This is a repository copy of Effect of power state on absorption cross section of personal computer components.

Transcription

1 This is repository copy of Effect of power stte on bsorption cross section of personl computer components. White Rose Reserch Online URL for this pper: Version: Accepted Version Proceedings Pper: Yn, Jiexiong, Dwson, John orcid.org/ nd Mrvin, Andy orcid.org/ (017) Effect of power stte on bsorption cross section of personl computer components. In: 017 Interntionl Symposium on Electromgnetic Comptibility - EMC EUROPE. EMC Europe 017, Sep 017, Angers. Reuse ["licenses_typenme_other" t defined] Tkedown If you consider content in White Rose Reserch Online to be in brech of UK lw, plese tify us by emiling eprints@whiterose.c.uk including the URL of the record nd the reson for the withdrwl request. eprints@whiterose.c.uk

2 017 IEEE. Personl use of this mteril is permitted. Permission from IEEE must be obtined for ll other uses, in ny current or future medi, including reprinting/republishing this mteril for dvertising or promotionl purposes, creting new collective rks, for resle or redistribution to servers or lists, or reuse of ny copyrighted component of this rk in other rks. 017 Interntionl Symposium on Electromgnetic Comptibility - EMC EUROPE, 4-8 September, 017 Effect of power stte on bsorption cross section of personl computer components Applictions to enclosure shielding Jiexiong Yn, John Dwson, nd Andy Mrvin Deprtment of Electronics University of York York, UK {jy936, john.dwson, ndy.mrvin}@york.c.uk Abstrct Kwledge of the shielding effectiveness of n enclosure is importnt for the electromgnetic comptibility of electronic systems. The shielding effectiveness of n enclosure depends on the bsorption cross section of its contents. It might be expected tht the energy bsorption in n electronic circuit chnges ccording to the operting stte of the semiconductor devices which compose the ctive components. In most published reserch, the bsorption cross section mesurements were performed when the contents were unpowered. In this pper we compre the mesured bsorption cross sections of the components of personl computer in powered nd unpowered sttes. Comprisons indicte tht power nd the prticulr operting configurtion do t hve significnt influence on the bsorption cross section. This mens tht the process of determining bsorption cross sections of circuit bords nd other components cn be esily chieved without the need to provide power nd define prticulr operting stte. Keyrds shielding effectiveness; borption cross section; reverbertion chmber I. INTRODUCTION The bility of n electronic device s enclosure to protect the device ginst electromgnetic interference is quntified by its shielding effectiveness (SE). Existing stndrds such s IEEE 99.1 [1] specify the mesurement of the SE of n empty enclosure. In rel pplictions, n enclosure is used with some contents inside it. Previous rk hs shown tht contents cn ffect the SE of n enclosure []-[5] since they bsorb some energy nd thus decrese the internl field. To ccurtely predict the shielding effectiveness of populted enclosures, it is necessry to understnd these effects. One successful pproch to nlyze the shielding problems of electriclly lrge enclosures with contents is the power blnce method (PWB) proposed by Hill et l [6]. The min dvntges of the method re tht it does t require the kwledge of the detiled geometry of contents nd it hs low computtionl cost compred with full wve solvers. In order to pply the power blnce method, the contents re chrcterized by their bsorption cross sections (ACSs); nd the pertures of the enclosure re chrcterized by their /17/$ IEEE trnsmission cross sections (TCSs). The ACS, σ (m ), of n object is the equl to the re of perfect bsorber tht uld bsorb the sme mount of energy. It is defined s the rtio of the bsorbed power (W) to the incident power density (W/m ). The TCS, σ t (m ), of n perture is n re equl to the rtio of the power trnsmitted through the perture (W) to the incident power density (W/m ). The SE of n electriclly lrge enclosure in reverbernt environment my be defined s the rtio of externl (incident) energy density ( S ) to the internl energy density o ( S i ) nd cn be expressed directly in terms of the ACS of the contents nd TCS of ny pertures [7]: SE S S o t = = (1) i σ + σ σ where the ACS nd TCS re ssumed to be vlues verged over ll ngles of incidence. In previous reserch, the ACSs of printed circuit bords (PCBs) hve been mesured in reverbertion chmber [5]- [8] nd the mesurements were performed when the PCBs were unpowered. We expected tht the power stte (power on or power off) might ffect the ACS of electronic equipment s the ctive devices in the circuit tend to be nconducting when powered off nd conducting only when powered on. Consider digitl circuit tht contins logic gte. When the circuit is t powered, the trnsistors of the logic gte re n-conducting, wheres when the circuit is powered, some of the trnsistors re switched into conduction to determine the logic stte of the gte; this chnges the circuit impednce nd hence we uld expect chnge in power bsorbed from n externl field. We uld lso expect the impednce of the circuit to chnge s the logic gte chnges stte. We hve observed this in the rerdited energy from digitl systems [9]. To test our ssumption, in this rk we mesured the ACS of the internl components of personl computer, in both powered nd unpowered sttes in reverbertion chmber. t

