Interaction Notes. Note October 2014

Transcription

1 Interacton Notes Note 6625 October 2014 APPLICATION OF ELECTROMAGNETIC TOPOLOGY AND POWER BALANCE CONCEPTS TO RADIO FREQUENCY COUPLINGS INTO BUILDINGS I. Junqua J.P. Parmanter - F. Issac - W. Quenum DEMR/CEM Onera The French Aerospace Lab, F-31055, Toulouse, France Abstract Ths paper focuses on the expermental and modellng work led n the context of the HIPOW European project that deals wth the protecton of crtcal nfrastructures aganst IEMI/NNEMP attacks. The objectve s to evaluate and analyze Rado Frequency transfer functons between dfferent areas of a buldng and the exteror of a buldng n order to assess EM envronment n potentally crtcal rooms submtted to an external EM threat. At a frst step, ths analyss s purely based on experments, whch mples to develop an optmzed expermental set-up. At a second step, Power Balance concepts ntally developed for hgh-q overszed systems have been evaluated n the case of Rado Frequency couplng n and through buldng rooms..

2 TABLE OF CONTENTS 1 Introducton Analyss of em nteracton through and nsde a buldng from expermental data Test Ste Test Set-Up Test confguratons Test results and man conclusons Intra measurements Inter measurements Ext/Int measurements Applcaton of EMT/PWB methodology Recalls on EMT/PWB methodology General concepts Topologcal network models EM resoluton of topologcal models Valdaton of EMT/PWB methodology Topologcal network model of the 1 st floor of the targeted buldng Basc confguraton: EM envronment modellng n the targeted offce More complex confguratons: sheldng effectveness n the laboratory room EMT/PWB modellng of addtonal confguratons Effcency of metallc roller shades for EM protecton 20 4 Concluson References Acknowledgements: Ths work has been founded by the European project HIPOW under the grant agreement FP7-SEC

3 1 INTRODUCTION A crtcal nfrastructure s a common term accepted by worldwde natons and governments whch refer to any asset essental for the socety and ts economy ncludng also health and securty. Consequently, protectng crtcal nfrastructures aganst any type of attack lkely to jeopardze ther normal operaton has become a major concern addressed at the hghest levels of all country organzatons. Among the large set of possble attacks (explosve, chemcal, bologcal, cyber-attack, armed attack ), the IEMI/NNEMP attack, as defned as the use of a rado frequency wave to nduce possble dsrupton or destructon of electroncs, s clearly dentfed and must be consdered n a rsk management process ([1], [2]). Ths s the reason why, for a few years, several projects n Europe and n the world have addressed ths specfc topc and ts consequences on crtcal nfrastructures. The work presented here has been carred out n the context of a European project named HIPOW [3] and funded by European Commsson (EC) under the grant agreement FP7-SEC HIPOW stands for protecton of crtcal nfrastructures aganst hgh power mcrowave threats as mentoned n the descrpton of work and ams at: - Conductng a threat analyss and rsk assessment of an IEMI/NNEMP attack onto crtcal nfrastructures, - Evaluatng the effcency of current protectons, - Investgatng the feasblty of hardenng measures, - Developng detecton and dagnoss systems, - Proposng gudelnes and nputs to standards for protecton. In ths work, we consder a buldng as an nfrastructure hostng crtcal devces operatng crtcal functons, as an hosptal or a bank. Fgure 1 shows such a buldng wth an IEMI/NNEMP attack carred out at dstance from a truck. The rsk analyss from ths EM threat wll be then derved from: - The crtcal functons ncluded n the buldng, each of them havng elementary levels of IEMI/NNEMP vulnerablty, - Ther dstrbuton, locaton and nstallaton nsde the buldng, - The nterdependences between crtcal functons, - The EM characterstcs of the buldng n terms of Electromagnetc (EM) attenuaton between the exteror and the nteror and between dfferent areas of the buldng, - The attack scenaro snce the effect onto the nternal functons of the buldng may sgnfcantly vary wth the mpact generated by the threat. Accordng to the defnton of IEMI/NNEMP (UWB and narrow band waveforms) gven n IEC standard , the frequency band of analyss consdered n ths work ranges from a few hundreds of MHz up to a few GHz (typcally 300MHz-3GHz). Fgure 1 IEMI/NNEMP attack of a buldng from outsde 2

4 The buldng beng seen as a passve envelope made of walls and contanng several electrcal/electronc functons; ths envelope has a strong nfluence on how the ncdent EM threat generated from the outsde wll enter the buldng and how t wll propagate nsde the buldng rooms up to crtcal equpment. Several EM couplng phenomena are nvolved n ths process together wth the topology of the buldng: the attenuaton of walls, the penetraton through apertures and wndows, the propagaton along the conducts and electrcal wres. All those EM phenomena contrbute to the buldng attenuaton or sheldng effectveness whch relates the ncdent EM level produced by the EM source and the resultng EM level at crtcal test-ponts n the buldng. The knowledge of ths nformaton thereby becomes a key-parameter to evaluate possble effects of an IEMI/NNEMP on crtcal devces of a buldng and to derve the most approprate protecton measures. The scope of ths work s delberately restrcted to the nvestgaton of nduced EM radated observables n the hgh frequency range [over 1GHz] and does not consder any conducted observables on cable networks. In ths context, the focus of ths paper wll be on the analyss of EM nteractons through and nsde a buldng n order to propose a methodology to estmate ths relevant transfer functons from whch EM levels nsde buldngrooms nduced by specfc external or nternal RF EM sources wll be deduced. For ths, we propose to apply the Electromagnetc Topology concepts and Power Balance concepts (EMT/PWB), even f those concepts have been ntally developed for electrcally large and hgh-q-factor (Q=Qualty) systems ([4], [5]) whch s not the general case of buldngs. Indeed ths methodology has been largely valdated for modelng and assessng EM couplng of a Hgh Intensty Radated Felds (HIRF) envronment onto an arcraft/rotorcraft [4]. In other words, the objectve s to see how the applcaton doman of the EMT/PWB methodology can be extended to the evaluaton of EM couplng n a low-q factor system as ths s the case n buldngs. Secton 2 of ths paper deals wth experments conducted n the authors premses n order to acqure a frst set of data and to analyze EM nteractons n ther buldng. For experments on such large scale systems, we wll show how the test set-up needs to be optmzed. Secton 3 s dedcated to the EMT/PWB approach: Frstly, some recalls on the methodology are gven. Then, the potentalty of the approach s evaluated. Fnally, some conclusons and prospects are gven about future applcatons of the EMT/PWB method for EM RF nteractons n buldngs. 3

