Design of Active Vibration Control System Based on TMS32OF2812

0 Introduction

Low-frequency vibration is a major problem that bothers people on modern ships and spacecraft. How to quickly eliminate vibration and reduce harm has become a problem that people are more concerned about. The main methods of vibration suppression are passive vibration control and active vibration control. Although passive vibration control does not require external energy and is easy to implement, it has limited control capabilities for low-frequency vibrations and poor adaptability to sudden environments. Active vibration control has greater flexibility and is particularly effective in suppressing low-frequency vibrations, and has become a method that people focus on studying.

In this paper, a multi-channel embedded active vibration control system is designed for low-frequency vibration control of a certain type of structure, with TMS320f2812 (hereinafter referred to as F2812) as the core, including sensors, conditioning circuits, D/A conversion circuits, power amplifiers and actuators. The structure is shown in Figure 1. First, the sensor picks up the vibration signal, which is converted and conditioned into a voltage signal of 0~3V by the charge amplifier, and then sent to the A/D conversion module inside the DSP for analog-to-digital conversion after filtering and limiting protection. According to the collected vibration signals, the DSP runs the control algorithm to output the control quantity, which is converted into a force or torque directly acting on the controlled object through the D/A conversion circuit and power amplifier to suppress or eliminate the vibration of the object. In order to reduce the high-frequency noise generated by the D/A conversion step signal, smoothing filtering is added. The DSP is connected to the host through the JTAG port and the emulator to compile and load the control algorithm. It communicates with the host serially through the built-in SCI module to transmit the vibration signal and control quantity to the host, so as to analyze and evaluate the control process and improve the algorithm.

1 Hardware Design

F2812 is a high-performance chip suitable for automatic control. It has a 32-bit fixed-point processor, a 150MHz clock frequency, 18K×16-bit SARAM and 128K×16-bit Flash memory. It implements multiplication operations in hardware and has a fast processing speed. In addition, it has some built-in devices and interfaces suitable for control systems, such as a 16-channel 12-bit A/D converter, an external memory parallel interface, an SCI serial communication interface, and a JTAG emulator interface. Therefore, in this system, a charge amplifier that does not require high cable distributed capacitance is selected as a preamplifier. The actuator can be a piezoelectric device or an electromagnetic actuator. According to the needs of the object, a corresponding power amplifier needs to be configured. F2812 itself does not have a D/A conversion module. This paper uses MAX547 to design and implement the D/A conversion circuit of F2812, which can output 8 analog signals at the same time.

1.1 Charge Amplifier

The commonly used charge amplifier is an inverting integrator circuit with DC feedback. To improve the low-frequency characteristics of the circuit, the resistance of the DC feedback resistor needs to be increased. For this reason, we use a T-type network to achieve a larger resistance. As shown in Figure 2, Ral7, Ral8 and Ral9 form a T-type network, and the equivalent resistance is:

By adjusting the size of Ral7, Ral8 and Ral9, a larger resistance value can be obtained. However, during the debugging process, we found that as the measurement time increases, the output will still drift to saturation. For this reason, we use a first-order in-phase integral link to extract the DC component of the output and feed it back to one end of Ral9 to form a self-following network, as shown in the left half of Figure 2. When using the integral capacitor, you should choose a capacitor with relatively small leakage current, small temperature drift, and relatively stable performance. The amplifier should select an operational amplifier with high gain, high input impedance, low bias current, and low temperature drift to ensure the performance of the charge amplifier.

1.2 Filter protection circuit

In the actual application of active vibration control, the signal frequency of concern is generally between 0.5 and 200 Hz. In order to filter out low-frequency drift and unnecessary high-frequency signals, this paper designs a bandpass filter consisting of a fifth-order Bessel low-pass and a first-order high-pass. The fifth-order Bessel low-pass is composed of two second-order low-pass and one first-order low-pass. Their parameters are: passband gain 1, cutoff frequency 475 Hz, Q value 0.577; passband gain 1, cutoff frequency 565 Hz, Q value 0.737; passband gain 1, cutoff frequency 530 Hz, and the first and second second-order low-pass are realized by unit gain KRC circuits.

Before the signal enters the ADC of the DSP, it should be adjusted to between 0 and 3V using an additive proportional circuit. If it is lower than 0V or higher than 3V, there is a risk of damaging the DSP. For this reason, we have added a limiting protection circuit, as shown in Figure 3.