3 Fig. 1. Photos of the computer under test. () Side view (without side pnel). (b) Front view (without cover). (c) Rer view. Section II provides the detils of the computer. Section III outlines the theory of ACS mesurement. Section IV presents the mesurement results nd section V is the conclusion. Fig.. Digrm of the reverbertion chmber set up for the ACS mesurements. II. COMPUTER UNDER TEST Figure 1 shows the personl computer under test which hs dimensions of 370mm 180mm 350mm. It cn be seen tht there re vrious pertures of different shpes nd sizes in the enclosure. Inside the computer there re motherbord, power supply unit, CD-ROM, disk drive, severl wires nd some extr structures. During the mesurements the keybord, mouse nd monitor of the computer were removed. Also, the side pnel of the computer ws removed to expose the circuitry to the electromgnetic energy in the chmber thus enbling the overll ACS to be mesured. III. MEASUREMENT OF ACS A. Mesurement Methodology The ACS mesurements were performed in reverbertion chmber with dimensions of 4.7m 3m.37m nd lowest useble frequency (t three times the first resonnce) of pproximtely 178MHz. The cross over frequency, or the Schroeder frequency of the empty chmber, t which the rtio of mode bndwidth nd mode spcing is 3, is bout 1.7GHz. A mechnicl stirrer ws fitted into the chmber to chieve field homogeneity. The digrm of the mesurement configurtion is shown in Figure. T Blde ntenns were used for the mesurements nd their detils cn be found in [11]. A vector netrk nlyzer ws used to collect S-prmeters from the t ntenns. Figure 3 is photo of the mesurement set up. The ntenns were plced t t opposite corners of the chmber. To reduce the coupling, they were t fcing ech other. The computer ws supported on polystyrene block. The ntenns nd the computer were t lest 0.m wy from the chmber wlls nd the floor. This gurntees tht they re t lest qurter wvelength wy from the chmber Fig. 3. Photo of the reverbertion chmber set up for the ACS mesurements. over the whole frequency rnge used. B. Frequency domin method Initilly we followed the method described in [8] where the verge ACS of n object in reverbertion chmber is given by: where λ 1 = 8π G 1 G σ () indictes n verge over number of mesurements with either different stirrer positions nd/or over rnge of frequencies; λ is the wvelength; nd G re the mesured verge net power trnsfer function between ntenns in the chmber with nd without the object respectively. The net power trnsfer function is obtined by: G