5 2 ANALYSIS OF EM INTERACTION THROUGH AND INSIDE A BUILDING FROM EXPERIMENTAL DATA At a frst step, experments have been conducted n ONERA premses n order: - To analyse RF EM envronment and RF EM nteractons n buldngs and to study to whch extend they can be represented by pseudo random varables as n perfectly mode strrng reverberatng chambers (MSRCs). - To dentfy potental EM areas of confnement of EM energy nsde buldng rooms and to evaluate the relevance of EM topology concepts based on volume decomposton of EM problems. 2.1 Test Ste Such transfer-functon tests requre realstc nstallaton features n order to represent correctly the topology of most common crtcal buldng nfrastructures as: - Several rooms of several sze and shapes - Fully equpped buldng wth wndows, wndow shades, doors - Fully electrcal and water ppes nstallaton Transfer functon measurements are low level measurements and do not requre any safety management as real hgh level tests. The ncdent power at the transmttng antenna nput s low enough to generate EM felds much below the safety levels gven n IEEE standards wth respect to human exposure to Rado Frequency EM felds. However obtanng agreements to perform them s generally tedous. Ths s why, for flexblty and safety reasons, we have chosen to carry-out those tests at our ONERA premses located n Toulouse, France. The ste s a cvl mult-buldng ste wth buldngs made of usual materals, ncludng doors and wndows and wthout any specfc EM protecton (Fgure 3). In addton, ths ste hosts several laboratory rooms and one buldng asle hosts the EM and Radar Department (DEMR) whch means that EM tests are not an ssue for ths buldng. Ths buldng contans three floors levels. It has been chosen as our targeted buldng. Besdes, snce ths buldng s made of several nterconnected asle-buldngs, t s possble to make the llumnaton from one of these asle buldngs onto the DEMR buldng. Such a soluton offers the possblty to keep usng the Ethernet network of the ste for transmsson of data and commands of measurement equpment. Consequently, the Spatal Envronment (DESP) buldng has been chosen as the llumnaton buldng (see Fgure 2). In addton, Fgure 3 presents several vews of the neghborng of the targeted buldng and from ts nternal nstallaton. Fgure 2 - ONERA s ste n Toulouse wth the DEMR asle-buldng (targeted buldng) and the DESP asle buldng (from whch the EM external llumnaton s made) Credts: Google Maps 4

6 Vew of the llumnaton buldng from the targeted buldng Typcal offce room Buldng corrdor Fgure 3 Targeted asle buldng vews 2.2 Test Set-Up In the approach, the natve expermental data that characterze EM nteractons nsde a system are transfer functons between a receved power on a recevng antenna and the ncdent power njected n a transmttng antenna for a gven geometry and locaton of both antennas and for gven frequences whch mples that only lnear EM mechansms are taken nto account. These transfer functons have been drectly measured wth a network analyzer. Nevertheless, such measurements generally requre connecton of the transmttng and recevng antennas to the ports of the network analyzer usng ether long measurement cables or optc-fber lnks. In case of a mult-room buldng, such measurement confguraton s not possble from the practcal pont of vew of the avalablty of long-enough measurement cables and deployment of those cables wthout generatng any trouble n the normal operaton of the buldng. Ths s the reason why, the measurement setup has been optmzed accordng to the expermental protocol descrbed below: - - On the transmttng sde, the measurement equpment ncludes a frequency syntheszer and the transmttng antenna. On the recevng sde, the measurement equpment ncludes a spectrum analyser and the recevng antenna. Both transmttng and recevng equpment (syntheszer, spectrum analyser, motor) are drven by a software program through a command PC. Instructons are carred va the bult-n nternal Ethernet network of the ste. At each sde of the lnk, both recevng and transmttng equpment are placed onto a trolley n order to allow easy dsplacement nsde the buldng areas. 5

