Design of neural signal conditioning circuit

The human body's neural signals directly represent the human body's self-meaning. Studying neural signals provides a way to understand and identify the human body. Over the years, the research mainly includes neural electrodes and neural signal conditioning circuits. Neural electrodes can extract neural electrical signals from the human body, while neural signal conditioning circuits perform denoising, amplification, and identification on neural signals.

Neural signals share some common characteristics with other biological signals of the human body, but also have some unique characteristics. According to neurobiological research, neural signals are pulse-like electrical signals with a frequency of about 1kHz and up to 10kHz. For example, a bundle of motor nerves that control muscles, when an impulse potential signal arrives, the muscle fibers will contract, and the strength of the contraction varies according to the frequency of the nerve impulse. Therefore, as long as the impulse potential is identified and processed into a digital control signal, specific applications such as prosthetic control can be carried out.

Of course, the detection of neural signals also has its difficulties. The human body's neural signals are low-frequency weak signals under strong noise interference. Because they are very weak, only at the microvolt level, and the interference is extremely strong, the effective signal is often submerged. Interference signals generally include high-frequency electromagnetic interference, 50Hz power frequency interference, and polarization voltage. Power frequency interference mainly exists in the form of common-mode signals, and the amplitude can usually reach several volts to tens of volts. The polarization voltage is a DC voltage generated by the chemical half-battery formed between the measured electrode and the biological tissue, which is generally tens of millivolts and can reach a maximum of 300mV. In addition, due to the complexity and particularity of the organism, its equivalent signal source output impedance is generally large, which can be tens of kilo-ohms, which is also necessary to consider. According to the above description of the characteristics of neural signals, a neural signal conditioning circuit with strong pertinence, superior performance, stability and reliability is designed.

1 Circuit system structure and principle

According to the characteristics of neural signals and universal electrodes, the conditioning circuit must have some necessary performance. First, the circuit must have a very high common-mode rejection ratio, which can suppress power frequency interference and physiological interference other than other measurement parameters. If the common-mode rejection ratio of the circuit is 120dB, the influence of the common-mode signal in the input signal will be weakened by 1 million times, and a 1V common-mode signal is equivalent to a 1μV differential-mode signal. At the same time, the input impedance of the circuit is also a very important parameter. High input impedance can effectively reduce the influence of high internal resistance of the signal source. As mentioned above, the output impedance of the equivalent signal source of the organism can generally be tens of kilo-ohms, which requires the input impedance of the designed circuit to be greater than 100 megohms. In addition, since the human tissues contacted by each electrode are different, they show unstable high internal resistance source properties, which will cause imbalance of the electrode input impedance and convert common-mode interference into differential-mode interference. Increasing the input impedance of the amplifier is conducive to reducing the impact of this conversion. At the same time, compared with the neural signal with an amplitude of microvolts, the low noise and low drift of the conditioning circuit are also extremely important. The structure of the neural signal conditioning circuit proposed in this paper is shown in Figure 1. The circuit system is divided into three parts: pre-input amplifier circuit, intermediate signal processing circuit and subsequent signal recognition and transmission circuit.

2 Circuit Design

Early biological signal circuits were mostly designed with discrete components. With the continuous development of microelectronics technology, many high-performance integrated instrument amplifiers have emerged. Due to the excellent performance of these devices and the avoidance of installation and debugging, they have been widely welcomed in the design of biomedical instruments. This paper makes full use of these devices in the design.

2.1 Pre-input amplifier circuit

The preamplifier mainly considers the influence of noise, input impedance and common mode rejection ratio. The circuit designed here consists of three parts: input buffer, high frequency filter and instrument amplifier. The circuit structure diagram is shown in Figure 2.

Since the input buffer adopts a direct voltage negative feedback design, the input impedance is infinite in theory, which effectively isolates the human body from the circuit system and eliminates the influence of high and unstable internal resistance of the signal source.

The low-pass filter network composed of R1a, R1b, C1b, and C2 can effectively remove the influence of high-frequency electromagnetic noise. The differential mode signal cutoff frequency BWDDFF and the common mode signal cutoff frequency BWCM of the circuit are shown in equations (1) and (2), where R1a, R1b, C1a, and C1b must be exactly equal, and C2>10C1. Generally speaking, instrument amplifiers have no common-mode suppression capability for signals above 20kHz. The use of this network can make instrument amplifiers work more effectively.

