Current design trends for high performance systems-on-chip (SoCs) are prompting designers to adopt more and more analog/mixed signal (AMS) contents in the overall design. Unlike digital designs that are quantized in the time and amplitude domains, AMS designs are more complex because performance, noise and other factors need to be controlled continuously in time and amplitude domain with very strict tolerances.
While the cost and performance benefits of SoCs are well known, in reality the complexity of design and verification of AMS ICs are making cutting-edge designs very time consuming and error prone, hence cost ineffective. The problem is further exacerbated by the lack of skilled AMS designers and adequate EDA tools.
Virtual prototyping in digital designs
In digital designs, the EDA tools are fairly successful in keeping up with the pace of technology scaling. To enable first pass silicon or at least to reduce the design iterations, virtual prototyping is used.
Virtual prototyping essentially refines a logical design based on the tentative estimates of physical design. For example, based on different floorplanning, layout and routing topologies, the synthesis is automatically refined so as to ensure that functionality, timing and power considerations are met.
This requires an integrated fast, adaptive, incremental physical architecture estimation tool along with the timing and power estimation tool that fine tunes the synthesis logic. Figure 1 demonstrates the stages of floorplanning involved in virtual prototyping.
Figure 1 — Stages of floorplanning in virtual prototyping
Virtual prototyping in AMS designs
How do we define virtual prototyping for AMS designs? Obviously, logic synthesis is not readily extensible to analog synthesis in terms of practice — in fact, there are no analog synthesis tools in the digital sense. In addition, the concepts of automated layout cannot be applied arbitrarily to AMS circuits as AMS circuits are very sensitive to layout topology (matching) and parasitics, including inductive coupling effects.
This article will progressively elaborate the concept of virtual prototyping in AMS designs. The end goal is to define a process that reduces design cycle by bridging the gap between physical and logical designs. The recipe for this objective is defined in the following sections.
Constraints bridge logical, physical and DFM worlds
The design intent in AMS designs is often specified but not captured, translated or adopted across the design tools. Experienced designers often impose the design intents manually through constraints. A constraint is a design restriction that is used during multiple design phases to enable better success in terms of end design goals. For example, a design constraint to have symmetrical routing across a current mirror circuit would minimize the wire parasitic mismatch and hence ensure better circuit performances.
Ideally, this routing constraint (design intent) would be entered at the schematic level and can be interpreted by the routing tool. Another example can be limiting current density through a wire at the schematic level, which would eventually be understood by circuit optimizer and electromigration tool. This in turn may reduce the IR drop across that wire, or improve the median time to failure of the circuit by lessening the electromigration effect.
EDA tools at different phases of the design cycle should be constraint aware and allow entry and visualization of constraints. To enable fluidity across design phases of constraints, it is imperative that constraints should have a standard semantics and be stored in a common database that is understood all tools in the flow.
Moreover, the constraints and designs need to be co-managed with the design management tools. Constraints will provide the infrastructure to define virtual prototyping in AMS designs and act as a bridge between logical, physical and DFM world.