On the relation between radiated and conducted RF emission tests

Transcription

1 Presented at te 3 t International Zuric Symposium on Electromagnetic Compatibility, February 999. On te relation between radiated and conducted RF emission tests S. B. Worm Pilips Researc Eindoven, te Neterlands Abstract Tis paper deals wit te RF emission from small electronic products wit a single (mains) cable in te frequency range from 30 to 30 MHz. Because it is important to ave EMC assessments in an early stage of te product development, tere is a need for simple, low-cost test metods tat can be applied in te design lab. A workbenc set-up, as described in IEC , wic is primarily intended for RF immunity testing, is also suitable for conducted RF emission testing. Te relation between conducted and radiated test results is discussed and special attention is paid to systematic measurement uncertainties tat may occur in bot setups. A simple teoretical relation is derived, wic corresponds well to te relation obtained from measurements, provided tat te resonance effects in te radiated set-up are suppressed, e.g. by applying a common mode termination impedance to te cable. Introduction. In tis paper we would like to discuss te measurement of RF emissions from small electronic products wit a single cable (i.e. mains) in te frequency range from 30 to 30 MHz. In order to reac compliance wit te EMC standards at a minimum of cost, it is useful to include some kind of EMC assessment in te early pases of te product s development, so tat te designers receive fast feedback on any canges tat tey make in teir design. Te standard radiated RF emission test procedure described in CISPR is not suitable for tis purpose, because it is too costly and too timeconsuming. A workbenc test set-up as described in standard IEC , primarily intended for conducted RF immunity testing using a reference plane and coupling/decoupling networks (CDNs), is muc more appropriate and can easily be installed in a design lab. It as been reported by several autors, e.g. [] and [], tat te radiation caracteristics of small electronic systems are determined mainly by te cables wic are connected to suc a system. Tis means tat, in particular for products wit a single cable, it must be possible to find a relation between te electric field strengt measured according to CISPR and te common mode current measured via a CDN in te conducted test set-up. We will take 30 MHz as te upper frequency of our considerations for te following reasons. Te conducted test set-up and corresponding measurement equipment of IEC are defined up to 30 MHz. CISPR as a 7 db relaxation of te limits above 30 MHz, and te products tat we are aiming at ave never posed any difficulties above tis frequency. Usually, EMC tests are not very accurate. A set-up itself may ave systematic errors (e.g. up to ± 4 db) and te layout of cables may add several dbs of uncertainties due to resonance effects. If we want to find a translation curve relating two different test metods, we sould terefore try to eliminate te systematic errors, in particular resonance effects, oterwise we could easily end up wit more tan 0 db of uncertainty in te translation curve. Numerical and analytical models for te test metods can be applied to support te measured results and to find a suitable curve for translating te conducted test results into CISPR limits. Figure. Open area test site (OATS) Te paper is arranged as follows: First a sort description will be given of te open area test site and te semi-anecoic room, wic serve as te reference sites for CISPR. It will be demonstrated

