numerical modeling for heatsink emissions in power ......curs plalce at different frequencies...

Adv. Radio Sci., 10, 239–243, 2012www.adv-radio-sci.net/10/239/2012/doi:10.5194/ars-10-239-2012© Author(s) 2012. CC Attribution 3.0 License.

Advances inRadio Science

Numerical modeling for heatsink emissions in power electronics

J. Kulanayagam1, J. H. Hagmann1, S. Schenke1, K. F. Hoffmann2, and S. Dickmann1

1Institute for Fundamentals of Electrical Engineering, Germany2Institute for Power Electronics Helmut-Schmidt-University/University of the Federal Armed Forces Hamburg, Germany

Correspondence to:J. Kulanayagam ([email protected])

Abstract. The parasitic coupling between power semicon-ductors and the heat sink is responsible for noise currentin Switching Mode Power Supply (SMPS) systems. Inthis paper, the variations in the radiation characteristics ofheatsinks are investigated with respect to their geometries byuse of numerical models. Analyses are facilitated by using amopole antenna as an EMI receiver and by using simplifiedheatsink models as EMI transmitters to model the heatsinkradiated emissions. In addition, the analysis is confirmed us-ing laboratory measurements.

1 Introduction

Power-switching semiconductors (MOSFET’s/IGBT’s) areusually mounted on heat sinks to keep the temperature belowan acceptable limit. Parasitic capacitances can be created be-tween the switching devices and the heatsink. The electricisolator (mica) acts as a dielectric of these capacitances. Theresult of this parasitic effects, the common-mode EMI is gen-erated in Switching Mode Power Supply circuits (Tihanyi,1995). If this current contains a high number of harmoniccomponents and its flows through the heatsink, the heatsinkbecomes an antenna. However, the maximum radiation oc-curs at different frequencies depending on the heatsink ge-ometry (Brench, 1994; Das and Roy, 1998). In consequence,this current could violate the EMC standards due to radiation.Therefore the heat sink should be well designed. A typicalheatsink configuration is shown in Fig.1.

In this paper, the variations in the radiation characteristicsof non-grounded and grounded heatsinks are investigated. Aprevious paper have shown that very little difference in thegeneral electromagnetic behaviors of a finned heatsink anda heatsink as a solid block (Dawson et al., 2001). In con-sequence, a heatsink as a solid block allows a much simplersimulation model.

Manuscript prepared for Adv. Radio Sci.with version 3.2 of the LATEX class copernicus.cls.Date: 24 July 2012

Numerical modeling for heatsink emissions in power electronicsJ. Kulanayagam1, J. H. Hagmann1, S. Schenke1, K. F. Hoffmann2, and S. Dickmann1

1Institute for Fundamentals of Electrical Engineering,2Institute for Power ElectronicsHelmut-Schmidt-University / University of the Federal Armed Forces Hamburg, Germany

Abstract. The parasitic coupling between power semicon-ductors and the heat sink is responsible for noise current inSwitching Mode Power Supply (SMPS) systems. In this pa-per, the variations in the radiation characteristics of heatsinksare investigated with respect to their geometries by use of nu-merical models. Analyses are facilitated by using a mopoleantenna as an EMI receiver and by using simplified heatsinkmodels as an EMI transmitter to model the heatsink radiatedemissions. In addition, the analysis is confirmed using labo-ratory measurements.

1 Introduction

Power-switching semiconductors (MOSFET’s/IGBT’s) areusually mounted on heat sinks to keep the temperature be-low an acceptable limit. Parasitic capacitances can be createdbetween the switching devices and the heatsink. The elec-tric isolator (mica) acts as a dielectric of these capacitances.The result of this parasitic effects, the common-mode EMIis generated in Switching Mode Power Supply circuits (Ti-hanyi, 1995). If this current contains a high number of har-monic components and its flows through the heatsink, whichbecomes an antenna. However, the maximum radiation oc-curs plalce at different frequencies depending on the heatsinkgeometry (Brench, 1994), (Das and Roy, 1998). In conse-quence, this current could violate the EMC standards due toradiation. Therefore, the heat sink should be well designed.A typical heatsink configuration is shown in Fig. 1. In thispaper, the variations in the radiation characteristics of non-grounded and grounded heatsinks are investigated. Previouspapers have shown that very little difference in the generalelectromagnetic behaviors of a finned heatsink and heatsinkas a solid block (Dawson et al., 2001). In consequence,

Correspondence to: J.Kulanayagam([email protected])

PCB

Power device

Heatsink

Mica

FR-4

Fig. 1. Schematic representation of parasitic capacitance

heatsink as a solid block allows a much simpler simulationmodel.

