Abstract: Since its invention, SPICE has been a primary workhorse in verification, primarily because it provides the closest, most accurate prediction of how physical devices are actually going to behave after fabrication. This level of accuracy comes at great expense, though, in terms of simulation performance and capacity. Since SPICE was unable to keep pace with growing design block sizes, the EDA industry created Fast-SPICE, which compromises some accuracy to get additional capacity and performance. Although this trade-off may be acceptable for some purely digital circuits and memories, it may not be acceptable when simulating the more sensitive analog portions of large mixed-signal blocks. Even with the trade-off, many Fast-SPICE tools have been unable to keep pace with growing block sizes and shrinking geometries, so simulation runs of days or weeks are still commonplace. This paper focuses on a new simulation approach that offers more than ten times the performance of other Fast-SPICE engines on the market, combined with SPICE-level accuracy for large mixed-signal blocks.

Click here to receive a complementary copy of this paper.

Abstract: Nascentric has a broad patent portfolio protecting many of the advanced technologies used in the OmegaSim SPICE simulation product line. This paper covers the highlights of some of these technologies, showing how OmegaSim simulators are able to provide the highest performance and accuracy combination of any other Fast-SPICE simulators on the market. In the process, OmegaSim has achieved a number of industry firsts: the first and only current-based SPICE simulator, the first and only SPICE simulator to offer both multi-threading and multi-processing, the first and only hardware-assisted SPICE simulator, and the first and only GPU-based EDA tool.

Click here to receive a complementary copy of this paper.

Hardware-Assisted SPICE Simulation

Abstract: SPICE simulation is particularly expensive in terms of processing power. Accurately evaluating a single BSIM transistor model, for instance, may require several hundreds of complex arithmetic operations. The subtleties of sub-65 nanometer design have only increased transistor model complexity, while millions transistors with millions of extracted parasitics are not uncommon block sizes. The computational resources needed to verify and simulate ICs have struggled to stay apace with IC growth. The Electronic Design Automation industry has a long history of using specialized hardware to accelerate verification. The hardware, however, consists of very expensive custom chips or boards. Even worse, due to Moore ’s law, the systems become obsolete after only a year or two of usage. Graphics Processing Units, or GPUs, offer an attractive alternative to customized verification hardware. Compared to multi-core CPUs, they offer a much greater degree of parallel computational capacity. Systems are commercial and off-the-shelf so the hardware cost is affordable. Due to heavy competition in the commercial graphics chip market, GPU advancements are rapid – closely tracking Moore 's law – and affordable. This paper shows how GPUs are used in OmegaSim GX, making it the first and only hardware-assisted SPICE simulator on the market.

Click here to receive a complementary copy of this paper.

The Role of Extraction in Fast-SPICE Verification and Analysis

Abstract: Designers tend to leverage the additional performance of Fast-SPICE to either perform a larger number of simulation runs to increase coverage or to perform a more detailed analysis of a specific design block. This paper addresses the second of these applications. In order to achieve detailed analyses, designers often dramatically increase the resolution of physical extraction, thereby generating highly detailed netlists with large numbers of parasitics. Paradoxically, this kind of parasitic overload may actually decrease the level of simulation accuracy while largely negating the performance advantage of Fast-SPICE. This paper demonstrates the negative effect of excessive parasitic extraction on simulation. It also provides a better strategy for optimizing Fast-SPICE performance through the intelligent application of parasitic extraction.

Click here to receive a complementary copy of this paper.

Abstract: EDA software is arguably at the pinnacle of high technology. Sadly and ironically, EDA sales can be even more frustrating and low tech than buying a used car. As with car sales, EDA tool list prices are set unreasonably high, but at least the car prices are prominently displayed on the window. EDA prices are usually hidden from the general public, buried within internal price books that are constantly being revised due to internal sales pressure to meet quarterly and yearly revenue goals. When directly asked about prices, EDA salespersons often try to evade the question or obscure the answer. When pressed, they may sheepishly reveal the price, but only after reciting a dizzying array of stipulations, qualifications, and contingencies. Fortunately, the selling and buying of EDA tools does not have to follow these legacy business practices. The purpose of this paper is to show that by reversing the perspective, and examining the purchasing process from the customers’ point of view, it is possible to devise a painless, hassle-free, sales experience more befitting to the high technology software being sold.

Click here to receive a complementary copy of this paper.