Raphael NXT

Overview
Raphael NXT is a fast field solver tool that provides 3D self and coupling capacitance extraction for silicon-accurate integrated circuit design. Raphael NXT complements Synopsys’ industry-leading StarRC™ full-chip parasitic extraction tool by providing an integrated solution for 3D capacitance extraction of critical nets, cells and blocks. Equipped with an ultrafast extraction engine and linearly scalable multicore technology, it enables users to efficiently extract large layout designs consisting of thousands of nets to achieve more predictable design closure. Raphael NXT supports the advanced process effects at 40-nm and beyond and it is closely correlated with Synopsys’ Raphael™, the gold-standard reference field solver.

Download Datasheet

Figure 1: Interconnect extraction hierarchy
Figure 1: Raphael NXT combines the accuracy of mesh-based field solvers with
an efficient algorithm capable of extracting larger critical circuits

Benefits

Description
Accurate interconnect parasitic extraction is critical to the success of high performance integrated circuits due to its profound impact on circuit timing, signal integrity and functionality. The accurate extraction requires detailed modeling of advanced process effects and accurate computation of the electrostatic characteristics of complicated 3-dimensional (3D) layout structures. Traditional 3D field solvers are very effective at extraction of small structures; however designers need more efficient and higher capacity extraction to handle larger sized blocks, cells or longer nets in their real designs. In addition, designers need a solution that easily fits into their existing full-chip extraction flow for ease-of-use and productivity.

Raphael NXT is a true 3D capacitance extractor that uses the Statistical Floating Random Walk method to accurately solve electrostatic equations to compute self and coupling capacitance of larger circuits. Raphael NXT accounts for all 3D effects associated with the complex geometries of interconnect structures to deliver silicon accurate extraction. It supports the latest process effects that include conformal dielectric layers, trapezoidal conductor cross sections, lithography effects, Low-K dielectric modeling, dielectric damage modeling and floating metal fill. Raphael NXT’s highly accurate capacitance extraction of critical nets, cells and blocks complements StarRC on the full-chip level as part of an overall interconnect extraction tool hierarchy (see Figure 1).

Raphael NXT can accurately extract the capacitance of thousands of critical nets in a design consisting of several million transistors. The statistical nature of the algorithm means that results are reported with statistical confidence limits. As the number of sampling random walks increases, the reported solution approaches the silicon value. The tool provides the flexibility to choose the default tolerance of 1σ at 3% total capacitance (error in total capacitance for 63% of the extracted nets is less than 3%) or specify a desired accuracy for self and coupling capacitance, if required.

Accuracy correlation with gold-standard Raphael
Raphael NXT is closely correlated with Raphael, the industry-standard field solver used by trusted major foundries as golden reference (interconnect parasitics generated by Raphael are included as part of foundry design reference guide). Figure 2 shows a correlation plot of Raphael vs. Raphael NXT for over 5000 test structures. The correlation results show a mean error of less than 2% and standard deviation of 2.5%. Though the two tools address different applications, the correlation ensures consistency of results for solvers using different solution algorithms.

Figure 2: Raphael NXT is closely correlated with gold-standard Raphael (45-nm process)

Multicore processing
One of the key advantages of the floating random walk method is its inherently parallel characteristics, i.e., each walk is independent of all others, so walks can be run in parallel with no loss of accuracy. Raphael NXT’s multicore processing capability allows users to run extractions in parallel on multiple cores with almost linearly scalable throughput (see Figure 3). The implementation uses a fine-grained parallel algorithm which runs efficiently even with different loads on different machines. As designs get larger and more nets require the accuracy of Raphael NXT, users can easily add more CPU cores to keep the total extraction time manageable.

Raphael NXT’s multicore technology is highly fault tolerant. If a CPU core is lost during a run, Raphael NXT selects a new core from the available clients. Raphael NXT allows processing across a mixture of compute resources such as Solaris and Linux.

StarRC Field Solver (FS) flow
StarRC, the full-chip parasitic extraction tool, uses pattern-matching algorithms based on interconnect structures calibrated with field solvers. In cases where higher accuracy is required, such as for clock trees, critical nets, cells or blocks, StarRC interfaces seamlessly with Raphael NXT for 3D capacitance extraction as shown in Figure 4. The StarRC integrated flow with Raphael NXT, called the Field Solver (FS) flow, has two different modes: FSCOMPARE and FS_EXTRACT_NETS.

The FSCOMPARE mode is widely used for the reference validation of full-chip extraction tools. It provides an automated push-button flow for analysis of test structures or selected nets from a chip using Raphael NXT. The results from Raphael NXT and StarRC are then compared to ensure consistency between the extractors.

The FS_EXTRACT_NETS mode is used for the high accuracy extraction of critical circuits to complement StarRC extraction. Within the StarRC environment, users can supply Raphael NXT with a list of nets that need the highest level of capacitance extraction accuracy. StarRC not only extracts the nets as in the regular flow, but also creates a subset of the design based on the user-specified nets to be extracted by Raphael NXT. StarRC merges Raphael NXT results with original StarRC results and outputs a single parasitic netlist.

Figure 3: Raphael NXT’s multicore technology offers linear performance scalability

Figure 4: Capacitance correlation between Raphael NXT and Raphael
Figure 4: StarRC ’s Field Solver flow uses Raphael NXT for high accuracy
3D capacitance extraction and reference validation

Advanced process modeling
When high accuracy is needed, detailed process effects must be included in the interconnect structures analyzed by Raphael NXT. Dummy metal fill, used to even out pattern density in chemical mechanical polishing (CMP) of copper conductors, can be extracted as floating metal without introducing extra nodes in the output SPICE netlist.

In dual-damascene processes, the cross section of the conductors is more trapezoidal than rectangular (Figure 5). The metal etching process in Low-K dielectrics creates damage near the metal-dielectric interface. Chemical mechanical polishing causes the thickness of metal etching to vary as a function of metal density. All of these effects can be modeled in Raphael NXT through the StarRC Field Solver (FS) options. Some of the advanced processes modeling features supported by Raphael NXT are listed below.

Conformal dielectric process support
Support of air gap
Via cap extraction
Support for background dielectric
Support of inter-layer dielectric
Support for co-vertical conductors
Support for nonplanarized metal
Accurate 3D interconnect modeling
Width and spacing-dependent thickness variation
Bottom thickness variation
Density-based thickness variation
Trapezoidal shape support
Dielectric damage modeling
Low-K dielectric, silicon-on-insulator (SOI) modeling

Figure 5: Scaling of solution time
Figure 5: Raphael NXT supports advanced process effects such as trapezoidal
conductor cross sections, Low-K damage and density-based thickness variation

Boundary conditions
Raphael NXT provides the ability to specify boundary conditions, which can be used for specifying the environment around a rectangular region being extracted. By default, it uses a boundary condition of zero volts (electrical ground) at and beyond the edges of the region – also called a Dirichlet boundary condition.

A reflecting boundary condition assumes the region beyond the boundary is a “mirror image” of the region inside it – also called a Neumann boundary condition. Raphael NXT also supports a “periodic” boundary condition, which assumes that the geometry of the region is repetitive in the x or y direction or both. This feature is useful in accurately extracting repetitive structures such as RAM cells.

Inputs Design file Technology file Control file (optional)

Outputs Capacitance report Capacitance matrix SPICE netlist (SPF format) Distributed RC report

Supported Platforms AMD64 (Opteron) 64-bit Sun Solaris 64-bit X86 (IA-32) Linux 32-bit