2 Experimental setup

The simplified model shown in Fig. 2 is used for experimen-tal purposes and theoretical simulations. In order to representa ground plane, a rectangular piece of aluminum (2x1.5 m)was used. The ground plane was placed on the wooden ta-ble inside an anechoic chamber. The edges of the groundplane were covered by microwave absorbers to limit the ef-fect of reflections from nearby objects. The block heatsink(88x34x53 mm) was supported by a small dielectric support88x34x5 mm (not included in the numerical model) over aground plane. A MOSFET (TO-247) was placed on theheatsink. The electric isolator was inserted between the caseof the power device and the heatsink. The thickness and di-electric constant of mica are 0.05 mm and 3.5, respectively.In order to create the excitation source the SMA- type con-nector was connected to the pin (drain) of the semiconductor.The access of the connector was connected to the network

Fig. 1. Schematic representation of parasitic capacitance.

2 Experimental setup

In order to represent a ground plane, a rectangular pieceof aluminum (2× 1.5 m) was used. The ground planewas placed on a wooden table inside an anechoic cham-ber. The edges of the ground plane were covered by mi-crowave absorbers to limit the effect of reflections fromnearby objects (Dawson et al., 2001). The heatsink block(88× 34× 53 mm) was supported by a small dielectric block88× 34× 5 mm (not included in the numerical model) overa ground plane. A MOSFET (TO-247) was placed on theheatsink. The electric isolator was inserted between the caseof the power device and the heatsink. The thickness and di-electric constant of Mica are 0.05 mm and 3.5, respectively.In order to create the excitation source the SMA- type con-nector was connected to the pin (drain) of the semiconductor.

Published by Copernicus Publications on behalf of the URSI Landesausschuss in der Bundesrepublik Deutschland e.V.

240 J. Kulanayagam et al.: Numerical modeling for heatsink emissions in power electronics

2 J.Kulanayagam et al.: ...

analyzer via coaxial cable. A 7.5 cm high monopole antenna(1 mm diameter) at 1 m from the centre of the block heatsinksensed the radiated emission of the heatsink for model valida-tion purposes. The network analyzer was used to determinethe transmission from heatsink to monopole (S21) in order tovalidate the model.

Network Analyzer

Monopole Antenna

Anechoic Chamber

Heatsink Setup

Ground Plane

Fig. 2. Measurement setup

3 Numerical modeling

In the numerical model all metallic parts are considered asa Perfect Electric Conductor (PEC) material. The blockheatsink was placed over the infinite ground plane withoutthe small dielectric support. It was assumed that the per-mittivity of the materials remain constant in the consideredfrequency range.

4 Results and discussions

Non-grounded and grounded heatsink models are necessaryto analyze the differing conditions of process of a heatsink.

4.1 Non-grounded heatsink

At first the simplified non-grounded heatsink model (seeFig. 3) is considered. The noise current is generated due tothe parasitic capacitances of the heatsink design. This noisecurrent can be calculated according to the following assump-tion. In the first approach for low frequencies is the capaci-tance between the case of the power device and the heatsinkCdh much greater than the capacitance between the heatsinkand ground plane Chg.The common-mode current does not

id

Ground

Non-grounded heatsink

ic/2 id

Chg

ic/2

Cdh

uds

unoise

Fig. 3. The noise currents in the non-grounded heatsink

take into account in this system. The differential-mode cur-rent can be calculated:

id(t) =Chg ·dunoise

dt. (1)

For low frequencies Cdh >> Chg, the noise voltage can becalculated as follows

unoise 'uds . (2)

The differential-mode current is given by following equation

id(t)'Chg ·duds

dt. (3)

Four Configurations were made for the non-groundedheatsink model. Fig. 4 shows the first configuration of themeasured and simulated coupling (S21) from the heatsink(frontally and laterally) to the monopole antenna for the non-grounded block heatsink. In order to change the parasitic

Fig. 4. The measured and simulated S21 from non-grounded blockheatsink in the frequency range for frontal and lateral measurements

Fig. 2. Experimental setup (heatsink frontally and monopole an-tenna vertically).

