Design Of True Random Number Generator Based On Thermal Noise

With the rapid development of the Internet of Things(IoT),information security has been widely promoted from the traditional political,military and other important fields to the field of people's livelihood and has received more and more attention.In order to achieve the confidentiality,integrity and controllability of information systems,data needs to be encrypted.The core of data encryption is the security key and its construction is based on the uncertainty and unpredictability provided by random numbers.Therefore,it is usually generated by True Random Number Generator(TRNG)in highly secured applications.TRNG uses the phenomenon with random characteristics in nature as an entropy source(such as thermal noise,nuclear decay,cosmic radiation,etc.),and the entropy source is converted into a random sequence by quantization and digital post-processing.In the field of integrated circuits,the random phenomenon that is easily available and utilized is the thermal noise.Therefore,this paper will focus on the design of true random number generator based on the thermal noise.The research contents mainly consist of the following three parts:1.After studies on the thermal noise model and the matching mechanism of sensitive amplifiers,a design of TRNG with adaptive matching ability is proposed.In this design,configurable NMOS array is embedded in the sensitive amplifier and the equivalent width-to-length ratio of the array is adjusted by applying dynamic compensation algorithm so as to balance the working current of the sensitive amplifier,thus the adaptive calibration of TRNG is achieved.The circuit is designed with TSMC 65nm technology,and the simulation results show that TRNG can operate stably in the supply voltage of 0.8V to 1.4V and the temperature of-40?to 120?.The maximum output rate is 1Gbps and the average energy consumption is 0.165pJ/bit.2.By studying the common-mode operating mode of comparator,a scheme for optimizing the randomness and robustness of TRNG with adaptive matching is proposed.In this scheme,the thermal noise inverters are connected end to end to form a loop so that the node voltage converges and remains at the metastable potential.The thermal noise at the output node of inverter in the metastable state is superimposed by the thermal noise of the comparator in the common-mode mode and is converted into random sequences by the TRNG.The proposed circuit is implemented in SMIC 55 nm process.The simulation results show that TRNG can work stably in five process corners(tt,ss,ff,sf,fs),and the uniformity of output sequences is significantly improved.The maximum output rate is 1Gbps and the average power is 0.426 pJ/bit.3.After studies on the frequency jitter mechanism,a design of TRNG based on Voltage-Controlled Oscillator(VCO)is proposed.The scheme amplified the thermal noise of resistance and took it as the control signal of VCO.Its oscillation frequency thus randomly jitter around the centre frequency.The slow oscillating signal generated by the VCO generate a raw random sequence by sampling the period-fixed high frequency oscillating signal,then using the post-processing circuit to improve the uniformity of sequence and to eliminate the autocorrelation.Adjusting the centre frequency of VCO is able to trade off the bit rate of sequence and the randomness.The proposed circuit is designed in SMIC 55nm technology with a chip area of 0.0124mm~2,a bit rate of 10Mbps and an average power of 0.81mW.The three designed TRNGs have excellent performance,and the output of true random number sequences can pass the NIST SP 800-22 randomness test.They have high security and can be widely used in signal encryption and other fields.