Instrument amplifiers have high input impedance and common mode rejection ratio due to their classic three-op-amp structure, and only need an external resistor to set the gain, so they are widely used in the field of biological signal processing. The AD8221 selected here is the latest model, which is an order of magnitude higher than the general AD620 in all aspects. In addition, due to the existence of polarization voltage, in order to avoid circuit saturation, the gain of the preamplifier circuit must be within tens of times and cannot be too large.

2.2 Intermediate processing circuit

The intermediate stage processing circuit is divided into a bandpass frequency selection network, a secondary amplifier circuit, a 50Hz notch filter and a gain adjustment circuit.

The bandpass frequency selection network is composed of RC passive network, which is simple and reliable, and the maximum range of the passband is set to 0.05Hz~10kHz. According to individual differences, the network can be selected by the digital control circuit to combine different frequency bands to meet the best signal state.

The structure of the secondary amplifier circuit is similar to that of the gain adjustment circuit, both of which are connected by an operational amplifier in the form of voltage negative feedback. The former amplifies the signal, while the latter controls the gain of the entire circuit, which can reach a maximum of 120dB. Its structural diagram is shown in Figure 3. Here, the operational amplifier uses OP27, and the voltage series negative feedback structure is used. Its advantages are simple structure and irreplaceable superior performance: (1) The input equivalent impedance is large, Ri=(1+AF)rid, and the output equivalent impedance is small, Ro=ro/(1+AF), where rid is the input impedance of the operational amplifier and ro is the output impedance. It not only completes the signal amplification function, but also acts as a buffer, effectively isolating the modules of the front and rear stages without the need to add additional impedance converters and matching modules; (2) The use of capacitor C makes the entire module have a low-pass function, which can not only remove high-frequency interference in the signal, but also compensate the high-frequency part of the effective signal due to its advance compensation function. Through reasonable design, the phase of the circuit frequency band will change smoothly. The neural signal mentioned above is a pulse-like signal with no obvious distortion in its shape, which has a positive significance when performing time domain processing on it.

The 50Hz power frequency notch filter adopts a typical active double-T notch network solution, with Q=2.5, which can effectively remove the power frequency interference in the signal. Its structural diagram is shown in Figure 4.

2.3 Signal recognition and photoelectric coupling circuit

The neural signal is a pulse-like analog electrical signal. Before transmitting this signal to the subsequent digital circuit, it must be regularized into a standard square wave signal. This is done by the signal recognition circuit, which consists of a hysteresis comparator. In order to ensure safety and prevent interference between analog and digital circuits, the optoelectronic isolation circuit is also an indispensable module. The signal recognition and optoelectronic coupling circuit is shown in Figure 5.

The hysteresis comparator circuit is a structure that uses an operational amplifier to form a positive feedback signal. It has strong anti-interference ability and the threshold can be adjusted by UR according to actual conditions. The threshold of the circuit is given by equations (3) and (4). In addition, since the effective frequency of the signal can reach 10kHz, the speed of the photocoupler is an important indicator. Here, the 6N137 photocoupler is selected.

3 Results and discussion

After the circuit debugging is completed, a function generator is used to generate a standard signal to perform a series of performance tests on the circuit. As discussed above, the neural signal is a weak analog electrical signal similar to a pulse, so the circuit's amplification performance and phase shift characteristics are the focus of the test. Figure 6 (a) and (b) are the results of the signal amplification performance and phase shift test, respectively. The function generator generates a 500mV/1kHz sine standard signal, which is attenuated by 40dB twice to obtain a 50μV/1kHz test input signal to test the amplification performance. As can be seen from the figure, the circuit effectively amplifies the signal by 100dB. The phase shift test is performed by a 1kHz square wave simulated neural signal generated by a function generator. If the system has significantly different phase shifts for each frequency band of the signal, it is very easy to cause distortion of the signal morphology, which is very unfavorable if the signal is to be processed in the time domain. From the results, it can be seen that except for the high-frequency component in the signal that does not belong to the effective frequency band and is filtered out, the overall phase shift of the signal is stable and maintains the original morphology. After identification and regularization, it can be fully appreciated in the time domain.

From the above discussion, it can be seen that the circuit proposed in this paper effectively solves the problem of neural signal conditioning, the circuit system solves the problem of neural signal conditioning, and the circuit system is practical and reliable. At present, the research on implantable electrodes that can extract neural signals has been completed, and clinical experiments and other research work will continue to be carried out using the circuit in this paper.