7 - A strrer s nstalled n the room hostng the transmttng equpment n order to generate a pseudoreverberatng chamber envronment. The strrer s drven by a motor enablng a step-by-step rotaton. Therefore, the transfer functon between the recepton pont and the transmttng pont could be measured on a frequency band from 300MHz up 6.3GHz for each poston of the strrer. - In order to renforce the strrng effect n the buldng rooms, the four polarsaton combnatons of transmttng and recevng antennas are consdered: HH, VV, HV, VH (V=vertcal; H=Horzontal). - The whole set of expermental data has fnally been post-processed for a gven room-locaton of transmttng and emttng antennas n order to extract hstograms and statstcal parameters over a complete rotaton of the strrer and for all combnatons of antenna polarsatons. A complete overvew of the test set-up s llustrated n Fgure 4. Antennas are classcal double rdged gude antennas 3119 (from Ets Lndgren). The strrer s a home-made strrer. Fgure 4 - Test set-up for transfer functon measurements n a buldng 2.3 Test confguratons Three man confguratons wth dfferent objectves have been consdered: - Intra confguraton: the transmttng and the recevng antennas are located n the same buldng room. The measured expermental data provde assessment of equvalent losses n the room. Accordng to the Power Balance Method, they can be expressed n terms of mean Q factors or so-called equvalent couplng-cross-sectons (CCS) [4] (or equvalent areas) whch fully characterze EM dsspatve effects n a sngle room (see paragraph 3.1). - Inter confguraton: the transmttng and the recevng antennas are located n two dfferent rooms of the same buldng. These measurements of transfer functons combned to prevous Intra confguratons gve the equvalent couplng cross secton of the propagaton path between the two rooms. - Ext/Int confguraton: one of the antennas s outsde the buldng and the other one s nsde a buldng-room. For safety reasons and because of the recprocty of the EM problem, the transmttng antenna has been postoned n the targeted room and the recevng antenna nsde a room of the llumnaton buldng. Ths soluton allows mantanng the emsson wthn the targeted buldng and to keep usng the Ethernet Network for data and command transfers. Both antennas are facng each other n front of two wndows of the two buldngs at a dstance of 20m. Of course, the post-processng and measurements are analysed as f the transmttng antenna was n the llumnaton buldng and the recevng antenna was nsde the targeted buldng (as n normal NNEMP/IEMI attack condtons). Fgure 5 llustrates the test setup n the room hostng equpment for emsson. 6

8 Fgure 5 - Transmttng equpment n buldng room at emsson sde In Fgure 6, the red-coloured wndow ndcates the wndow at whch the recevng antenna s placed (second floor of the llumnaton buldng) n confguraton Ext/Int. The pcture s taken from the wndow of the targeted buldng room at the frst floor hostng the transmttng system (agan usng the property of recprocty of the problem). Fgure 6 - Confguraton Ext/Int Vew from the targeted room under test 2.4 Test results and man conclusons Intra measurements Fgure 7 llustrates a typcal set of results obtaned n an offce room of 50m 3, n Intra confguraton (we remnd that the objectve s to characterze EM envronment n a room n terms of statstcal dstrbuton of the transfer functon). In ths confguraton, the EM envronment s expressed as the S 21 parameter between the transmttng and the recevng antennas, both placed n the same room for each poston of the strrer and for each combnaton of polarzaton of the antennas. Fgure 7 thereby presents the cumulatve hstogram of S 21 measurements at a frequency equal to 2GHz. As seen n Fgure 7 (but ths result s also confrmed by all measurements), EM envronment n 7

9 buldng rooms cannot be characterzed as n mode strrng chambers wth Raylegh probablty functons because of ther low Q factor (or hgh losses features). However a Weybull law can be a good approxmaton[6]. Nevertheless, Weybull laws are a 2-parameters functons whch can be deduced from the average and standard devaton of the set of measurements. At ths stage of the work and snce the measurements have been carred out at ONERA s premses only, t s not possble to fnd general laws correlatng these 2 parameters wth typcal characterstc features of buldngs (such as materals, volumes, ). Fgure 7 - Example of cumulatve dstrbuton obtaned from Intra confguraton measurements n a typcal room of ONERA s buldng At a second step, the S 21 results obtaned n Intra confguraton n several typcal rooms of the targeted buldng have been post-processed n terms of mean Q factor or n terms of mean CCS as n the Power Balance approach as follows ([4], [5]):. 2 16π V Q E [ 2 = S21 ] 3 λ 2πV CCS = λq V s the volume of the room under test, λ, the wavelength, and E[ S 21 2 ] the statstcal 2 nd moment of S 21 over a rotaton of strrer and combnatons of antennas polarzatons. In Fgure 8, the equvalent CCSs are deduced from the S 21 -measurements and are drawn for several rooms of the targeted buldng n the whole frequency range of nterest. In ths fgure, reference means CCS of ONERA s empty MSRC. Ths reference confguraton n MSRC s overlad n order to put to the fore the dfference of EM physcs n a reverberatng chamber and n a standard buldng room. Note that the equvalent couplng cross-secton of a common buldng room of about 50 to 60m 3 s relatvely flat wth respect to the frequency; the values vares between 100 and 500m 2, whch s 20 to 30dB hgher than n a MSRC of the same sze. The comparson wth the results of the MSRC wth the open door s partcularly nterestng: they seem to ndcate that EM scatterng through wndows or doors s not sgnfcant n buldngs compared to other EM dsspatve mechansms such as classcal Joule or absorpton effects. (1) 8

