Sensor Noise and Its Suppression Methods

【Abstract】 This paper analyzes the noise sources of sensor circuits in detail, and provides practical solutions such as shielding, isolation, etc., as well as signal processing circuits such as filtering and detection. Keywords: sensor, noise, signal processing 1 Introduction As the frontier sentinel of the automatic control system, the sensor receives the measured information like an electronic eye and converts it into an effective electrical signal, but at the same time, some useless signals are also mixed in it. We call these useless signals noise. It should be said that noise exists in any circuit, but its impact on sensor circuits is particularly prominent. This is because the output impedance of the sensor is generally very high, which makes its output signal attenuated severely. At the same time, the sensor itself is easily submerged by the noise signal. Therefore, the existence of noise will inevitably affect the accuracy and resolution of the sensor, and the sensor is the primary link in detecting the automatic control system, so it is bound to affect the performance of the entire automatic control system. Therefore, the study of noise is an important link that must be considered in the design of sensor circuits. Only by effectively suppressing and reducing the impact of noise can the sensor be effectively used and the resolution and accuracy of the system can be improved. However, there are many types of noise, the causes are complex, and the interference ability to the sensor is also very different, so the methods of suppressing noise are also different. The following is a more comprehensive study of the noise problem of the sensor. 2 Sensor Noise Analysis and Countermeasures The sources of sensor noise can be divided into internal noise and external noise according to the noise source. 2.1 Internal noise - noise from sensor devices and circuit elements 2.1.1 Thermal noise The mechanism of thermal noise is that the potential difference fluctuates when the free electrons in the resistor perform irregular thermal motion. It is caused by temperature and is proportional to it, and is expressed by the following Nyquist formula:

Where, Vn: effective value of noise voltage; K: Boltzmann constant (1.38×10 ^-23 J·K ^-1 ); T: absolute temperature (K); B: bandwidth of the system (Hz); R: resistance of noise source (Ω).
Noise sources include the internal resistance of the sensor itself, circuit resistance elements, etc.

As can be seen from formula (1), thermal noise comes from the device itself and cannot be eliminated fundamentally. It is advisable to choose resistors with smaller resistance as much as possible.

At the same time, thermal noise has nothing to do with the frequency, but is proportional to the bandwidth, that is, there is a uniform power distribution corresponding to different frequencies, so it is also called white noise. Therefore, choosing a narrow-band amplifier and phase-sensitive detector can effectively reduce noise.

2.1.2 Amplifier noise

2.1.3 Shot noise

The source of shot noise is the transistor, and its mechanism is formed by the fluctuation of charged particles reaching the electrode causing the fluctuation of current. The noise current In is proportional to the current Ic reaching the electrode and the bandwidth B, which can be expressed as:

It can be seen that when using a bipolar transistor preamplifier to amplify the output signal of the sensor, the Ic value should be as small as possible. At the same time, a narrow-band amplifier can also be selected to reduce the shot noise current.

2.1.4 1/f noise

1/f noise and thermal noise are the main noise sources inside the sensor, but the mechanism of its generation is still controversial. It is generally believed that it is a body noise, not a surface effect, caused by lattice scattering. Near the PN of the transistor is the noise generated by the irregularity of electron-hole recombination. The power distribution of this noise is inversely proportional to the frequency, hence the name. Its noise voltage is expressed as:

Hooge also proposed an empirical formula to explain 1/f noise in 1969:

In the formula, SRH and SVH are the power noise density corresponding to the resistance fluctuation and voltage fluctuation, V is the bias voltage applied to R, N is the total number of free carriers, and α is called the Hooge factor, which is a constant independent of the device size and an important parameter for judging material performance.

For rectangular resistors, the total number of free carriers N = PLWH, where P is the carrier concentration, and L, W, and H are the length, width, and thickness of the resistor.

Therefore, we can conclude that 1/f noise is related to the geometric parameters of the force-sensitive resistor. Generally, for a certain material, expanding the resistor area can increase N and reduce 1/f noise. At the same time, experiments show that blindly increasing the size will reduce sensitivity and increase the vibration amplitude of the noise spectrum, and selecting L/W = 10, L is more suitable between 100μm and 200μm.