The access of the connector was connected to the networkanalyzer via coaxial cable. A 7.5 cm high monopole antenna(1 mm diameter), 1 m from the centre of the heatsink block,sensed the radiated emissions from the heatsink for modelvalidation purposes. The network analyzer was used to de-termine the transmission parameters from the heatsink to themonopole antenna (S21) in order to validate the model. Thesimplified model (without microwave absorbers around thetable) shown in Fig.2 is used for experimental purposes.


In order to compute the electromagnetic field outside theheatsink model in the far field environment, CST MicrowaveStudio was used. In the numerical model all metallic partsare considered as Perfect Electric Conductor (PEC) materi-als. The heatsink block was placed over the infinite groundplane without the small dielectric support. It was assumedthat the permittivity of the materials remained constant in theconsidered frequency range.


Non-grounded and grounded heatsink models are necessaryto analyze the differing process conditions of a heatsink.


The noise current is generated due to the parasitic capaci-tances of the heatsink design. This noise current can be cal-culated according to the following assumption. In the firstapproach for low frequencies, the capacitance between thecase of the power device and the heatsinkCdh is much greaterthan the capacitance between the heatsink and ground plane



Network Analyzer

Monopole Antenna

Anechoic Chamber

Heatsink Setup

Ground Plane








id

Ground


ic/2 id

Chg

ic/2

Cdh

uds

unoise




dt. (1)


unoise 'uds . (2)


id(t)'Chg ·duds

dt. (3)



Fig. 3. Noise currents in the non-grounded heatsink.

Chg. The common-mode current is not taken into accountin this case. Firstly, the simplified non-grounded heatsinkmodel (see Fig.3) is considered.

The differential-mode current can be calculated:

id(t) = Chg·dunoise

dt. (1)

For low frequenciesCdh >> Chg, the noise voltage can becalculated as follows

unoise' uds . (2)


id(t) ' Chg·duds

dt. (3)

Four configurations were made for the non-groundedheatsink block model. Firstly, the heatsink block was sup-ported by a small dielectric block with the relative permit-tivity εr = 1 over a ground plane. A MOSFET was placedon the heatsink. The electric isolator (Mica) was insertedbetween the case of the MOSFET and the heatsink. Fig.4shows theS21 parameter for the first configuration of themeasured and simulated coupling (S21) for the frontally andlaterally polarized heatsink. In order to compare the radi-ated emissions of the heatsink, the measured result of thefrontally polarized heatsink was used as the reference curvefor the non-grounded and grounded heatsink.

In order to change the parasitic capacitance between theheatsink and the power device, the thickness of the Mica wasdoubled in the second configuration. The results of the sec-ond configuration are shown in Fig.5. As expected, the re-sults of the first and second configuration are the same. The

Adv. Radio Sci., 10, 239–243, 2012 www.adv-radio-sci.net/10/239/2012/

J. Kulanayagam et al.: Numerical modeling for heatsink emissions in power electronics 241



Network Analyzer

Monopole Antenna

Anechoic Chamber

Heatsink Setup

Ground Plane








id

Ground


ic/2 id

Chg

ic/2

Cdh

uds

unoise




dt. (1)


unoise 'uds . (2)


id(t)'Chg ·duds

dt. (3)



Fig. 4. Comparison of measured and simulated results ofS21 for thefrontally and laterally polarized non-grounded heatsink with singleMica.

