Filter Optimisation For High Power Converters
An Example Of Modelling Including Enclosure Parasitic Effects
EMC Filter Simulation
A ‘bad’ filter based on a topology used in an automotive onboard charger was simulated.
Including Enclosure Parasitic Effects
The model included parasitic effects of the PCB and enclosure using FEA in Ansys Q3D.
Validation Against Measurement
The model was validated against measurement on a prototype.
Filter Circuit Diagram
Circuit diagram of the filter topology to be simulated
Component Parasitic Effects Included
The X and Y capacitors were modelled using a series R-L-C model to include parasitic inductance and ESR.
The model parameters were adjusted to match manufacturer data.
Common-Mode Choke Model (Part 1)
Frequency Dependence In Common-Mode
The frequency dependence of inductance and resistance of the common-mode choke in the common-mode was modelled using complex permeability data for the ferrite material.
Differential mode attenuation was not modelled in
this piece of work.
Common-Mode Choke Model (Part 2)
CM Choke Model Using S-Parameters
Four-port S-parameters were derived and formatted into a Touchstone file to give a means of importing frequency-dependent CM choke impedance into the simulator.
Common-Mode Choke Model (Part 3)
CM Choke Model
The CM choke was measured in common-mode using a network analyser then the model inductance was adjusted to give an acceptable
match between measured and modelled CM choke common-mode impedance.
PCB Layout And Export To Siwave
The PCB was laid out in Mentor Xpedition Layout and exported to ODB++.
PCB Import Into Siwave
The ODB++ file was imported into Ansys SIwave and configured with contact points for use when exported to Ansys Q3D for integration with the enclosure model. PCB FEM simulation is possible in SIwave, however, Q3D was used here to allow this integration with the enclosure model.
PCB Export To Q3D
The PCB was then exported to Q3D for integration with an enclosure model.
Enclosure In Q3D
An ‘off-the-shelf’ enclosure was selected and modelled in Ansys Q3D. Representations of brass stand-offs were added.
Why This Odd Arrangement?
An enclosure with long stand-offs was chosen to accentuate the problems inherent with using conduction via stand-offs and the enclosure as a current path in a filter rather than, for example, a good continuous ground plane on the PCB.
Combined Enclosure And PCB Model
Incorporation Of PCB Into Enclosure
The PCB model exported from SIwave into Q3D was positioned inside the enclosure, making contact with the stand-offs.
Connections For Top-Level Model
‘Sources’ and ‘sinks’ were configured on each conductor for connection of the lumped models of capacitors, CM chokes, source and load.
Simulation And S-Parameter Export
A frequency sweep was run, resulting in a simulation of parasitic resistance, inductance and capacitance inherent in the combined enclosure, PCB and stand-offs. This data was exported as a
32-port S-parameter file.
Top-Level Circuit Model
Incorporation Of All Model Components
The lumped component capacitor models, S-parameter models of common-mode chokes, S-parameter model of combined PCB, enclosure and stand-offs and a lumped model of 50Ω source and load were combined into a top-level circuit model.
Prototype With Measurement Connections
A prototype was built. BNC connectors were mounted in copper brackets for connection of the filter in common mode, i.e. with both live and neutral both connected to the centre pin of the corresponding BNC connector, using the enclosure as the ‘ground’.
Why Not Use PCB Mounted BNCs?
The intention was to inject noise in a manner similar to the real application, via wiring carrying the noise currents, not by an intimately connected PCB mounted BNC connector.
A signal generator and spectrum analyser controlled by EMC32 were used to measure the filter attenuation versus frequency. Short coaxial cables were used and normalisation was performed as part of the measurement to compensate for cabling and other signal path attenuation.
Measurement vs Simulation (Part 1)
Simulations were performed using both the full model (green trace) and a model with the enclosure, PCB and stand-off model replaced with ideal connections (blue trace).
Ideal vs Full Simulation
The ideal and full simulations match identically up to ~400kHz, suggesting that enclosure, PCB and stand-off parasitic effects are not significant below this frequency. Above 400kHz, the effect of the enclosure model becomes apparent.
Measurement vs Simulation (Part 2)
Full Simulation vs Measurement
The match between the full model and simulation is satisfactory up to ~300kHz.
Between 300kHz and 2MHz there is an anomaly which is probably explained by insufficient measurement resolution in the experimental setup.
From 2MHz upwards, the model appears to capture the general trend of reduced attenuation, given the presence of enclosure, PCB and stand off parasitic effects.
Measurement Vs Simulation (Part 3)
For approximate comparison a rectangular loop of 40mm x 40mm has an inductance in the region of 100nH.
These dimensions roughly represent the loop formed by the enclosure, the PCB and a pair of stand-offs forming part of the poor grounding
structure in this design.
At 400kz a 100nH inductor has a reactance of ~0.25Ω, i.e. its impedance is starting to become significant compared to a good ground plane.
Filter Simulation Including Enclosure
An outline of a method used to simulate a filter including its enclosure has been given.
Match Between Simulation And Measurement
The match between simulation and measurement is reasonable but requires further investigation has given the differences observed and crude measurement method.
This baseline model can be used to rapidly optimise the filter design in a simulation environment.
Graham Roberts MEng (Hons) PhD CEng INCE MIET – TURNTIDE – EMC and Compliance International Presentation 22nd March 2022.