At the same time, 1/f noise is also related to the material. Experiments show that single crystal silicon is significantly better than microcrystalline silicon, and microcrystalline silicon is slightly better than polycrystalline silicon. The main reason is that single crystal silicon has a more complete lattice structure. In addition to lattice defects, the movement of hydrogen atoms or atomic groups in the material and the boundaries of grains are also another major cause of 1/f noise.

From the above formula, it can be seen that the carrier concentration is inversely proportional to the 1/f noise, and different doping concentrations correspond to different carrier concentrations, so the doping concentration is also a factor that affects the 1/f noise. Experiments show that for every 10-fold increase in doping concentration, the 1/f noise decreases by 36% to 50%, but the optimal doping concentration is generally selected as 5×10 ¹⁵ cm ^-2 .

2.1.5 Noise generated by switching devices

Generally, when using analog multi-way switches to allow many sensor outputs to alternately use an amplifier circuit (such as MOS image sensors), the opening and closing of the switch generates corresponding noise interference, which is superimposed on the output signal.

The suppression of switching noise is usually achieved by setting up a corresponding pseudo-sensor circuit.

2.2 External noise
External noise is caused by artificial or natural interference outside the sensor circuit. The main reason is electromagnetic radiation. The noise sources are very extensive, including almost all electrical and electrical machinery, as well as natural phenomena such as lightning and atmospheric ionization. At the same time, when the analog-to-digital part of the system has a common ground and a common power supply , the frequent current changes of the digital signal generate noise in the analog circuit. They exist in the sensor circuit in the form of electrostatic coupling, electromagnetic coupling and leakage current, as shown in Figure 1.

In view of the above causes, it is necessary to adopt electrostatic shielding and magnetic field shielding for the sensor circuit to reduce the electrostatic and magnetic coupling between the noise source and the sensor circuit, so as to achieve the purpose of suppressing external noise. The measures usually taken are:

2.2.1 Processing of analog-digital mixed circuits

It is required to separate the power supply and ground wire of the analog circuit and the digital circuit, and make the internal resistance of the DC power supply of the digital circuit as small as possible to reduce the influence of the digital signal on the analog circuit.

2.2.2 Anti-interference of stray electromagnetic fields

Shielding is the main method to reduce external noise interference. Shielding is to surround components, transmission wires, circuits and assemblies with low resistance materials to isolate the mutual interference of internal and external electromagnetic or electric fields. Shielding

is generally divided into three types: electric field shielding, magnetic field shielding, and electromagnetic shielding.

Electric field shielding is mainly used to prevent interference caused by distributed capacitance coupling between components or circuits. Generally, materials with high conductivity such as copper, iron and other metals are selected. The electric field shielding body must be reliably grounded.

Magnetic field shielding is mainly used to eliminate interference caused by parasitic coupling of magnetic fields between components or circuits. Generally, materials with high magnetic permeability such as soft iron and Permalloy are selected.

Electromagnetic shielding is mainly used to prevent interference from high-frequency electromagnetic fields. Therefore, it is effective to use materials with high electrical conductivity such as copper, silver and other metals. They use electromagnetic fields to generate eddy currents inside the shielding metal to absorb its energy and achieve the purpose of shielding.

2.2.3 Isolation

Isolation is to separate the signal ground terminals of the front and rear circuits from the circuit, because they are prone to form loop currents and cause noise interference. The main isolation methods are transformers and photoelectric couplers. Transformer isolation is only applicable to AC circuits. In DC or ultra-low frequency measurement systems, photoelectric coupling isolation is often used.

2.2.4 Requirements for output lines, power lines, wiring, and wiring

The output lines of the sensor should be twisted together to reduce the influence of external magnetic lines of force. At the same time, the output line should be as short as possible.

If noise current flows into the power line and wiring, it will radiate noise magnetic fields, and will also be induced by the electromagnetic field of the noise source to pick up noise, that is, it is easy to send and receive noise. Therefore, it is necessary to make each wiring have no antenna effect. Paired wires and twisted wires can eliminate magnetic fields, but cannot completely eliminate electrostatic effects. Coaxial cable can eliminate electromagnetic fields at the same time.

When the wiring is in a ring shape, the electromotive force caused by the magnetic lines of force that cross the ring will generate noise. Therefore, the wiring should be as close as possible to the inlet and outlet wires of the current and twisted.