J.Kulanayagam et al.: ... 3

capacitance between the heatsink and the power device, thethickness of the mica was doubled. The results of the sec-ond configuration are shown in Fig. 5. As expected the re-sults of the first and second configuration are equal. The

Fig. 5. The measured and simulated S21 from non-grounded blockheatsink with an extra electric isolator in the frequency range forfrontal measurement

block heatsink was supported by a Polyvinyl Chloride (PVC)block over the ground to change the capacitance between theheatsink and ground plane. The measured and simulated re-sults of third configuration are shown in Fig. 6. The radiatedemission has reduced in the frequency range due to the re-duction of the impedance between the heatsink and groundplane.

Fig. 6. The measured and simulated S21 from non-grounded blockheatsink in the frequency range for frontal measurement-supportedby PVC

Furthermore, by adding a ferrite core around the metalliccylinder of the heatsink (see Fig. 7) the input impedance ofthe heatsink can be increased (Kulanayagam et al., 2011).In consequence, the noise current can be minimized, whichresults in a reduced radiation of the heatsink (see Fig. 8).

MOSFET

Heatsink

PCB

Mica

Ferrite core

FR-4

Fig. 7. Schematic representation of filter design

Fig. 8. The measured and simulated S21 from non-grounded blockheatsink with the ferrite core in the frequency range for frontal mea-surement

4.2 Grounded heatsink

Finally the simplified grounded heatsink model (see Fig. 9)is considered. The differential-mode current is given by

id(t) =Cdh ·duds

dt. (4)

The noise voltage can be calculated as follows

unoise ' 0 . (5)

Fig. 10 compares the simulated and measured S21 when theheatsink is grounded at both sides of the heatsink with a cop-per tabs (8x1x5 mm). At frequencies below 500 MHz, the

Fig. 5. Comparison of measured and simulated results ofS21 forthe frontally polarized non-grounded heatsink with two Micas.

Fig. 4 and the Fig.5 show nearly 8 dB deviation in the fre-quency range 500-580 MHz between measurement and sim-ulation. In a real situation, there are always some reflectionsdue to objects inside the anechoic chamber.

The difference between the first and the third configurationis the heatsink block was supported by a Polyvinyl Chloride(PVC) block over the ground to change the capacitance be-tween the heatsink and ground plane. The measured and sim-ulated results of the third configuration are shown in Fig.6.The Fig. 6 shows nearly 5 dB deviation in the frequencyrange 800-900 MHz between measurement and simulation.The radiated emission was reduced in the frequency rangedue to the reduction of the impedance between the heatsinkand ground plane. In consequence, the radiated emissionswere now less than the reference model.







MOSFET

Heatsink

PCB

Mica

Ferrite core

FR-4





id(t) =Cdh ·duds

dt. (4)


unoise ' 0 . (5)


Fig. 6. Comparison of measured and simulated results ofS21 forthe frontally polarized non-grounded heatsink supported by PVC.







MOSFET

Heatsink

PCB

Mica

Ferrite core

FR-4





id(t) =Cdh ·duds

dt. (4)


unoise ' 0 . (5)


Fig. 7. Schematic representation of the filter design.

Furthermore, by adding a ferrite core around the metalliccylinder of the heatsink (see Fig.7) the input impedance ofthe heatsink can be increased (Kulanayagam et al., 2011). Inconsequence, the noise current can be minimized, which re-sults in reduced radiation from the heatsink.In the fourth configuration two different ferrite cores wereused, which consisted of different material characteristics(see Fig.8). The radiated emissions were reduced by addingthe ferrite core around the current path between the MOSFETand the heatsink.

www.adv-radio-sci.net/10/239/2012/ Adv. Radio Sci., 10, 239–243, 2012

242 J. Kulanayagam et al.: Numerical modeling for heatsink emissions in power electronics







MOSFET

Heatsink

PCB

Mica

Ferrite core

FR-4





id(t) =Cdh ·duds

dt. (4)


unoise ' 0 . (5)


Fig. 8. Measured results ofS21 for the frontally polarized non-grounded heatsink for two different ferrite cores.


