Low Power Radio Frequency Transceiver System Research Based On Artificial Neural Network

With the rapid development of the Internet of Things,artificial intelligence and semiconductor manufacturing technology,human beings are gradually entering a society of informationization and intelligence,and billions of wireless devices are continuously transmitting and receiving data.The radio frequency transceiver is the basic component of the wireless communication device and is the necessary medium for wireless commu-nication between the electronic devices.Its performance directly determines the commu-nication quality,communication mode,stability and standby time between the wireless devices.This paper reviews the development trend of ultra-low-power RF transceivers and integrated circuit design from the perspective of using cost-effective CMOS technol-ogy nodes.Combined with the basic theory of the front end of the RF transceiver,with the goal of reducing system power consumption and improving overall efficiency,the RF transceiver and its key module design have been extensively and deeply studied.Thanks to the development of mechanical learning technology in recent years,this paper inte-grates neural networks into RF transceivers for the first time,and studies and implements automatic power and channel switching based on mechanical learning technology.The main research results are as follows:1.Design of ultra low power receiver:The battery of the implanted device is of-ten impossible or extremely difficult to replace,which severely limits the power con-sumption of the wireless transceiver.Ultra-low-power RF receivers typically use extreme low-power modes of operation to conserve power.In this case,the LO's lock-up time determines the lower limit of the duty cycle.This paper first analyzes the key constraints of receiver design,and then proposes a low-power LNA based on noise cancellation tech-nology,which cancels the in-phase noise in the differential signal to achieve noise opti-mization.Then,a low power frequency combiner with frequency multiplier function is proposed.This low-power frequency combiner allows the oscillator to operate at 1/9 of the carrier frequency,with a shorter start-up time,lower operating frequency,and lower power consumption than conventional oscillators.The ultra-low power receiver is imple-mented in a 180 nm CMOS process.The test results show that the receiver has a gain of 61.2 dB,a noise figure of 16.8 dB,IIP₃ of-15.8 dBm,and a power consumption of73W.2.Design of transmitters under low voltage process:Highly integrated RF trans-mitters can reduce the manufacturing cost of wireless communication systems.How-ever,most power amplifiers still need to be implemented in high voltage process.Low breakdown voltage and high corner voltage are major drawbacks of modern nano and deep sub-micron CMOS processes,limiting the available voltage swing at the load and limiting output power.In this paper,the nonlinear problem and efficiency problem in transmitter design are studied and analyzed.On this basis,this paper proposes a method to improve the output power of the transmitter by using a power synthesizer in the low-voltage process.The Wilkinson power synthesizer and the transformer power combiner are designed and compared.A high conversion efficiency transformer power combiner is designed.Finally,a high-integration,high-efficiency,high-output RF transmitter with low-voltage process was designed and implemented using the transformer power com-biner.The simulation results show that the transmitter has an output power greater than23.3 dBm in the range of 5.2-5.8 GHz and a drain efficiency of 25.7%.3.Design of Low Power RF Transmitter Based on Mechanical Learning Technol-ogy:There are countless wireless devices in use around the world,with an annual power consumption of about 61 trillion watts.Traditional transmitters don't”know”the trans-mission distance and target,so PA usually needs to use full power to ensure that the receiver can receive signals.Artificial neural networks can also be used in RF circuit design.In this paper,the feedforward multi-layer perceptron is integrated into the trans-mitter to realize intelligent data transmission.This paper first introduces the key prob-lems and solutions of the machine learning algorithm.Then,a multi-layer perceptron with data recognition capability that can automatically switch the transmitter's transmit channel and transmit power is studied and designed.Then a voltage controlled oscillator with programmable output frequency and a high efficiency power amplifier with variable output power are designed and implemented.Finally,the overall simulation and verifica-tion of the smart transmitter proposed in this paper is carried out.The test results show that the MLP network of the transmitter can identify the data and automatically adjust the transmitter's transmit power and transmission channel,and the recognition accuracy is 95.1%.The transmitter has a peak output power of 14.9 dBm,an operating voltage of1.5 V,and a power dissipation of 34.3 mW.