2 tat te common mode termination impedance near te ground plane as a strong influence (about 5 db) on te measured electric field strengt due to resonance effects on te mains cable. A conducted set-up for measuring te common mode current on te cable and possible causes of systematic errors will be discussed in section 3. In comparing te measured field strengt and te measured common mode current it turns out tat a smoot translation curve can be obtained if in te radiated set-up a CM termination is cosen suc tat te VSWR on te cable remains low (section 4). Tis fact can be confirmed via numerical analysis (section 5) and even by means of a simple analytical model of te radiated set-up (see appendix). Tese teoretical considerations lead to a simple translation curve tat can be applied in te design lab during workbenc testing. Figure. Spectrum of a 0 MHz oscillator measured in a SAR wit various CM termination impedances. Te radiated RF emission test set-up Standard radiated RF emission tests in te frequency range above 30 MHz are described in CISPR. Te metod requires an open area test site (OATS) wit a conducting ground plane and a test distance of 3 m or 0 m; see Figure. Te device under test is placed on a turntable at 0.8 m above te ground plane. At eac measurement frequency te turntable is rotated stepwise to cover te full 360. Te waves emitted by te device under test (DUT) can reac te receiving antenna via a direct pat and an indirect pat (reflection at te ground plane), so interference between te two waves may occur. Tis interference causes a reduction in te sensitivity of te set-up in certain frequency ranges, wic can be overcome by scanning te eigt of te receiving antenna (between and 4 m). Te wole sequence is performed for orizontal and vertical polarisation of te receiving antenna, so te metod becomes quite elaborate. As an alternative to te open area site a semi-anecoic room (SAR) may be used. Te requirement for bot te OATS and te SAR is tat te normalised site attenuation lies witin ± 4 db from te teoretical site attenuation. Since we wanted to use te SAR as a more accurate field strengt measuring instrument (e.g. ± db), we ave performed an additional calibration of tis test set-up. Suc a calibration is normally not used in EMC tests and te difference between actual and teoretical site attenuations is a generally accepted measurement uncertainty. According to CISPR te mains cable sould be plugged into a mains outlet in te ground plane, wic means tat te common mode termination impedance is undefined. It will be sown next tat tis common mode termination impedance is an important source of measurement uncertainty. Influence of te CM termination impedance In tis section te influence of te common mode termination impedance on te RF emission, measured in a semi-anecoic room, will be investigated. We need a device under test (DUT) tat offers sufficient field strengt levels over te entire frequency range to enable good comparison wit te conducted emission results later on. Te applied DUT is a small box (5 cm x 8 cm x cm), containing a crystal oscillator wit a fundamental frequency of 0 MHz, positioned 80 cm above te ground plane. Te receiving antenna is placed at a orizontal distance of 3 m, in vertical polarization and at m eigt. Since te mains (adapter) cable does not yield sufficient emission levels, we will use te onboard battery supply and connect te signal output to a single vertical wire, wic is connected to te ground plane via a termination resistor (see Figure ). Figure sows te electric field strengt measured for te following values of te termination resistor: 0 Ω, 50 Ω, 50 Ω, 300 Ω, 500 Ω and open circuit. Because discrete spectra are difficult to compare, te levels are sown at te armonic frequencies only and connected by straigt line sections. A low CM impedance termination yields a systematic over-estimation of te field strengt at frequencies between 30 and 80 MHz and an under-estimation between 00 and 00 MHz. A ig impedance termination as te opposite effect. Te influence of te termination is small above 00 MHz. Te total variation in te curves is about 5 db. Systematic errors of tis kind could be reduced by coosing a common mode termination impedance tat lowers te VSWR on te cable, as will be discussed later on. 3 Te conducted RF emission test set-up In te conducted test set-up te DUT is placed 0 cm above a large reference plane; see Figure 3. Te signal output is connected via a single sort cable to a CDN, wic offers a common mode termination impedance of 50 Ω. Te spectrum of te common mode current troug te CDN test port is measured as a voltage over te 50 Ω input impedance of te spectrum analyzer. People wo prefer te results in terms of terminal voltage sould add 0 db for te 50 / 50 Ω conversion.