Ground

id

idic/2

Cdh

uds

unoise

ic/2

Grounded heatsink

Fig. 9. The noise currents in the grounded heatsink

grounded heatsink is effective. The grounding wires havea high impedance in the high frequency range. In conse-quence, above frequencies of 500 MHz, the radiated emis-sions are higher than with the reference setup. The choice ofthe grounding wire plays a big role in SMPS circuits. It hasbeen also observed that the measured noise floor is higher un-til frequencies 200 MHz for the non-grounded and groundedheatsink design.

Fig. 10. The measured and simulated S21 from grounded blockheatsink in the frequency range for frontal measurements

5 Conclusions

In this paper, the variations in the radiation emissions ofheatsinks are investigated with respect to their geometries byuse of numerical model and laboratory measurements. Thenon-grounded and the grounded heatsink models were con-sidered. For the non-grounded heatsink, the radiated emis-sions depend on the dielectric material between the heatsinkand the ground plane. By adding the ferrite core between

the heatsink and the power device, the radiated emission ofthe non-grounded heatsink can be minimized. In order to re-duce the radiated emission, the heatsink should be grounded.The radiated emissions of the grounded heatsink depend onthe grounding wires, which are connected from the heatsinkto the ground plane. The grounding wire should have smallimpedance. Otherwise its becomes an antenna. A good cor-relation between the numerical model and measurements hasbeen obtained. Again, our proposed model gives the radiatedemission of the heatsink.

References

Brench, C.: Heatsink radiation as a function of geometry, in: IEEEInternational Symposium on Electromagnetic Compatibility, pp.105–109, 1994.

Das, S. and Roy, T.: An investigation on radiated emissions fromheatsinks, in: IEEE International Symposium on Electromag-netic Compatibility, vol. 2, pp. 784–789, 1998.

Dawson, J., Marvin, A., Porter, S., Nothofer, A., Will, J., and Hop-kins, S.: The effect of grounding on radiated emissions fromheatsinks, in: IEEE International Symposium on Electromag-netic Compatibility, vol. 2, pp. 1248–1252, 2001.

Kulanayagam, J., Hagmann, J. H., Hoffmann, K. F., and Dickmann,S.: Reduction of heat sink common-mode currents in switchingmode power supply circuits, Advances in Radio Science, 9, 317–321, 2011.

Tihanyi, L.: Electromagnetic Compatibility in Power Electronics, J.K. Eckert & Company, Inc., Sarasota, Florida, 1995.

Fig. 9. Noise currents in the grounded heatsink.


Finally, the simplified grounded heatsink model (see Fig.9)is considered. The differential-mode current is given by

id(t) = Cdh·duds

dt. (4)


unoise' 0 . (5)

Fig. 10 compares the simulated and measuredS21 when theheatsink is grounded at both sides of the heatsink with cop-per tabs (8x1x5 mm). At frequencies below 600 MHz, the


Ground

id

idic/2

Cdh

uds

unoise

ic/2

Grounded heatsink

Fig. 9. The noise currents in the grounded heatsink

grounded heatsink is effective. The grounding wires havea high impedance in the high frequency range. In conse-quence, above frequencies of 500 MHz, the radiated emis-sions are higher than with the reference setup. The choice ofthe grounding wire plays a big role in SMPS circuits. It hasbeen also observed that the measured noise floor is higher un-til frequencies 200 MHz for the non-grounded and groundedheatsink design.

Fig. 10. The measured and simulated S21 from grounded blockheatsink in the frequency range for frontal measurements

5 Conclusions

In this paper, the variations in the radiation emissions ofheatsinks are investigated with respect to their geometries byuse of numerical model and laboratory measurements. Thenon-grounded and the grounded heatsink models were con-sidered. For the non-grounded heatsink, the radiated emis-sions depend on the dielectric material between the heatsinkand the ground plane. By adding the ferrite core between

the heatsink and the power device, the radiated emission ofthe non-grounded heatsink can be minimized. In order to re-duce the radiated emission, the heatsink should be grounded.The radiated emissions of the grounded heatsink depend onthe grounding wires, which are connected from the heatsinkto the ground plane. The grounding wire should have smallimpedance. Otherwise its becomes an antenna. A good cor-relation between the numerical model and measurements hasbeen obtained. Again, our proposed model gives the radiatedemission of the heatsink.

