The Circular Electron Positron Collider (CEPC) is designed to operate at center-of-mass energies of 240 GeV as a Higgs factory, as well as at the Z-pole and the WW production threshold for electroweak precision measurements and study of flavor physics. A good particle identification on charged hadrons is essential for the flavor physics and jet study. To meet this requirement, a tracker with a drift chamber between the silicon inner tracker (SIT) and silicon external tracker (SET) is proposed in the CEPC 4th conceptual detector design. The drift chamber is expected to provide excellent PID with cluster counting technique.
In our study, a waveform-based full simulation has been performed, which includes waveform generation with Garfield++ program, simulation of electronics response and noise effects, as well as waveform analysis by utilizing effective peak finding algorithms. We developed a peak finding algorithm before based on traditional differential approach, it is further adapted using the realistic noise and electronics response parameters. We also explore the advantage of neural network for resolving a time-sequence problem, a recurrent neural network (RNN) algorithm shows great peak detection ability on MC simulation. Several optimizations in terms of the size of cell, choice of gas mixtures are performed to improve the PID capability. Preliminary results show that the K/pi separation power could be more than 2 $\sigma$ up to 20 GeV/c. The design of the drift chamber will be optimized based on the simulation study and the prototype test.