Distributed and multiple time-constant electro-thermal modeling and its impact on ACPR in RF predistortion

Abstract

The performance of power amplifier linearization techniques is often impaired by thermal and electrical memory effects when the signal bandwidth is changed. Effective design of linearizers requires transistor models to be able to correctly reproduce these effects. We are presenting a distributed and multiple time-constant electro-thermal LDMOSFET model which is more accurate in predicting distributed and transient thermal effects. The distributed temperature field across the fingers is obtained using a recently reported 3D image method [Rinaldi, 2002]. A distributed equivalent electro-thermal model is then constructed to accurately and efficiently predict the average thermal resistance of large devices and is therefore particularly useful for device scaling. The electro-thermal model is further used to more accurately predict the transistor's transient thermal response so that the thermal memory effects in RF predistortion linearization be investigated separately from electrical memory effects. For this purpose an improved hybrid image method in the time-domain is derived and compared to FEM to find the LDMOSFET's transient thermal response. The transistor's transient thermal response is fitted using a multiple time-constant RC network to implement this time-dependent thermal model in an harmonic balance simulator. The complete distributed electro-thermal LDMOSFET model is then used in a predistortion linearization system. It is verified that the impact of thermal memory effects on adjacent-channel power-ratio for various modulation bandwidth is only dominant over electrical memory for signals with modulation bandwidth below 1 MHz for the device considered featuring a thin Si layer. Under such low modulation bandwidth, the multi-stage thermal model is found to be more accurate than the single-stage thermal model. The methodology should also be useful for optimizing the device layout as it establishes a relationship between thermal memory effects and the device structure.