The balanced-unbalanced transformer has high impedance to common mode noise and low impedance to normal noise, thus absorbing the noise in the process of balanced wiring caused by unbalanced wiring to twisted wires.

2.3 Signal processing circuit to reduce noise

The sensor circuit first needs to amplify the sampled weak signal. However, there are many noise sources at the same time: sensor internal resistance, cable resistance, amplifier circuit, and electromagnetic interference sources around the circuit. Therefore, low-pass filters and differential amplifiers are usually used to suppress differential mode noise and common mode noise (as shown in Figure 3).

Assume that Vs is the signal voltage of the sensor; Vn1 and Vn2 are the induced noise voltages of the external noise source on the cable; Vns is the circuit noise. Therefore, the output voltage Vo of the differential amplifier is:

The following is a detailed description of the relevant circuit.

2.3.1 Differential amplifier circuit

The input stage of the integrated operational amplifier uses the symmetry of the differential circuit to not only eliminate the zero drift phenomenon, but also reduce the common mode signal, improve the common mode rejection ratio, and eliminate noise interference.

2.3.2 Filtering Circuit

The main function of the filtering circuit is to process the input signal, usually to filter out the noise. There are many types of filters, which can be divided into two categories: classical filters and modern filters. Classical filters can filter out noise that occupies different frequency bands from the useful signal, but they are powerless when the spectrum of the useful signal and the noise overlap. Figure 4 shows the spectrum of the signal and the noise. In Figure 4, S(f) is the useful signal, the frequency band is f1~f2, and N(f) is the noise. After filtering, only the noise outside f1~f2 can be filtered out, and the overlapping part of the noise and the signal cannot be filtered out.
Modern filters regard both signals and noise as random signals, use their statistical characteristics to derive a set of optimal estimation methods, and then implement them in hardware or software. The Villa filter is a representative example.

Usually, the classical filter is used the most. The filter circuit can be composed of passive components R, L, C, such as the line filter shown in Figure 5; it can also contain active components (as shown in Figure 6). Its advantages are mainly that it has certain signal amplification and load capacity. The filter circuit can be divided into four categories from the functional point of view: low-pass (LP), high-pass (HP), band-pass (BP), and band-stop (BS) filters. Each type has two forms: analog (AF) and digital (DF) filters. Since the sensor signal is generally a slowly changing signal, the low-pass filter is used to suppress the high-frequency noise signal the most. Usually such as Butterworth filter, Chebyshev filter, elliptical filter, etc. The commonly used low-pass filter circuit is shown in Figure 6.

2.3.3 Phase detection circuit

When the signal is known to be periodic, sampling and outputting synchronously with the signal period can be adopted to improve the signal-to-noise ratio. Although the noise is random, after N samplings, the signal-to-noise ratio can be improved by N ^1/2 times. Its principle diagram is shown in Figure 7.

If the signal and the switching cycle are T, and the off time in one cycle is ΔT, under the condition of τ＝CR》T, the noise voltage across the capacitor is:

thus,

On the other hand, the signal has gradually approached the average value Es after τ/(ΔT/T) time. Therefore, after sufficient time,

If the switch is closed every half cycle, ΔT = T/2, then the S/N is improved by 2(τf) ^1/2 times.

2.3.4 Digital Signal Processing

Digital signal processing technology (DSP) uses hardware such as microcomputers, single-chip microcomputers, DSP chips, etc., and writes software based on numerical calculations to realize signal processing. It has the advantages of precision, strong anti-interference, and fast speed, which are unmatched by analog signal processing technology. As an emerging discipline, digital signal processing technology has developed rapidly in the information age and has become another advanced method for filtering sensor systems.

3 Conclusion

The noise of the sensor inhibits the effective realization of its accuracy and has become a problem that the sensor circuit must pay attention to. However, through the analysis of the source of sensor noise, it is completely possible to use corresponding methods and signal processing circuits to effectively suppress it and ensure the normal operation of the sensor.

References
1 Hu Guangshu. Digital Signal Processing. Beijing: Tsinghua University Press, 1997
2 Qu Zugeng. Analog Electronic Technology. Beijing: Machinery Industry Press, 1995
3 Niu Defang. Principles and Applications of Semiconductor Sensors. Dalian: Dalian University of Technology Press, 1993