Improved Behavioral Modeling Based on the Input Output Buffer Information Specification.

Abstract

High level behavioral modeling is widely used in lieu of low level transistor models to ascertain the behavior of I⁄O drivers and receivers. IBIS (Input Output Buffer Information Specification) is one of the most widely used methodology to model digital drivers as it satisfies the basic requirements of a behavioral model such as intellectual property protection, simple structure, fast simulation time and good accuracy.

As driver technology gets increasingly complicated and rise time of input signal gets increasingly smaller, important phenomenae such as simultaneous switching noise (SSN) becomes a major consideration when simulating multiple I⁄O drivers in the integrated circuit. Misrepresentation of noise might result in overestimation of signal strength and quality resulting in a high bit error rate and poor signal to noise ratio at the receiver end. This might lead to total failure of the system. With the size of the transistor shrinking, more and more I⁄Os can be accommodated on the chip, resulting in a greater probability of more drivers switching simultaneously, hence increasing the problems of SSN furthermore.

Experiments show that IBIS models over-represents noise in the quiet line when placed in an environment where multiple drivers were present and switching simultaneously. A thorough analysis is performed to determine the inadequacies in IBIS with regards to SSN. A method is presented for compensating for the missing information by complimenting the IBIS model with a black box that is simulator independent, without compromising with the speed that IBIS enjoys over the transistor models.

Model validation and system level tests are done with the IBIS model in conjunction with the black-box by passing a pseudo random bit sequence (PRBS) at the input. Eye diagrams are used to determine the quality of the received signal at the input for the behavior models as well as transistor models. These tests demonstrated a better response of the behavior model than the plain IBIS model when compared to the circuits with the transistor model. The enhancement of the IBIS models resulted in at least a 20% improvement in the simulation accuracy with behavioral models.