3 Spectrum Analyzer 0 MHz oscillator folding or meandering of te cable sould be avoided. For accurate measurements at ig frequencies eiter te cable lengt sould be kept as sort as possible or te caracteristic impedance of te cable above te reference plane sould be close to 50 Ω (i.e. te ratio of te cable eigt and its diameter sould be about 3). CDN cable 0 cm Figure 3. Reference plane Conducted RF emission test set-up Measurement uncertainties of te conducted set-up Te battery-powered 0 MHz oscillator was measured in a conducted test set-up according to IEC Because we want to compare te result wit te results from te SAR measurements, te common mode current generated by te DUT was not measured on te mains wire but on te signal output. Te output wire was connected to a CDN (type S, were te 50 Ω coupling occurs via te sielding of a single coaxial wire). Figure 5. Variation in te spectrum measured at te CDN normalized to te result obtained at L 0 cm. (solid curves: eigt 0 cm solid, dased curves: eigt 5 cm) Figure 4. Variation in te spectrum measured at te CDN wit a varying set-up geometry (see text). Te results sown in Figure 5 can be explained by means of transmission line teory. Consider a transmission line wit a caracteristic impedance Z 0,fed by a current source I S and terminated in a load impedance Z L. Te ratio of te load current I L wit respect to te source current is given by I L () IS cos( β L) jsin( βl) Z L Z0 If we relate te load current measured wit a certain lengt of cable to te load current tat would ave been measured wit zero lengt, we obtain curves as sown in Figure 6 for Z 0 /Z L, L 0 cm, 5 cm and m. In tis example te error wit L 0 cm remains witin db. Figure 4 sows te variations in te measured spectrum due to some geometry variations in te set-up: - cable lengt was 0 cm, 5 cm or m - cable straigt or meandering (only wit m lengt) - eigt of DUT and cable was 5cm or 0 cm. Below 00 MHz te variation in te results is less tan 5 db. At iger frequencies te set-up geometry becomes more important. It can easily be sown tat te variations in Figure 4 are of a systematic nature. Assume tat te curve measured at 0 cm distance yields te true common mode current generated by te DUT and use tis curve to normalize te oter ones. Te resulting curves are sown in Figure 5. Variations in te eigt of te DUT and te cable yield only small deviations ( or db). Tey are caused by te fact tat te common mode current itself will be sligtly iger wen te DUT is brougt closer to te reference plane. Variations in cable lengt and routing are more severe. In particular Figure 6. Error in measured common mode current for various cable lengts wit Z 0 /Z L. Furter uncertainties may be encountered in te common mode impedance of te CDNs temselves (see IEC

4 for specifications). CDNs constructed for a restricted frequency range could ave narrower tolerances tan te ones tat are intended for te full range from 50 khz to 30 MHz. te absorbing clamp as te opposite effect. A CDN gives a rater smoot translation curve between conducted and radiated test results. If te mains cable as a connector for wic no CDN is available, te EMclamp is a good alternative. 4 Comparison of conducted and radiated test results In te previous sections we ave found te spectrum of te electric field strengt E in te radiated set-up and te common mode current I cm in te conducted set-up. If we normalize te E-field curves of Figure to te common mode current corresponding to te curve wit L 0 cm from Figure 4 we obtain Figure 7, again for a set of termination resistors of 0 Ω, 50 Ω, 50 Ω, 300 Ω, 500 Ω and open circuit. Figure 8. Measured common mode impedance of practical termination devices. Figure 7. Relation of field strengt E z to common mode current I cm for various termination resistances. It is clear tat te translation curve from conducted to radiated test results depends on te CM termination impedance used in te radiated set-up. An average translation curve is obtained wit 50 Ω or 300 Ω. In practice we can of course not apply a resistive termination to te mains wire of a device under test. In our EMC test lab we ave te following termination devices at our disposal: a LISN, various CDNs, an absorbing clamp ( MDS clamp) and an EM-clamp ( injection clamp). Te measured frequency dependency of te CM impedance of tese devices is sown in Figure 8. Te CDNs and te EM-clamp are designed to ave a CM impedance of 50 Ω (wen terminated properly). A LISN is intended for te frequency range below 30 MHz. It sows a low CM impedance between 5 and 00 Ω for frequencies up to 00 MHz. Te absorbing clamp as a rater ig impedance below 60 MHz. Te radiated RF emission measurements for te small oscillator were repeated wit four practical termination devices. Te relation between te radiated and conducted test results is sown in Figure 9, were we recognize tat te LISN gives a systematic overestimation at frequencies between 30 and 80 MHz and an under-estimation between 00 and 80 MHz. Using Figure 9. Relation of field strengt to common mode current for practical termination devices. So, applying an appropriate termination device in te radiated test set-up can reduce te uncertainties due to resonance effects on te mains cable and improve te agreement wit conducted measurements. 5 Numerical model of te radiated test set-up In order to gain more insigt into te relation between te measured electric field strengt and te measured common mode current sown in Figure 7 and Figure 9, a numerical analysis as been carried out, using te antenna software package EMIR [3]. In te model an equivalent voltage source is placed between te vertical wire and te DUT, wic is modelled as a rectangular conducting plate. Te resulting vertical electric field strengt E z at 3 m distance, normalized to te current at te DUT output port, is sown in Figure 0 for various values of te CM termination resistor (0 Ω, 50 Ω, 50 Ω, 300 Ω, 500 Ω and open circuit).