4 Fig. 4. Digrm of the distribution of the frequency segments for the time domin method. τ nd τ re the verge chmber power time constnts with nd without the object respectively. This method lso hs the dvntge of being independent of the efficiency of the ntenns used for the mesurement. The chmber time constnt is obtined by pplying the n-liner curve fitting technique described in [13] to the verge power dely profile (PDP) of the input energy t ech frequency of interest. The PDP is given by: G S 1 ( 1 S11 )( 1 S ) = (3) where S 1 is the trnsmission coefficient between the trnsmitting nd receiving ntenns, S 11 nd S re the reflection coefficients of the t ntenns. The losses within ech ntenn re ssumed to be negligible. The mesurement uncertinty is defined s: std σ α std( σ ) σ = (4) where ( ) is the stndrd devition of the ACS over number of mesurements with either different stirrer positions nd/or over rnge of frequencies. In [10], Flintoft et l include model to clculte the uncertinty of ACS mesurements in reverbertion chmber. Both (4) nd the method in [10] led to similr uncertinty, which is bout 0%. The ACS of the computer ws mesured from 1GHz to 18GHz. For the frequency domin method, the mechnicl stirrer ws set to move 100 uniformly spced positions over one rottion equl-spced points were recorded over the frequency rnge. The sweep time of the netrk nlyzer ws 4s. The frequency stirring technique, using bndwidth of 100MHz, ws pplied to the mesured dt to further reduce mesurement error. C. Time domin method A mesurement uncertinty of 0% ws considered too high to llow smll chnges to ACS to be observed. Therefore, we lso mesured the ACS of the computer by using the time domin method, presented in the pper of Zhng et el [1], which llows significntly lower mesurement uncertinty. The verge ACS is determined from the verge chmber time constnts: σ = V c 1 τ 1 τ where V is the chmber volume, c is the velocity of light, (5) PDP ( f ) IFFT[ S ( f, n) ( f f )] = (6) 0 1 win where S ( f, n) 1 is the trnsmission coefficient between the trnsmitting nd receiving ntenns t frequency f nd stirrer position n ; IFFT is the inverse fst Fourier trnsform opertion; win( f 0 ) f is window function which selects suitble bnd of frequencies bout the desired frequency of interest f 0 ; nd indictes n verge over ll the stirrer positions. In [1], Zhng et l described how to estimte the mesurement uncertinty. By using the Monte Crlo method, the uncertinty of the mesured ACS is clculted to be pproximtely 1%. As the time domin method requires very smll frequency steps to obtin sufficiently long time-spn in the IFFT, segmented frequency sweeping technique ws used; insted of sweeping the entire frequency bnd, only nrrow bnd round ech frequency of interest is considered. In this wy, the mount of dt collected nd the mesurement time cn be optimized. The distribution of frequency segments, which is depicted in Figure 4. A totl of 171 segments re uniformly distributed from 1GHz to 18GHz. Ech segment hs bndwidth of 5MHz nd contins 51 equl-spced points. In this pper we used rised cosine window function. The continuous stirring technique ws used nd 800 stirrer positions were smpled t ech frequency. D. Powered nd unpowered mesurements According to () nd (5), for both frequency nd time domin methods, t mesurements re required to obtin the ACS. First the net power trnsfer function or the time constnt of the empty chmber ws mesured, followed by tht of the chmber loded by the computer. The ACS of the unpowered nd powered computer were ech mesured three times. When the computer ws powered, the Windows 7 operting system ws ctivted. To mke the computer run t full cpcity, stress test progrm, HevyLod, ws used [14]. It tests the CPU, GPU nd hrd drive simultneously. In ddition, music CD ws plyed by using the Windows medi plyer to mke full use of the CD-ROM. During ll mesurements the ntenns nd the computer were kept t the sme plces in the chmber. The only thing tht chnged ws the power stte of the computer. 0

5 Fig. 5. Mesured ACS of the unpowered nd powered computer in the reverbertion chmber (obtined by the frequency domin method) Fig. 6. Mesured ACS of the unpowered nd powered computer in the reverbertion chmber (obtined by the time domin method) IV. RESULTS Figure 5 shows the ACS of the unpowered nd powered computer mesured in the reverbertion chmber by using the frequency domin method. It cn be seen tht the ACS of the computer hs vlue ner 0.0m over the whole frequency rnge. It is pprent tht the mesured ACSs of the computer re similr despite of different power sttes. Although they re t identicl, the generl fetures re the sme. Whilst it is difficult to tell if there is ny smll difference in ACS between the powered nd unpowered sttes, ny difference tht does exist is less thn the ~0% sttisticl vrition in the mesurements. As hs been mentioned, the frequency domin method gives reltively lrge mesurement vribility; therefore we decided to repet the mesurements with incresed ccurcy using the time-domin method. Figure 6 presents the mesured ACSs of the unpowered nd powered computer obtined by using the time domin method. It cn be seen tht the ACS of the computer is bout 0.0 m, which is similr to the results obtined by the frequency domin method. Here we cn see tht ny vrition in ACS between powered nd unpowered sttes is t obvious within the mesurement uncertinty, which is ~1%. Figure 7 shows comprison between the ACS of the computer obtined by both frequency nd time domin methods. The results shown in Figure 7 re the verges of the six mesurements presented in Figure 5 nd 6 respectively. It cn be seen tht both methods led to similr ACS. The time domin method hs much lower mesurement uncertinty thn the frequency method. Idelly, for the frequency domin method, incresing the number of stirrer positions uld reduce mesurement uncertinty, but tht uld gretly increse the time of mesurement. In this study, for instnce, the frequency domin method took bout Fig. 7. Comprison between the ACS of the computer obtined by frequency nd time domin methods. 40 minutes to finish 100 stirrer positions while the time domin method used only 10 minutes to finish 800 positions. Also the time domin method does t require the kwledge of the ntenn efficiency. For the frequency domin method we llowed for the ntenn reflection loss in (3) but on ny loss within the ntenn. However the ntenns used hve very low internl loss, so the ACSs mesured by the t techniques re close. V. CONCLUSIONS In this pper we hve presented results demonstrting how power stte ffects the ACS of electronic equipment. The ACS of personl computer t different power sttes hs been mesured in reverbertion chmber over the frequency rnge of 1GHz to 18GHz by using both frequency