10 Fgure 8 - Equvalent mean Couplng Cross-Sectons of dfferent rooms n the targeted buldng Inter measurements At a thrd step, EM couplng between several rooms of the targeted buldng has been measured n Inter confguraton n order to characterze EM attenuaton nsde the buldng. Fgure 9 presents a mappng of EM attenuaton deduced from the measurements at a frequency of 5.2GHz between the Emsson room, n whch the transmttng antenna s placed, and neghbourng rooms hostng the recevng antenna. Ths fgure shows that EM propagaton n the buldng depends on EM dsspatve mechansms that are specfc to each elementary room. Note that ergodcty between pont locatons n rooms and measured samples (ncludng strrer rotaton and polarsaton confguraton) has been appled to obtan ths mappng. Note that n smlar offce rooms, we almost obtan the same attenuaton level. The measured data are fnally post-processed to extract the transfer CCSs between the Emsson room and the other elementary rooms. Fgure 9 - Attenuaton between an emsson room and other neghbourng rooms n the targeted buldng (blue zones correspond to zones for whch no measurement s avalable) 9

11 2.4.3 Ext/Int measurements Fnally, the last confguraton ntends to measure EM transfer functons between an ncdent external feld and the resultng nduced EM felds n several rooms of the targeted buldng ( Ext/Int confguraton). These data wll be used later on as reference data for the valdaton of the PWB modellng of EM nteractons wth buldngs. Ext/Int measurement confguratons: We remnd that the ncdent feld s generated by the transmttng antenna (double rdge antenna) located at 20m far from the targeted room of the buldng as llustrated n Fgure 10. The emsson antenna s postoned n front of the wndow of the targeted offce room. Other observaton rooms are the Laboratory room and two other Offce rooms, all of them located at the frst floor of the targeted buldng. Fgure 10 - Sheldng effectveness of targeted offce and laboratory rooms for several open-closed confguratons of offces The wndows of the offce rooms are made of 4 glass panels, fully transparent to EM waves. Those wndows can be consdered as the only relevant Pont of Entry (POEs) of EM energy for our problem. All the offce room wndows are equpped wth metallc roller shades lkely to provde a sgnfcant attenuaton of the EM external nterference and act as an EM protecton aganst an external EM threat. The roller shades are made of metallc blades that can be set horzontally leavng 60 equvalent slots of 2cm wdth and 0.775m long n each of the four wndow panels. The dmensons of wndows of the targeted offce wth open roller shades and half-open roller shades are llustrated n Fgure 11. Fgure 11 - Offce-room wndows and roller shade confguratons 10

12 Therefore, 3 confguratons of wndows can be dstngushed: - Fully open when the roller shades are up - Half open when the roller shades are down n half open poston - Closed when the roller shades are down n closed poston In addton, n order to evaluate the nfluence of the poston of the blades of the roller shades, horzontal and vertcal polarzatons of the transmttng antenna have been consdered. Influence of roller shades poston: In order to evaluate the nfluence of the roller shades, varous confguratons of llumnaton of the targeted offce have been consdered dependng on the poston of the roller shades (open, half-open, closed), both n emsson and recepton rooms and dependng also on the polarzaton of the external ncdent EM feld: ) Horzontal polarzaton open roller shades n both rooms ) Horzontal polarzaton half-open roller shades n both rooms ) Vertcal polarzaton half-open roller shades n both rooms v) Horzontal polarzaton half-open roller shades n recepton room (n targeted offce room) open roller shades n emsson room v) Horzontal polarzaton closed roller shades n both rooms In all confguratons, the attenuaton brought by the roller-shades s defned as the transfer functons measured between the transmttng and the recevng antennas and normalzed to the reference confguraton n whch the roller shades are open n both rooms. The attenuatons brought by the metallc roller shades (closed or half-open) dependng on the polarzaton of the ncdent nterference are plotted n Fgure 12 up to 6GHz. The 0dB-level represents the reference confguraton wth open roller shades n both rooms. The reference confguraton does not depend on the polarzaton of the external feld. Fgure 12 - Effects of metallc roller shades onto buldng attenuaton n the targeted buldng room 11

13 The followng conclusons can be drawn: - Metallc roller shades when completely closed brng a sgnfcant attenuaton of about 40dB n the whole frequency range onto the EM nduced levels n the targeted offce. - When half open n both rooms, the attenuaton of metallc roller shades depends on the polarzaton of the external EM nterference. - Wth ncdent vertcal polarzaton, slots beng horzontal, the attenuaton s about 8dB less n the whole frequency range than the n full open confguraton. Ths result s consstent wth the fact that the equvalent area of the 60 slots s also about 8dB less than the open wndows. - Wth ncdent horzontal confguraton and n half-open confguraton of roller shades, the attenuaton vares wth frequency and the typcal behavour of slots correspondng to hgh-pass flters can be seen. The cut-off frequency depends on the dmensons of the slots. At hgh frequency (wth respect to the dmensons of the slots) the slots (n half open roller shades confguraton) let the EM energy penetrate through the wndow and the attenuaton s the same n both polarzatons of the ncdent external nterference. - Wth ncdent horzontal confguraton, the effect of hgh pass flter s more obvous when both rooms have half-open roller shades whch can be explaned by the fact that both half open roller shades act as two successve hgh-pass flters. Ths s also why the varaton wth frequency between 2.5GHz and 4GHz s much more abrupt than wth a sngle half open roller shade. Sheldng effectveness measurement results n targeted buldng: Fgure 13 llustrates an example of sheldng effectveness deduced from expermental results. Here the sheldng effectveness s defned as the rato between the ncdent electrc feld n free space at the level of the targeted room and the average electrc feld nduced n another room. The ncdent reference electrc feld n free space s supposed to be the one nduced by the transmttng antenna at the level of the targeted offce. It s calculated analytcally gven the gan of the antenna, the dstance antenna-targeted offce and the ncdent power at antenna nput. The results have been obtaned n two dfferent rooms of the targeted buldng: - The targeted offce room of the buldng n two confguratons of ts wndow shades (metallc roller shades open or closed), - The laboratory room for two confguratons of wndow shades (metallc roller shades open or closed) of the neghbourng offces. In ths analyss, the roller shades n the targeted buldng-room are n open poston but the roller shades of the two other buldng-offces are ether open or closed. 12