5 and still ave some margin in practice. Te CISPR class B limit of 40 dbµv/m at 3 m distance can tus be translated into a common mode current limit of 0 dbµa or a terminal voltage of 54 dbµv in te range from 00 to 30 MHz. Te limit is 0 db iger at 30 MHz and varies linearly wit te logaritm of te frequency between 30 and 00 MHz. Figure 0. Relation between field strengt and current obtained by numerical modelling. Te calculated results sown in Figure 0 are quite similar to te measured results of Figure 7. Again, we find a smoot translation curve for moderate values of te termination impedance (in tis example 300 Ω yields te best curve). One more question as to be investigated. In tis modelling section te electric field strengt and te current at te DUT output ave bot been considered in te radiated set-up, wereas in te previous sections te common mode current was taken from measurements in te conducted set-up. Weter te common mode current on te mains cable is approximately te same in bot set-ups depends on te size of te DUT and on te coupling mecanisms involved. In bot cases an RF signal is generated at te DUT. A common mode current can flow via te cable to te ground plane and te loop is closed via a capacitance from (part of ) te DUT to te ground plane. If te coupling mecanism is voltage-driven, we ave to consider te stray capacitance from te disturbing voltage node to te ground plane. Tis stray capacitance does not vary muc for eigts between 0 cm and 80 cm in te case of a PCB wit a side lengt smaller tan 0 cm (factor G in [3]). If te side lengt is larger, tere will be some difference in te stray capacitance between te two set-ups and te orientation (track up or down) will play a role. In te case of a current-driven coupling mecanism we ave to consider te total capacitance between te PCB and te environment. Wit a small PCB (0 cm square) te common mode current will be about.5 db iger at 0 cm tan at 80 cm eigt. Wit a large PCB (50 cm square) te difference will still be less tan 6 db. Figure. Relation between E z and I DUT obtained wit equation (5), using 0.8 m, d 3 m, s m. Straigt lines are an asymptotic boundary curve Appendix Derivation of a simple analytical model In te appendix a very simple analytical model is derived, based on te assumption of a sinusoidal current distribution on te vertical wire. Figure sows some results of tis analytical model for a pure traveling wave (Γ 0) and for pure standing waves (Γ ±). In te case of a traveling wave te relation E z /I DUT as an almost constant level of 30 dbω/m for te frequency range from 00 to 30 MHz and it is approximately proportional to frequency from 30 to 00 MHz. At 30 MHz te results obtained wit te analytical and numerical models are in good agreement. Above 00 MHz tere is about 4 db difference. Tis is probably due to te fact tat te real current distribution is not perfectly sinusoidal (and te analytical model does not include radiation losses). So, if we define an asymptotic boundary curve, proportional to frequency between 30 and 00 MHz and constant at 30 dbω/m between 00 and 30 MHz (as indicated in Figure ), we can use tis curve to translate conducted measurement results into radiated results Z Z Figure. DUT I(z) image R a R R b Geometry of te radiated set-up. Assume tat te device under test (DUT) is small wit respect to te considered wavelengts, so only te mains wire yields a contribution to te radiated field. Suppose tat te mains wire runs vertically (z-direction) and tat te current distribution on tis wire can be d ground plane s

