Complex-Block Characterization includes computationally intensive verification tasks from pre- and post-layout simulation to variation, noise, and RF analysis on circuits such as PLLs, DLLs, ADCs, SerDes, Tx chains, Rx chains, and memory interfaces.

Many complex blocks are RF in nature and most naturally lend themselves to periodic analysis. RF designers run into accuracy, performance, and capacity limitations with traditional RF simulators (Shooting-Newton and harmonic balance) that are similar to the traditional SPICE limitations.

Traditional RF simulator accuracy, performance, and memory consumption depend upon the user-selected number of sidebands or harmonics. This is not a problem for simple blocks where the tools’ defaults apply, but it quickly becomes an issue for larger, nonlinear complex blocks (e.g., those with dividers). In these cases traditional RF simulator runtime and memory consumption can increase quadratically with the number of sidebands/harmonics needed – assuming the tool can even get periodic steady state (PSS) convergence. Designers are left simplifying their circuit and using a multi-pass methodology to ensure convergence, “good enough” accuracy, and reasonable runtime on computer with enough memory.

Berkeley Design Automation RF FastSPICE (RFS) uses a unique proprietary periodic analysis engine that does not tradeoff accuracy and performance. RFS has vastly superior PSS convergence to traditional RF simulators and delivers the equivalent of “infinite” sidebands/harmonics every run. As a result, designers can run their original circuit without simplification and utilize a 1-pass methodology that always delivers full accuracy results.

Complex-Block Periodic Analysis Examples

The table above contains a range of RFS comparisons with traditional RF simulators. The dual-conversion receiver and transmit modulator with LO generator did not converge in traditional RF simulation. The dual-conversion receiver contained 2 dividers. The designer in that case was able to get their traditional RF simulator to converge and complete RF analysis by removing one divider at a time. This typifies the types of simplifications, workarounds, and shortcuts that traditional RF simulation tools require. RFS completed the two circuits – something that no other tool had accomplished – in just 5.3 hours and <12 hours respectively.

Traditional RF simulation did converge on the 2.3 GHz receiver and the LC-tank VCO; however, RFS was 5.5x and 10x faster on each. This is the kind of savings that make an enormous difference on RF circuits in which designers make numerous iterations to optimize the circuit. Knowing that the results are always full accuracy is just as important. The last circuit is a relatively straightforward ring-oscillator VCO. Again, traditional RF failed to converge; RFS completed PSS and oscillator phase noise within 0.5 hrs.

It should be noted that RFS is the only simulator that produced impulse sensitivity function (ISF) information during oscillator phase noise, and it does so without any simulation overhead. Traditional RF simulators produce contribution information for every node in the circuit during oscillator phase noise. This is helpful for identifying those nodes that generate the most noise, but it can be very misleading because it says nothing about how sensitive the VCO output phase noise is to contribution from a given node. RFS provides contribution, sensitivity (ISF), and the product for every node in the circuit and shows how all three values vary over the period. This provides VCO designers all of the analysis information they need to reliably optimize their VCOs.