14 Fgure 13 - Sheldng effectveness of targeted offce and laboratory rooms for several open-closed confguratons of offce wndows The followng conclusons can be drawn: - In the targeted offce, the sheldng effectveness s low when the roller shades are n open poston because ts wndow s n front of the llumnatng antenna and consttutes the man POE of the EM energy n ths room. Off course, when the roller shades are closed, the sheldng effectveness ncreases. - In the laboratory-room, the sheldng effectveness has an average level of about 30dB when all the roller shades are open and 30dB when they are open n the targeted offce only. Ths result clearly shows that the two wndows of the two other buldng rooms also contrbute to the penetraton of EM energy nsde ths room. Man lessons: The lessons from analyss of the measurement results n varous confguratons of the llumnaton can be summarzed as follows: - EM couplng nsde buldng-rooms s manly due to nternal losses. Leakages through openngs (wndows and doors) are not sgnfcant n terms of loss of EM energy. Such a stuaton contrbutes to the localzaton of the average EM energy n rooms. - Openngs are the man POEs for the transfer of EM energy from the outsde or between rooms. The fact of openng or closng those openngs clearly nfluences the transfer functons n logcal addtve way. As a general concluson, even f the feld dstrbuton n buldng rooms cannot be modeled wth smple and effcent dstrbuton laws as n MSRC, the couplng between each room and the couplng between rooms approach can be characterzed by specfc CCS. Ths stuaton seems favorable to the applcaton of the network formulaton of the EMT/PWB theory. Ths s the subject of the next secton. 13

15 3 APPLICATION OF EMT/PWB METHODOLOGY As prevously mentoned, the dea s to use the expermental data base prevously presented n order to valdate EMT/PWB concepts appled to buldngs by comparng them to numercal data resultng from the EMT/PWB modellng. Recalls on EMT/PWB methodology are gven n paragraph 3.1. Paragraph 3.2 s dedcated to the valdaton of the EMT/PWB method when appled to modellng EM nteractons n buldngs. In paragraph 3.3, addtonal confguratons of the buldng are smulated n order to demonstrate the potentalty of the modellng method for assessng EM envronments n varous buldng rooms nduced by an IEMI/NNEMP attack. 3.1 Recalls on EMT/PWB methodology General concepts The prncples and the theoretcal formalsm of the EMT/PWB methodology to model EM nteractons n electrcally large systems has been fully detaled n ([4],[5]). Let us recall that the modellng methodology conssts n combnng Electromagnetc Topology concepts (EMT) of a system decomposed n topologcal volumes [8] and a PoWer Balance (PWB) quantfcaton of EM nteractons nsde and between those volumes. A qualtatve analyss of the EM problem enables buldng an nteracton dagram sequence n whch nodes represent volumes n whch dsspatve mechansm happen and branches descrbng transfer-of-energy paths between volumes. The quanttatve analyss s made n two steps: - The buldng of the topologcal network model - The resoluton of the topologcal model Topologcal network models The network model s made of tubes connectng junctons and derved from the nteracton dagram sequence and s flled wth relevant physcal features of EM nteractons usng PWB concepts. Let us pont out here the advantage of a network model compared to a smple transfer functon model appled on the nteracton dagram sequence. The network model ncludes n ts formulaton the fact that EM nteractons are of course b-drectonal snce backnteractons of one volume on the other are possble. In a transfer functon model, EM nteractons are undrectonal and do not account for any back couplng. In EMT, ths absence of reacton couplng s known under the Good- Sheldng approxmaton but t must be appled wth cauton. The network formulaton s also more general n the sense that sources can be appled anywhere on the network. PWB concepts assume that the structure under test s large enough compared to the ncdent wavelength and complex enough to assume that the exact geometry of the problem s not fully controlled. Consequently, as n classcal MSRCs, the nduced EM envronment can be modelled by pseudo-random varables and reduced to energetc (scalar) quanttes such as mean power denstes, mean dsspated power, equvalent CCSs. CCSs are defned as the rato between the dsspated power and the ncdent EM power densty. It has the property to be an ntrnsc characterstc (n the sense that t does not depend on the exctaton source and the test envronment used to measure t). Two types of CCSs can be dentfed: - EM energy dsspaton n volumes (Joule losses, scatterng through exteror openngs ) descrbed mathematcally by a matrx or a scalar number - Transfer of energy between volumes descrbed mathematcally by a matrx. In the PWB network formulaton we have chosen, the two types of CCSs are appled n junctons. Therefore, three types of junctons are defned: - POA junctons that contan the CCS scalars number descrbng the dsspaton of energy at POAs (Ponts Of Absorpton). - POE junctons that contan the matrces related to the transfer of external energy through POEs (Ponts Of Entry). - POC junctons that contan the matrces related to the transfer of nternal energy through POCs (Ponts Of Couplng). 14