6 described by means of a sinusoidal distribution. In tat case we can apply a general formula from antenna teory [4], wic states tat te electric field component in te z-direction due to suc a wire segment can be written in terms of te current and its derivatives at te ends of te wire segment: E were ( ) R d z z and R ( ) d z z are te distances from te two end points of te wire segment to te observation point z at a orizontal distance d; see Figure. η0 0π Ω and k π / λ (for wave propagation in 0 free space). A sinusoidal current distribution on te mains wire can be written as jβz jβz e I( z) I 0 ( z > 0) (3) e were I 0 is te current at te DUT port (z ) and β π / λ describes te wave propagation on te wire. Γ relates te amplitudes of te waves in positive and negative z-directions. In transmission line teory tis would correspond to te voltage reflection coefficient at z 0. In tis simple model it is not necessary to know te caracteristic impedance Z 0 of te vertical transmission line. () Te effect of te ground plane can be included by assuming a current distribution on an image wire, wic satisfies te requirement I(-z) I(z), or (4) Equations (), (3) and (4) can be combined in a straigtforward manner to obtain te vertical component of te electric field at z s. Usually only te far field terms are of interest: E z jkr η 0 e 4π jkr z _ far jkr e jkr were d ( s ), R d ( s ) b R a d s. R I di dz jβz jβz e 0 < 0 e ( z) I ( z ) jβη0i 4π z 0 jkr z z e I z R jkr jkr jkr z z e I R jkr jkr di dz e e e Γe ( Γ) jβ jkr e jkr jkr e jkr jkr e jkr (5) and a a ( z ) b b ( z ) In te case of near field conditions (wit a measuring distance d 3 m tis occurs at f < 0 MHz) te following near field terms sould be added: jkr b η 0I 0 s e E z _ near 4π Rb jkrb jkrb jkr a s e Ra jkra jkra An example of te results is sown in Figure. 6 Conclusions Radiated RF emission tests according to CISPR are not convenient for testing during te early pases of a design. Te conducted test set-up described in IEC is more appropriate. For small devices wit a single cable a teoretical relation between radiated and conducted RF emission test results can be found, wic corresponds well to te measured relation, provided tat te effects of resonances in te radiated set-up are suppressed, e.g. by applying a CM termination to te cable. Te teoretically derived ideal translation curve yields sufficient margin to allow its application in less ideal practical situations. We agree wit Leuctmann et al. [5] tat it is not reasonable to state tat te radiated metod is te only reference, because ten we would be forced to reproduce te systematic errors from te radiated set-up into te conducted metod, or end up wit unpractically large margins. However, we tink tat CDNs are more suitable tan an MDS clamp in a design lab. Furter investigations sould include statistical evaluation of a number of products of different sizes to ceck te validity of te ideal translation curve and find out ow to proceed wit products aving more tan one cable. (6) References [] B. Szentkuti, Give up radiation testing in favour of conduction testing, 8 t Int. Symposium on EMC, Zuric 989. [] T.H. Hubing and J.F. Kaufman, Modelling te electromagnetic radiation from electrically small table-top products, IEEE Transactions on EMC, Vol. 3, No., 989, pp [3] J.R. Bergervoet, G.P.J.F.M. Maas and M.J.C.M. van Doorn, Te common-mode skeleton model for assessment of electromagnetic compatibility at te system level. t Int. Symposium on EMC, Zuric 997. [4] S.A. Scelkunoff and H.T. Friis, Antennas, Teory and Practice, J. Wiley, 95. [5] P. Leuctmann, H. Ryser and B. Szentkuti. Conducted versus radiated tests - Numerical field simulation and measured data. 9 t Int. Symposium on EMC, Zuric 99.