Electric Energy Meter

The rapid development of real-time digital signal processing and ultra-large-scale integrated circuit technology has continuously promoted the improvement of the performance of digital signal processors, making them play an increasingly important role in the fields of signal processing, military and civilian electronic technology, and their application breadth and depth are also constantly expanding and deepening. Digital signal processing has great advantages over analog signal processing, mainly in terms of high precision, strong flexibility, good reliability, easy large-scale integration and storage, and can adopt a variety of high-performance digital signal processing methods and algorithms . The core and symbol of real-time digital signal processing technology is the digital signal processor. The introduction of practical algorithms such as fast Fourier transform has promoted the development of digital signal processing. Digital signal processing lies in the real-time nature of calculation and processing.

As a tool for measuring electric energy, electric energy meters have been highly valued by the national electric power department for many years , and electric energy meter manufacturers have spared no effort in designing and developing them. However, the current design level of electric energy meters in China is still relatively backward, and high-precision electric energy meters mainly rely on imports. Traditional 4-bit and 8-bit microcontrollers are inevitably stretched in terms of high-precision electric energy measurement due to their own performance limitations. The application of DSP technology in electric energy meters has brought new hope for a substantial improvement in the accuracy of electric energy measurement.

Application of DSP in Electric Energy Meter

According to the function and error accuracy requirements of the electric energy meter, we selected TI's TMS320VC5402 chip. In addition to completing the fast data processing work in the program design, we also corrected and compensated for the system nonlinear distortion.

Acquisition data processing and calculation

In practical applications, the power signal is collected into the energy meter through a mutual inductor, and then converted into a digital signal through a 6-channel 16-bit analog input front-end processor (AD73360) (A/D) and transmitted to the DSP, and then the sampled data is digitally filtered. The application of sampling technology in DSP requires a fast ADC, that is, to sample analog signals at a very fast speed, and a fast DSP is required to perform digital low-pass filtering and extraction. In digital signal processing, filtering plays an extremely important role. It solves the problems of voltage drift, temperature drift and noise that analog filters cannot overcome, thereby improving the jitter of digital signals and making the waveforms of voltage and current signals tend to be ideal. The principle block diagram of the energy meter is shown in Figure 1.

Figure 1 Schematic diagram of electric energy meter

In the sampling process, the first problem is the selection of sampling frequency. The Nyquist sampling theorem points out that if the continuous signal x(t) is of limited bandwidth, the highest frequency of its spectrum is fc. When sampling x(t), if the sampling frequency fs≥2fc is guaranteed, then x(t) can be restored from the sampled signal. When sampling x(t), we must first understand the highest cutoff frequency fc of x(t) to determine the sampling frequency fs that should be selected. If x(t) is not of limited bandwidth, an anti-aliasing filter should be used to analog filter x(t) before sampling to remove the high-frequency components of f＞fc. Therefore, before A/D conversion, an analog low-pass filter with a sharp roll-off characteristic is required to limit the spectrum of the analog signal. An ideal filter should allow all frequencies below fs/2 to pass through, while completely blocking all frequencies greater than fs/2. Usually, the selection of the filter and sampling frequency is to limit the frequency band of interest to between DC and fs/2.

First, the voltage and current input signals are sampled and filtered by RC filter network , and then A/D conversion is performed. After the A/D conversion is completed, an interrupt is generated, and the result of each conversion is read out in the interrupt service subroutine as the input of the digital low-pass filter. The input of DSP is the digital signal obtained after A/D conversion. DSP processes the input digital signal and obtains the corresponding energy after certain calculations and conversions.

Digital filter design

The difference in the operation structure of the digital filter will affect the performance indicators of the system operation, such as accuracy, error, speed and economy. In general, it is required to use as few constant multipliers and delays as possible to implement the system, and
the operation error is required to be as small as possible. We mainly use the Moving Average Filter with FIR structure .

In digital signal processing applications, it is often necessary to design a linear phase filter. FIR filters can easily achieve strict linear phase characteristics while ensuring that the amplitude characteristics meet technical requirements. In order for the filter to meet the linear phase condition, its unit impulse response h(n) is required to be a real sequence and meet the even symmetry and odd symmetry conditions, that is, h(n)=h(N-1-n) and h(n)=-h(N-1-n).

It can be seen that the FIR filter continuously delays the input sample x(n), then performs a multiplication and accumulation algorithm to output the filtering result y(n). Therefore, FIR is actually a multiplication and accumulation operator. In digital filters, the most important feature of FIR filters is that there is no feedback loop, so there is no instability problem; at the same time, the amplitude characteristics can be set arbitrarily while ensuring accurate and strict linear phase. Stability and linear phase characteristics are the outstanding advantages of FIR filters. The following is a subroutine for FIR filter design:

.TEXT
BEGIN LDP 80H,DP
LDI @STACK_ADDR,SP
LDI 21,BK
LDI 19,RC
LDI @XN_ADDR,AR1
LDI @XNNEW_ADDR,AR2
LDI @OUTNEW_ADDR,AR3
LOOP LDF *AR2,R6
STF R6,*AR1++(1)%
LDI @HN_ADDR,AR0
CALL FIR
STF RO,*AR3
BR LOOP

Data processing method

Data processing mainly involves calculating and processing the collected discrete signals and analyzing the electrical signals using the fast Fourier algorithm (see Figure 2).

Figure 2 Data processing

The most commonly used method in DSP is frequency domain analysis. For some sequences with small length, discrete Fourier transform (DFT is precisely defined as: X(m)=) is usually used, while for sequences with large length, fast Fourier transform (FFT) is usually used. The operation speed of FFT is much faster than that of DFT, but DFT is more flexible. If a small number of frequency domain values ​​need to be calculated, the DFT method can be less computationally intensive than FFT, the length of the data sequence can be arbitrary, and the calculation of N output values ​​is independent of each other. Since the output of DFT is a complex number, the real and imaginary parts are contained in two arrays of length N, and the output results can be simulated by computer software (MATLAB). The output results of DFT are calculated and plotted, and the Nth harmonic components are calculated through the application of FFT in DSP, thereby improving various technical parameters on the electric energy meter.