16 - Volume junctons that sum-up the CCS from the POA, POE and POC junctons The consequence of such a descrpton s that the tubes are deal connectons that only carry a unt transfer functons. Nevertheless, they are the only elements of the network supportng the applcaton of sources as n usual EMT network models [7]. Ths PWB network model descrpton s the one chosen n the PWB computer code developed at ONERA and appled n the followng for quanttatve calculatons. As an example of EMT/PWB modellng of EM nteractons nsde a buldng, let us consder the case of a buldng made of three rooms as shown n Fgure 1 at the begnnng of ths paper. Let us also assume the IEMI attack scenaro n whch the RF source s amng at the buldng nsde a van parked onto the access road (supposed to be the only possble access to the buldng from the exteror). The topologcal decomposton of the problem can be summarzed as follows: - The buldng s along an access road, - Two rooms (room A and room B) have wndows facng the access road, - The thrd room (room C) s a control room wth no wndows and s supposed to contan crtcal equpment. To buld the equvalent EMT/PWB network of ths scenaro of attack, t s necessary to frst analyse from a qualtatve pont of vew the propagaton path of the RF ncdent EM energy from the source up to the nternal rooms by consderng all possble EM couplng paths. The analyss here s qute obvous: - The RF energy can enter the buldng by both wndows of rooms A and B (wth possble dfferent levels of EM felds at the levels of the wndows dependng on the relatve postons between wndows, road and EM source). - Once n room A, RF energy can ether dsspate n the room, go out through the wndow or go to room B and room C through the doors and the corrdor. - Once n room B, RF energy can ether dsspate n the room, go out through the wndow or go to room A and room C through doors and the corrdor. - Once n room C, RF energy wll dsspate n the room and nteract wth other rooms. All these EM nteractons are put together n the EMT/PWB network schemed n Fgure 14. From a quanttatve pont of vew, the CCSs are the relevant physcal nput parameters and they are assgned to each juncton representng an elementary EM mechansm. Fgure 14: EMT/PWB network of the smple buldng 15

17 In ths network model we can dentfy the followng elements: - Blue tubes ndcated that an equvalent source (ether an ncdent power or an ncdent power densty) s appled to ths access pont. - A red tube ndcates the EM volume of nterest (for us the room contanng the crtcal equpment) n whch EM levels must be evaluated. - Room-A, Room-B and Room-C junctons are volume junctons related to the 3 man rooms of the buldng. - Wndow-room A, Wndow-Room-B, are POE junctons. - Transfer-Room-A_Room-B, Transfer-Room-A_Room-C, Transfer-Room-B_Room-C are POC junctons. - Dsspaton-Room-A, Dsspaton-Room-B and Dsspaton-Room-C are POA junctons EM resoluton of topologcal models In the EMT formalsm, the unknowns are defned as ncomng and outgong waves travellng along tubes. These waves are then gathered nto the BLT network equatons recalled hereafter where [W(0)], [W(L)] and [Ws] are respectvely, the ncomng, outgong and source waves. [S] s the scatterng matrx and [Γ], the propagaton matrx along tubes. In ths formaton, no sources are appled on a port of a juncton. If a source must be appled on a juncton, t s appled on the tube connected to ths port: ths formalsm s an analogy wth what s done on multconductor network formalsm. Propagaton equaton: ( ) = [ Γ] ( ) + [ ] WL.W0 W s Scatterng equaton: W0 ( ) [ S.WL ] ( ) = BLT equaton: ([ I] [ S ].[ Γ] ). W( 0) = [ S ].[ W s ] Here, n order to match the EMT formalsm wth PWB concepts, two waves are appled on each tube n order to descrbe ncomng and outgong waves at juncton level. They are defned as a lnear combnaton of dsspated power P and power densty S as llustrated n Fgure 15. Ac s called the characterstc CCS chosen arbtrarly as the couplng cross secton of a perfectly matched antenna and equal to λ 2 /8π. (2) W n (0) W n (L ) W s n Juncton n Tube Juncton m W (0) = A.S W n m c (0) = A c.s n m P n P m W (L ) = A.S W n m c (L ) = A c.s m n + P + P m n (3) W s m W m (L ) W m (0) Fgure 15 Defnton of waves on a tube n EMT/PWB formalsm Junctons are characterzed n our formalsm by scatterng parameters [S] lnkng at each juncton ncomng and outgong waves. S-parameters are derved from the dual CCS matrx [A] of the juncton. These relevant physcal features can be ether derved from analytcal and theoretcal expressons or deduced from measurements of Q factors n volumes or measurements of transfer functons between areas as presented n paragraph 2.4 n Intra and Inter confguratons. 16

