Constraint Management for Collaborative Electronic Design

Juan Antonio Carballo and Stephen W. Director

University of Michigan, Ann Arbor, MI, USA

Design Automation Conference 1999, New Orleans, USA [PDF file]

Abstract

Today's complex design processes feature large numbers of varied, interdependent constraints, which often cross interdisciplinary boundaries. Therefore, a computer-supported constraint management methodology that automatically detects violations early in the design process, provides useful violation notification to guide redesign efforts, and can be integrated with conventional CAD software can be a great aid to the designer. We present such a methodology and describe its implementation in the Minerva II design process manager, along with an example design session. This session partially describes the collaborative design of an integrated circuit, including complex microelectromechanical devices, and its manufacturing process.

For information complementing this paper, please look at the Odyssey project.

Description

As shown below, our methodology is based on a set of distributed software modules that communicate with standardized message protocols, and the designers, who communicate through a user interface. A design process manager interacts with an arbitrary number of designers to help them manage the design process, supplementing the services of existing tools or frameworks. However, unlike conventional design process managers, ours carries out constraint generation at each step and sends the result to a constraint manager. This manager then handles constraint evaluation by running a constraint propagation algorithm. During this algorithm, existing CAD tools and constraint-based systems are leveraged to deal with the details of evaluating each constraint. Constraints of arbitrary form are thereby evaluated without reinventing existing methodology. After constraint evaluation, the conflict information is returned to the design process manager. The notification manager then carries out constraint violation handling, which ends in notification of new conflicts. This manager controls the notification policy, thereby preventing design information-overload.

Demo

Consider the team-based design of a MEMS system including pressure microsensors and analog circuitry. Examples of top-level constraints such a system must meet are timing, yield, resolution, dynamic range, and cost. Below we show a snapshot of the Minerva II design problem status window at a point in the design process. Designers may search for areas of the design that are likely to require redesign by examining the right most column (Conflicts),which shows the number of conflicts affecting the properties associated with each design problem. Specific types of conflicts can be viewed, e.g., only timing constraints (see lower right button). In the snapshot below the dev. Structure problem (selected row in white pane), for example, has 5 constraint violations.

In our example, the estimated yield has fallen below its minimum required value. The snapshot below shows the window by which DENIM (the notification manager connected to Minerva II) notifies the sensor device designer of this constraint violation. The figure shows that the device cross section and the process module specification are involved in the conflict. The notification suggests possible resolution strategies, in particular to reduce the top-layer thickness in the cross section.

A crucial element of searching for an effective design fix is to review the history behind the decisions that cause the conflicts. DENIM makes such a review possible. The snapshot below illustrates doing so for one of the properties involved in the yield conflict, the cross section. The history indicates that the objective of the design problem whose solution set the cross section value was to optimize the dynamic range of the sensor. The goal was to achieve at least 20 psi, and the achieved result was 25 psi. Note that designers do not need to pursue time-consuming manual input of this design history - Minerva II captured it automatically as designers designed with its support. This 5 psi margin suggests that it might be possible to fix the yield conflict by reducing the top-layer thickness in the cross section, without violating the dynamic range requirement.

For more information, please read DAC'99 paper [PDF file].

Contact email: Juan Antonio Carballo