Colin Warwick

ST Microelectronics & University of Lyon Predict EMI from Switched-mode Power Converter Using ADS

Blog Post created by Colin Warwick Employee on Jul 25, 2016

Note: A more polished version of this Case Study is available in PDF format as STMicroelectronics and University of Lyon Predict EMI Using ADS.

 

This Case Study highlights work in a paper N-Conductor Passive Circuit Modeling for Power Converter Current Prediction and EMI Aspect by Roberto Mrad, Florent Morel, Cael Pillonnet, Christian Vollaire, Phillppe Lombard, and Angelo Nagari. The high di/dt edges in switched-mode power supplies (SMPSs) can create EMI problems especially if the layout has long power and ground traces. To explore the design space, you need a good model of the impairment you are trying to mitigate. Prior modeling work used lumped elements, but these are not great a modeling traces, which are transmission lines. You'd need an infinite number of infinitesimal LC ladder segments to model a trace exactly. Mrad took a leaf out of the RF/microwave playbook and used network parameters instead.

The Challenge

Conducted electromagnetic emissions from switched-mode power supplies (SMPSs) interferes with the adjacent circuits they are powering. To avoid failure, EMI filtering is added. To design cost effective EMI filtering, it is necessary to explore the design space efficiently. Cut-and-try is too time consuming and so simulation is preferred. But to do simulation, you need accurate models and conventional lumped element models of the traces are not adequate.

The Solution

Traces and vias are essentially delay lines (exp(s) in Laplace space) whereas lumped element L's and C's are integrators/differentiators (s or 1/s in Laplace space). For this reason, lumped element ladders are not efficient at modeling traces. Fortunately network parameter models are readily available and work efficiently in both frequency and time domain simulation. They work directly in AC or harmonic balance (HB) frequency domain circuit analysis or via convolution in the time domain. (Convolution is a way of building a time domain delay line model from frequency domain data using a "smart" inverse fast Fourier transform (IFFT).) The authors used Advanced Design System because it handles network parameters both from measurement instruments and from transmission line models.

The Results

 

The authors obtain excellent correlation with measurement as you can see in Figure 18 of their paper:

figure-18.png

They went on to use the model to design a compact and low cost EMI filter using a sophisticated optimization algorithm, also using ADS. That paper is called Discrete Optimization of EMI Filter Using a Genetic Algorithm but we'll save that Case Study for another day.

 

 

Are you working on switched-mode power converters? Do you face the same challenge? Or something else? Please log in and leave a comment and/or "like" this posting!

Best regards,

-- Colin

PS Here's the link to request an ADS evaluation license if you want to try it.

Outcomes