Analysis and Estimation on EMI Effects in AP-DRAM Interface for a Mobile Platform

Abstract

Electromagnetic Interference (EMI) is an increasingly important factor in determining whole-system performance of a mobile system. This is driven by size reduction of the mobile platform and the ever-increasing density of electronic components. In this work, we suggest an analysis approach for EMI effects in the interface between an application processor (AP) and dynamic random-access memory (DRAM) by using an I/O driver model including power delivery network (PDN) effects. By applying this approach, the EMI effect by the AP-DRAM interface is able to be accurately estimated when its operating frequency is shifted up/downward depending on the scheme of mobile operations. The EMI effect in the AP-DRAM interface can be characterized by the (1) on-chip metal layout and package layout of the I/O PDN, (2) inserted decoupling capacitance of the I/O PDN, (3) I/O driver properties that define the rising/falling time of signals, and (4) channel properties including crosstalk between adjacent lanes and their frequency response at data transfer. In a time-domain simulation, the extracted I/O driver model was used with extracted AP and memory package models, a channel model, and an interposer. The simulation was carried out by varying input stimulus depending on a bit density ratio, bit length, and a test pattern. As a result, a wide range of spectrum generated by the AP-DRAM interface up to 6 GHz was calculated. The bandwidth of data queue (DQ) signals at the interposer was found to be around 170MHz with multiple harmonics of the fundamental 1.6GHz frequency. In comparison to the measurement, the difference in bandwidth was found to be less than 10MHz and the voltage difference at the harmonics was found to be less than 10dB. Consequently, we proposed an analysis and estimation approach for EMI effects in an AP-DRAM interface which affects the degradation of communication performance in a compact mobile platform and successfully demonstrated its applications for predicting EMI effect in the communication band of interest.