1.3 D/A conversion circuit

This paper uses MAX547 as the core device to design and implement the D/A conversion circuit of F2812. MAX547 contains 8 13-bit voltage output D/A converters. Each DAC has an input latch and DAC latch before it, which can be selected separately to perform 8-way D/A conversion. Every two DACs share a reference voltage, and a total of 4 independent external reference voltages are required. The MAX547 interface signal has 3-bit address lines, which address and select 8 channels respectively, 13-bit data lines, and control signals such as chip select/CS, write/WR, asynchronous input/LD and clear/CLR. The control signals are all level triggered. These interface signals are all matched with TTL/CMOS levels, so F2812 can be directly connected to MAX547 without level conversion. MAX547 is powered by ±5V dual power supply, and the output voltage swing is -4.5V to 4.5V.

When the write signal /WR and chip select signal /CS of MAX547 are both low, and the address signals A0, A1, and A2 are valid, the input latch of the corresponding channel is turned on and the conversion value is read from the data line. When one of /WR and /CS becomes high, the data is locked into the corresponding input latch. /LD is responsible for switching the DAC latch. When /LD is low, the DAC latch is turned on and the data enters the DAC latch from the input latch. When /LD becomes high, the data is kept in the DAC latch and the DAC performs digital-to-analog conversion. When /LD, /WR, and /CS are all low, the data can be directly transferred to the DAC latch, but /LD should be 50ns later than /WR to reverse the high level. /CLR can set the DAC conversion content to 1000H, so that the analog output AGND voltage.

F2812 has an external interface (XINTF) that can map five independent external storage spaces, each of which has a chip select signal. The eight channel addresses of MAX54.7 are assigned to the external storage area 0, and the connection with F2812 through the external interface bus is shown in Figure 4. To start the digital-to-analog conversion of DACA, DACB, DACC, DACD, DACE, DACF, DACG, and DACH inside MAX547, write the data to be converted to the addresses D9H, B2H, B3H, F4H, F5H, 7EH, and 7FH respectively. For the four independent reference voltages required by MAX547, a precision reference voltage chip REF02 is used in the system to provide the standard voltage, which is buffered by four voltage followers. REF02 outputs a 5V standard voltage, and after conversion, it can output an adjustable standard voltage. The follower uses MAX494. In order to reduce the interference of the circuit leads during PCB design, it is best to connect the input end of the MAX494 op amp directly to the reference voltage input end of MAX547, as shown in Figure 4. In addition, in order to reduce the high-frequency noise caused by the D/A conversion step wave, a low-pass filter is also designed.

1.4 RS232 interface

There are two asynchronous serial interfaces (SCI) inside F2812, each with a transceiver buffer register, a transceiver shift register and a 16-level deep FIFO for receiving and sending. The baud rate of receiving and sending can be programmed, up to 64K. In the system, F2812 needs to communicate with the host through SCI, and the F2812 peripheral interface is 3.3VCMOS color level, and the computer serial port is RS-232 level, so level conversion is required between the two. MAX3232 is a commonly used RS-232 level conversion chip, and the interface circuit is shown in Figure 5.

2 Software Design

The DSP control algorithm consists of two parts: the main program and the interrupt service subroutine. The program flow chart is shown in the figure. In the main program, the system control register, watchdog, system clock, GPIO port, interrupt vector table and some peripherals are initialized, the sampling sequence of ADC, the communication baud rate of SCI and the access timing of the external memory interface are set, and then the timing cycle, counting mode and timing interrupt of the timer are set, and the state of waiting for timing interrupt is entered. The interrupt service subroutine mainly completes the following functions: first, the interrupt is turned off and the calculation variable is initialized; then the input signal of the corresponding channel of ADC is converted by A/D, the sampling value is taken out for preprocessing, the control algorithm is run to calculate the control quantity, and the control quantity is output by writing the channel address of D/A for D/A conversion, and finally the control analog signal is output to drive the actuator to reduce vibration; communicate with the host through SCI to upload the vibration signal and control quantity; finally, the ADC, SCI and timer interrupt are set, and the global interrupt is opened to prepare for the next operation of the interrupt service subroutine. After the interrupt service subroutine is completed, it returns to the main program and enters the state of waiting for interrupt, waiting for the next timing interrupt, and so on.

3 Conclusion

The embedded active vibration control system designed in this paper takes TMS320F2812 as the control core. It can collect multiple vibration signals at the same time by using the built-in A/D module of F2812. The D/A conversion circuit implemented by MAX547 can output 8 control signals, which can realize multi-input and multi-output active vibration control and is close to engineering practice. This system has been successfully used to carry out multi-input and multi-output active vibration control on a ship raft.