6 nd time domin methods. The results suggest tht for this computer, power stte hs little influence on its ACS. Therefore, it cn be concluded tht for this computer, the ACS vries slowly with frequency nd is independent of power stte within 1% mesurement vribility. When performing similr ACS mesurements in the future, there will be need to consider power stte, which will bring convenience to the mesurements. REFERENCES [1] IEEE Stndrd Method for Mesuring the Shielding Effectiveness of Enclosures nd Boxes Hving ll Dimensions between 0.1 m nd m, IEEE Std , 014. [] D. W. P. Thoms, A. C. Denton, T. Konefl, T. Bensen, C. Christopoulos, J. F. Dwson, A. C. Mrvin, S. J. Porter nd P. Sewell, Model of the electromgnetic fields inside cuboidl enclosure populted with conducting plnes or printed circuit bords, IEEE Trnsctions on Electromgnetic Comptibility, vol. 43,., pp , My 001. [3] A. C. Mrvin, J. F. Dwson, S. Wrd, L. Dwson, J. Clegg nd A. Weissenfeld, A proposed new definition nd mesurement of the shielding effect of equipment enclosures, IEEE Trnsctions on Electromgnetic Comptibility, vol. 46,. 3, pp , August 004. [4] A. Rusieski, K. Aniserowicz, A. P. Duffy nd A. Orlndi, Internl stirring: n pproch to pproximte evlution of shielding effectiveness of smll slotted enclosures, IET Science, Mesurement Techlogy, vol. 10,. 6, pp , August 016. [5] S. L. Prker, I. D. Flintoft, A. C. Mrvin, J.F Dwson, S.J. Ble, M. P. Robinson, M. Ye, C. Wn nd M. Zhng Predicting shielding effectiveness of populted enclosures using bsorption cross section of PCBs, Proceedings of the 016 Interntionl Symposium on Electromgnetic Comptibility, pp , September 016. [6] D. A. Hill, M. T. M, A. R. Ondrejk, B. F. Riddle, M.L. Crwford nd R. T. Johnk, Aperture excittion of electriclly lrge, lossy cvities, IEEE Trnsctions on Electromgnetic Comptibility, vol. 36,. 3, pp , August [7] A. Gifuni, Reltion between the shielding effectiveness of n electriclly lrge enclosure nd the wll mteril under uniform nd isotropic field conditions, IEEE Trnsctions on Electromgnetic Comptibility, vol. 55,. 6, pp , December 013. [8] I. D. Flintoft, S. L. Prker, S. J. Ble, A.C. Mrvin, J. F. Dwson nd M. P. Robinson, Mesured verge bosrption cross-sections of printed circuit bords from to 0 GHz, IEEE Trnsctions on Electromgnetic Comptibility, vol. 58,., pp , April 016. [9] I. D. Flintoft, A. C. Mrvin, M. P. Robinson, K. Fischer nd A. J. Rowell, The re-emission spectrum of digitl hrdwre subjected to EMI, IEEE Trnsctions on Electromgnetic Comptibility, vol. 45,. 4, pp , November 003. [10] I. D. Flintoft, S. J. Ble, S. L. Prker, A.C. Mrvin, J. F. Dwson nd M. P. Robinson, On the mesurble rnge of bsorption cross section in reverbertion chmber, IEEE Trnsctions on Electromgnetic Comptibility, vol. 58,. 1, pp. -9, Februry 016. [11] A. C. Mrvin, G. Esposito, J. F. Dwson, I. D. Flintoft, L. Dwson, J. A. K. Everrd nd G. C. R. Meli, A wide-bnd hybrid ntenn for use in reverbertion chmbers, Proceedings of the 013 Interntionl Symposium on Electromgnetic Comptibility, pp. -6, August 013. [1] X. Zhng, M. P. Robinson, I. D. Flintoft nd J. F. Dwson, Inverse Fourier trnsfer technique of mesuring verge bsorption cross section in the reverbertion chmber nd Monte Crlo study of its uncertinty, Proceedings of the 016 Interntionl Symposium on Electromgnetic Comptibility, pp , September 016. [13] X. Zhng, M. Robinson nd I. Flintoft, On mesurement of reverbertion chmber time constnt nd relted curve-fitting techniques, Proceedings of the 015 Interntionl Symposium on Electromgnetic Comptibility, pp , August 015. [14] HevyLod. [Online]. jm-softwre.com. Avilble: [Accessed: 14/10/016].