Design of a Two-Stage Operational Amplifier Using Artificial Neural Network

Design of complex analog integrated circuits requires the appropriate choice of various design parameters such as MOSFET’s aspect ratio, compensation capacitance and load capacitance in a way that improves user’s desired parameters like gain, bandwidth, power dissipation and phase margin. Considering previous works, in this paper, a two-stage miller compensated operational amplifier with PMOS input pair is designed using artificial neural network. The inputs of the neural network are design parameters including DC gain, bandwidth, power dissipation and phase margin and in its output, the sizing of transistors and the amounts of reference current supply, compensation capacitance and load capacitance are acquired. In this design method, a sampling method based on parallel HSPICE simulations is employed for data acquisition from the 15-dimensional design space which results in simplicity and automation of the dataset collecting procedure and reduces the total sampling time and then this data is used for training the neural network model. In the next stage, a range sampling method is applied for making new designs from the trained model which has facilitated the design procedure and made the user-desired tradeoffs between different performance parameters of the operational amplifier possible. Moreover, if the amplifier performance figure of merit (FOM) is defined as the result of the multiplication of unity gain bandwidth and load capacitance divided by power consumption, the comparison between obtained designs of this paper’s proposed method and the results of some other methods applied for designing operational amplifiers with relatively similar topologies in previous works, indicates that this parameter has increased by 154% at the minimum.