18 Let us fnally note that n the BLT equaton formulaton of PWB, propagaton of waves along tubes s taken nto account n the super matrx [Γ] whch s reduced to the unt matrx at ths moment n the PWB formalsm. The EMT/PWB formalsm as descrbed above has been mplemented at ONERA n a dedcated computer code named PWB [9]. It s based on the CRIPTE code ntally developed at ONERA n the 90 s to model and evaluate EM nteractons on multconductor cable networks [10]. The EMT/PWB formalsm and ts related numercal code PWB had been fully valdated onto dfferent test cases, from a generc smplfed structure up to an arcraft [5]. 3.2 Valdaton of EMT/PWB methodology Topologcal network model of the 1 st floor of the targeted buldng The EMT/PWB network model correspondng to the confguratons of llumnaton of ONERA s targeted buldng s deduced from the analyss of the topology of the buldng; the topologcal network model s llustrated n Fgure 16 together wth a geometrcal scheme of the 1 st floor of the targeted buldng. In ths fgure, we can dentfy the followng PWB network elements: - Blue tubes ndcate couplng paths between the exteror and the buldng. - Junctons 2, 10 and 15 represent wndows (POEs). Let us recall that n the smulaton, the wndows are consdered as four free apertures (1.9m x 0.775m) when there s no roller shade and as 60 slots of 2cm wdth and 0.775m when the roller shade s half-open. In confguraton wth closed metallc roller shades, the source terms on assocated blue tubes wll be removed n the EMT/PWB modellng. - The room junctons (volume junctons 3, 6, 11 and 16). Such a juncton corresponds to a so-called Ideal juncton [11] whch makes the deal connecton of all the ports of the juncton (equvalent to sum-up all the powers on the connected tubes) - The couplng paths between room junctons (POC junctons 5, 12, 18). The assocated CCSs are deduced from the Inter confguraton transfer functon measurements presented n secton 2. - The varous global ntrnsc dsspatve mechansms (POA junctons 4, 8, 13 and 17) and the recevng antennas (POA junctons 7, 19 and 20) located n rooms under test (connected to purple tubes). As for POC junctons, the assocated CCSs are deduced from the Intra confguraton transfer functon measurements presented n chapter 2. Note that the whole buldng has been reduced to offces and corrdor located next to the targeted offce at the frst floor n order to smplfy the EMT/PWB modellng. 17

19 Fgure 16 - EMT/PWB network model of the llumnaton confguraton of ONERA s targeted buldng Pror to model the entre buldng llumnaton and n order to valdate the EMT/PWB approach appled on buldngs, some specfc expermental confguratons wth ncreasng levels of complexty have been consdered and smulated. They are reported n the followng paragraphs Basc confguraton: EM envronment modellng n the targeted offce Ths basc confguraton conssts n measurng and smulatng the sheldng effectveness n the targeted offce whch s drectly llumnated by the external ncdent EM feld n horzontal polarzaton. For ths we consder an deal far feld model of the antenna at a dstance d=20m (dstance between the llumnaton and targeted buldngs as n Fgure 6). In ths confguraton the man couplng path s the llumnated wndow of the targeted offce wth, n the one hand, no roller shades and, n the other hand, the half-open metallc roller shade. As far as the orders of magntude and global frequency varaton are concerned, Fgure 17 shows that the PWB modellng gves a good agreement n the whole frequency range between the measured and smulated sheldng effectveness obtaned n ths basc confguraton. 18

20 Fgure 17 - Sheldng effectveness of targeted offce smulaton and measurement Horzontal polarzaton More complex confguratons: sheldng effectveness n the laboratory room The next valdaton step conssts n measurng and smulatng the EM envronment nduced n the laboratory room whch s not n the lne of sght of the external llumnaton. Therefore, the path to reach the laboratory room s not drect and can be decomposed as follows: the external feld penetrates n parallel n the buldng through the wndow of the targeted offce room and through the wndows of the neghbourng offce room. Once nsde, the EM energy scatters dependng on dsspatve mechansms and nternal couplng paths between elementary rooms of the buldng. Two confguratons are consdered: - Confguraton A: all roller shades are closed except the wndow of the targeted offce - Confguraton B: all roller shade wndows (targeted offce and 2 neghbourng offces) are open. Measured and smulated sheldng effectveness are drawn respectvely n Fgure 18 and Fgure 19 n confguratons A and B. Fgure 18 - Sheldng effectveness of laboratory room confguraton A 19

21 Fgure 19 - Sheldng effectveness of the laboratory room confguraton B The agreement between measurements and smulatons s qute consstent n the whole frequency range f one consders the orders of magntude and global frequency varaton. Of course, some dscrepances can be noted. These dscrepances can be explaned as follows: - In both confguratons, the EMT/PWB network has been reduced to the man couplng paths between the targeted room, two neghbourng offces and the laboratory. Ths representaton may be nsuffcent and addtonal EM nteracton paths mght have been ncluded n the model such as, for example, the couplng paths between neghbourng laboratory room or other rooms whch can be n the radaton pattern of the llumnaton antenna. - The external ncdent EM feld has been also reduced n the EMT/PWB modellng to an deal plane wave mpngng globally onto all llumnated wndows. The ampltude of the plane wave has been deduced analytcally from expermental condtons as the dstance between antenna and targeted room, the gan of the antenna. Therefore, scatterng due to the buldng and due to the other surroundng buldngs has been neglected and ths assumpton may be too smplfed for our expermental setup. 3.3 EMT/PWB modellng of addtonal confguratons Effcency of metallc roller shades for EM protecton The last step of ths work conssts n smulatng addtonal confguratons of IEMI/NNEMP attacks (for whch no comparsons wth measurement s avalable) onto the buldng n order to demonstrate how ths modellng approach may be used n order to draw protecton measures. We have chosen to analyse the effcency of metallc roller shades as EM protecton devces. Let us thereby assume the followng attack scenaro n whch: - The IEMI/NNEMP weapon s located outsde the buldng facng the prevously targeted offce and radates a plane wave wth horzontal polarzaton, - The targeted room s not equpped wth metallc roller shade, the wndow s supposed to be free. Ths mght be an operatng constrant mposed to the buldng. - The crtcal equpment s located n the laboratory room - The neghbourng rooms are equpped wth metallc roller shades whch can be ether open, half open or closed. 20

