Periodic Noise Analysis of Circuits with Sharp Transitions

Accurately analyzing the periodic noise of circuits with sharp transitions requires including device noise from a large number of harmonics (a.k.a., sidebands) – typically hundreds and often thousands. Traditional frequency-based periodic noise analysis (pnoise) employ limited-spectrum techniques that typically default to ~50 sidebands—producing approximations that may be wildly off the actual results in such circuits. While it is possible to increase the number of harmonics or sidebands, doing so dramatically increases runtime and leads to severe convergence problems, making accurate analysis impractical or impossible.  

Analog FastSPICE RF (AFS RF) includes the industry’s only pnoise analysis that includes full-spectrum device noise every run with absolutely no tradeoff in accuracy versus runtime. The tool produces these results by employing a time-based technique that directly analyzes the periodic steady state (PSS) and then translates the results to the frequency domain. The resulting pnoise is always nanometer SPICE accurate and >5x faster than frequency-based techniques for circuits with sharp transitions. AFS RF also includes vastly superior PSS convergence, handling circuits with >100K elements.  

As an example, consider the series of pnoise results for a periodic circuit shown below, which compares frequency-based pnoise runs with an increasing number of harmonics to AFS RF time-based pnoise analysis:  

The first three frequency-based pnoise results in the table above (50, 100, and 200 harmonics) are so far off that they do not even appear in the plot. The next three results are color coded to match their plotted results. The frequency-based pnoise results change dramatically with the number of harmonics, monotonically approaching the time-based pnoise results.  

The AFS RF full-spectrum pnoise analysis ignores any specified number of sidebands and always produces identical, accurate results. In this case it does so ~5x faster than even the least accurate frequency-based run and >100x faster than frequency based runs that produce reasonable approximate results. Because AFS RF is full-spectrum, it is never necessary to run the same analysis more than once.  

The minimum number of harmonics required for acceptable accuracy in traditional frequency-based pnoise analysis can be determined by using Fourier analysis. AFS RF includes the Device Noise Analysis Advisor (DNA Advisor) that provides this capability. DNA Advisor analyzes transient or PSS waveforms and recommends the parameters for each type of device noise analysis including, transient noise "noisefmax," time-domain PSS "maxacfreq," harmonic balance simulation harmonics, and frequency-based pnoise sidebands or harmonics. The figure below shows the DNA Advisor GUI.

The table below shows typical DNA Advisor results for minimum number of harmonics required when running frequency-based pnoise for various types of blocks:

As shown in the table, circuits with sharp transitions, such as phase frequency detector (PFD) plus charge pump (CP) require 1700 harmonics and the divider requires 500 harmonics. These blocks need to be simulated with time-based pnoise which only AFS RF provides in order to obtain accurate results in a reasonable timeframe.

See additional Hot Topics here.

For more information, contact your BDA application engineer or click here for a web request.