References

Brench, C.: Heatsink radiation as a function of geometry, in: IEEEInternational Symposium on Electromagnetic Compatibility, pp.105–109, 1994.

Das, S. and Roy, T.: An investigation on radiated emissions fromheatsinks, in: IEEE International Symposium on Electromag-netic Compatibility, vol. 2, pp. 784–789, 1998.

Dawson, J., Marvin, A., Porter, S., Nothofer, A., Will, J., and Hop-kins, S.: The effect of grounding on radiated emissions fromheatsinks, in: IEEE International Symposium on Electromag-netic Compatibility, vol. 2, pp. 1248–1252, 2001.

Kulanayagam, J., Hagmann, J. H., Hoffmann, K. F., and Dickmann,S.: Reduction of heat sink common-mode currents in switchingmode power supply circuits, Advances in Radio Science, 9, 317–321, 2011.


Fig. 10. Comparison of measured and simulated results ofS21 forthe frontally polarized grounded heatsink with a single Mica.

grounded heatsink is effective. The grounding wires havea high impedance in the high frequency range. In conse-quence, above frequencies of 600 MHz, the radiated emis-sions are higher than with the reference setup. The choiceof the grounding wire is very important in SMPS circuits.It has also been observed that the measured noise floor ishigher up to a frequency of 200 MHz for the non-groundedand grounded heatsink design.

5 Conclusions

The main objective of this work was to investigate the radi-ated EMI of the heatsink. The work is based on numericalmodeling and the experimental EMI measurements for thenon-grounded and the grounded heatsink.The results of the numerical modeling and the experimen-tal EMI measurements were compared for the non-groundedheatsink with a small dielectric support, varying thicknessof the dielectric material layer between the power deviceand the non-grounded heatsink, the non-grounded heatsinkwith PVC support and the grounded heatsink with a smalldielectric support. The results of the experimental EMI mea-surements were also presented for the adding of the fer-rite core between the power device and the non-groundedheatsink. The capacitance between the heatsink and theground plane is responsible for the EMI radiation of thenon-grounded heatsink, while the capacitance between theheatsink and power semiconductor is responsible for thegrounded heatsink. By adding the ferrite core betweenthe extended heatsink and the power device, the radiatedemissions from the non-grounded heatsink can be mini-mized. In order to reduce the radiated emissions, theheatsink should be grounded. The radiated emissions fromthe grounded heatsink depend on the grounding wires, which

Adv. Radio Sci., 10, 239–243, 2012 www.adv-radio-sci.net/10/239/2012/

J. Kulanayagam et al.: Numerical modeling for heatsink emissions in power electronics 243

are connected from the heatsink to the ground plane. Thegrounding wire should have minimal impedance, otherwiseit can become an efficient antenna. A good correlation be-tween the numerical model and measurements was obtained.

References

Brench, C.: Heatsink radiation as a function of geometry, in:IEEE International Symposium on Electromagnetic Compatibil-ity, 105–109, 1994.

Das, S. and Roy, T.: An investigation on radiated emissions fromheatsinks, in: IEEE International Symposium on Electromag-netic Compatibility, 2, 784–789, 1998.

Dawson, J., Marvin, A., Porter, S., Nothofer, A., Will, J., and Hop-kins, S.: The effect of grounding on radiated emissions fromheatsinks, in: IEEE International Symposium on Electromag-netic Compatibility, 2, 1248–1252, 2001.

Kulanayagam, J., Hagmann, J. H., Hoffmann, K. F., and Dickmann,S.: Reduction of heat sink common-mode currents in switch-ing mode power supply circuits, Adv. Radio Sci., 9, 317–321,doi:10.5194/ars-9-317-2011, 2011.


www.adv-radio-sci.net/10/239/2012/ Adv. Radio Sci., 10, 239–243, 2012

numerical modeling for heatsink emissions in power ......curs plalce at different frequencies...

Documents