22 Let us suppose that we do not want to overcome the susceptblty level of the crtcal equpment n room C and we need to defne the approprate protecton to be appled at the levels of the wndows. For ths we run dfferent confguratons of the metallc roller shades n the other offce rooms n order to assess the level of protecton brought by the roller shades: - Confguraton 1: metallc roller shades of both neghbourng rooms are closed, - Confguraton 2: metallc roller shades are half-open n both neghbourng rooms, - Confguraton 3: one open wndow n one neghbourng room and metallc roller shade are half-open n the other neghbourng room, - Confguraton 4: open roller shades n both neghbourng rooms. The smulated EM levels nduced n the laboratory room normalzed to Confguraton 4 are drawn n Fgure 20. The plots ndcate the attenuaton effcency of metallc roller shades regardng the frequency of the IEMI/NNEMP attack of the buldng. Dependng on the level and frequency range of EM susceptblty of the equpment and the frequency of the IEMI/NNEMP attack, the user wll be able to decde wth these smulatons whch confguraton s approprate for EM protecton and operatonal condtons. For example below 3GHz, metallc roller shades do not need to be completely closed to obtan a suffcent EM attenuaton of 20dB. Ths concluson s of course not vald anymore at hgher frequency; for the same level of 20dB attenuaton, metallc roller shades must be closed. Fgure 20 Attenuaton brought n the laboratory room by other rooms metallc roller shades 21

23 4 CONCLUSION The work presented n ths paper focused on the analyss of transmsson of EM RF energy nsde buldng rooms through buldng wndows at whch the frst level of possble EM protecton versus external attack can be performed. The paper also nvestgated the transmsson of the EM energy nsde a buldng. In parallel to experments carred out n ONERA s buldng n specfc rooms, numercal modellng have been conducted n order to valdate the relevance of EMT/PWB methodology to assess EM envronment nduced n a buldng by an IEMI/NNEMP attack. The modelng results are very encouragng but some addtonal expermentaton n dfferent types of buldngs (wth renforced concrete walls for example) would be nterestng n order to enlarge the conclusons on the scope of applcaton of EMT/PWB concepts. In parallel, a specfc work s requred to obtan theoretcal analytcal models of losses n rooms dependng on ther man features (as volumes, materals, openngs,..) and of POEs. Expermental characterzaton of EM losses and EM couplng paths n a gven buldng can be made effcently by usng smple laboratory equpment and transmsson of data onto the bult-n Ethernet network. In the future, we can thnk of developng dedcated and lghter equpment n order to obtan those measurements. Especally transmsson of data wth RF sgnals may be of nterest f the Ethernet access s not possble. Once those measurements avalable (even partally), ths methodology and the assocated model can be now used to play varous attack scenaros and obtan the nduced EM levels at the level of crtcal equpment or nternal functons nsde a buldng. From now on, EMT/PWB modelng of EM nteractons n a buldng can also be used n order to help to the dagnoss of an IEMI/NNEMP attack. Two dfferent ways can be proposed: - Evaluaton of crtcal EM envronments for crtcal equpment nsde the buldng rooms gven the EM sources measured by detectors at POEs - Evaluaton of external ncdent EM source ampltude and frequency spectrum gven the EM envronment measured nsde the buldng by detectors (nverse problem). 22

24 5 REFERENCES [1] W. A. RADASKY, C. E. BAUM, Introducton to the specal ssue on hgh-power electromagnetcs (HPEM) and ntentonal electromagnetc nterference (IEMI), IEEE Transacton on Electromagnetc Compatblty, Vol. 46, No. 3, pp , 2004 [2] R. HOAD, W. RADASKY, Progress n IEC SC 77C standards addressng HPEM threats to systems and nfrastructure, EUROEM 2012, Toulouse, France, July 2012 [3] O-H. ARNESEN, Protecton of Crtcal Infrastructures aganst Hgh Power Mcrowave Threats HIPOW - An EU 7th framework project on protecton of electronc systems aganst natural and manmade electromagnetc threats, n AMEREM2014 proceedngs, Albuquerque USA, July 2014 [4] I. JUNQUA, J-P. PARMANTIER, F. ISSAC, A Network Formulaton of the Power Balance Method for Hgh- Frequency Couplng, Electromagnetcs 25:7-8, pp , or n Interactons Notes, Note 576, November 2002 [5] I JUNQUA, Hgh frequency couplng mechansms n complex systems: analyss and assessment by the PoWer Balance method, Ph.D. Dssertaton of Llle Unversty, June 2010 (n French) [6] C. LEMOINE, P. BESNIER, M. DRISSI, Investgaton of reverberaton chamber measurements through hgh power goodness-of-ft tests, IEEE Transacton on Electromagnetc Compatblty, Vol. 49, No. 4, pp , 2007 [7] I. JUNQUA, J-P. PARMANTIER & Al, Combnng Asymptotc Methods and Power Balance Approach to smulate HIRF HF scenaros, n EUROEM2012 proceedngs, Toulouse France, July 2012 [8] C.E. BAUM, The theory of electromagnetc nterference control, Interacton Notes Note 478, December 1989 also n modern rado scence 1990, pp , Oxford Unversty Press. [9] I. JUNQUA, S. BERTUOL, J-P. PARMANTIER, Power Balance TM User s Gude Verson , IDDN.FR S.P , October 2009 [10] J.P. PARMANTIER, X. FERRIÈRES, S. BERTUOL, C. E. BAUM, Varous Ways to Thnk of the Resoluton of the BLT Equaton wth an LU Technque. Interacton Notes. Note 535. January 1998 [11] J.P. PARMANTIER, An effcent technque to calculate deal juncton scatterng parameters n multconductor transmsson lne networks. Interacton Notes. Note 536. February