Electromagnetic compatibility aware design and testing of intermodulation distortion under multiple...

www.ietdl.org

Published in IET Science, Measurement and TechnologyReceived on 17th October 2011Revised on 11th January 2012doi: 10.1049/iet-smt.2011.0198

ISSN 1751-8822

Electromagnetic compatibility aware design andtesting of intermodulation distortion under multipleco-located sources illuminationA. Biondi F. Declercq D. De Zutter H. Rogier L. VallozziInformation Technology Department, Ghent University, St. Pietersnieuwstraat 41, 9000 Ghent, BelgiumE-mail: [email protected]

Abstract: Current electromagnetic immunity tests mainly rely on single-frequency sources. However, the evolution of electronicsystems leads to miniaturisation and low-cost solutions, in which filters are omitted in front of active non-linear components, alsofor efficiency reasons. As a result, intermodulation products may leak into the band of operation. The authors propose acomprehensive strategy consisting of design and test methodologies to evaluate in-the-band leakage of out-of-band undesiredcomponents, using multiple-tone excitation and relying on an anechoic chamber as test facility. The aim of this study is todemonstrate that an anechoic chamber together with a dual-source network analyser represents an optimal facility toinvestigate signal integrity issues owing to leakage of intermodulation products.

1 Introduction

In the last few years, a great interest has emerged in thedevelopment of compact, low-cost, low-power andreconfigurable microwave systems. This evolution led tominiaturised, low-cost solutions, in which filters in front ofactive devices [1] may be omitted. For instance, in softwaredefined radio, a very broadband signal is received by theantenna and directly amplified, before being converted tothe digital domain, where further processing, includingfiltering, takes place [2]. Non-linearities in these devicescombined with the lack of shielding and filtering maycompromise performance, signal integrity andelectromagnetic (EM) compatibility [3–5] owing to in-the-band leakage of out-of-band undesired signals interferingwith the desired signal. Indeed, when two or more signalsare present at the input of a non-linear component, thesignal at the output of that component includes otherfrequency components in addition to the wanted signalsbecause of intermodulation (IM) distortion. This issue inparticular arises when several sources illuminate devicesunder test (DUT), where none of the sources willindividually compromise signal integrity (SI) or EMimmunity (EMI) but the total effect of these sources is suchas to potentially cause SI and EM compatibility (EMC)problems [6].

Therefore in contrast to existing established EMC tests,it is important to develop and understand approaches to testelectronic systems under multiple-sources illumination, inparticular when IM products occur that might potentiallydisturb the DUT and cause SI/EMI issues. Existing EMCtests rely on single-frequency sources, and thereby do notalways provide a realistic transmission scenario [6, 7].

IET Sci. Meas. Technol., 2012, Vol. 6, Iss. 2, pp. 105–112doi: 10.1049/iet-smt.2011.0198

1.1 Contributions

Hence, in this paper, we present a comprehensive experimentalstrategy to model and test radio frequency (RF) devices undermultiple-frequency sources illumination, in the presence ofout-of-band frequency components leaking into the band ofinterest. Specifically, the novel contributions are as follows:

1. A new EMC aware computer-aided design strategy isoutlined providing guidelines to account for in-the-bandleakage of undesired out-of-band signals because of IMdistortion. It is demonstrated how the current use ofstandard parameters describing non-linearities, such asP1dB, OIP2 and OIP3 often fail to capture leakage effects.2. Novel precompliance tests using a dual-source vectornetwork analyser are proposed at the subsystem/device level toquickly identify and correct for IM distortion during the design.3. New EMC tests are defined to characterise in-the-bandleakage of a DUT, using an anechoic chamber and multipleco-located interfering sources transmitted along the mainbeam of the illuminating antenna, as such representingthe worst-case transmission scenario. Test procedures areoutlined where the DUT is tested at different radiated powerlevels and at varying distances.

To the best of authors’ knowledge, this is the first paperpresenting a real comprehensive multi-frequency modellingand test strategy in the EMC context.

1.2 Organisation of the paper

First, the different mechanisms that lead to in-the band leakageof undesired signals via IM distortion are discussed in the next

105

& The Institution of Engineering and Technology 2012

www.ietdl.org

section. This allows us to define modelling and test proceduresduring the design, precompliance as well as compliancetesting phases. Sections 3 and 4 discuss the practicalimplementation of this strategy for a representative circuit,consisting of an antenna that picks up the disturbing signals ina frequency-selective way combined with a low-noiseamplifier (LNA) that creates intermodulation distortion (IMD)because of its non-linearity. Specifically, in the design andprecompliance testing phase, as discussed in Section 3, thefocus is on the non-linear device, tested and modelled in a50 V environment. Section 4 focuses on testing the completeDUT for EMC compliance. Section 5 summarises theconclusions of the paper.

2 Accounting for IM distortion in EMC awaredesign and testing

In this section, we put forward a strategy to evaluate thebehaviour of a DUT (Fig. 1) under multiple-sources’illumination. In particular, we must account for all differentcauses of IM products that might give rise to in-the-band leakage.

2.1 Causes of IM distortion

Electronic devices are susceptible to different kinds ofdisturbances depending on their nature. In order for acomponent device or system to be susceptible to multi-frequency signals it must consist of a concatenation of asusceptible linear device, and a non-linear componentcreating IM products. Specifically, it is the absence of afilter in between the susceptible device and the non-linearcomponent, as shown in Fig. 1, which allows undesiredout-of-band signals to leak into the frequency band ofoperation. To provide a clear view to the EMC awaredesign and test engineer, we first present the differentscenarios resulting in IM distortion. Each cause of IMproducts, leading to potential in-the-band leakage ofunwanted signals because of IM distortion, corresponds to aspecific multi-frequency transmission scenario that will berecreated in the design, precompliance and compliancetesting phase by means of modelling and experiments.Fig. 2 shows the case in which a transmitter illuminates theDUT with a single-frequency component (tone) at f0, and ahigher-order (the second-order in the figure) harmonic fallsinto the band of interest as unwanted signal (scenario 1).Once this occurs, the unwanted signal can no longer beremoved by filtering. Therefore it is important to testwhether the power of this in-the-band disturbance remainssmall enough compared to the desired signal, in order notto affect the EMC performance of the DUT.

Fig. 1 General DUT

Fig. 2 Propagation of harmonics in the output spectrum of theDUT (scenario 1)

106


The second scenario, represented in Fig. 3, depicts the caseof two tones, f1 and f2, a desired one f1 falling in the bandand an unwanted one f2 falling out-of-band, radiated by theco-located sources illuminating the DUT. In this scenario,the two frequency components result in an intermodulationproduct falling in the band of operation of the device.Again, the power of the in-the-band interference should bemonitored to validate the EMC behaviour of the DUT. Notethat also two undesired out-of-band signals generatingrelevant in-the-band interference should be considered inthis scenario, although such a configuration is not illustratedin the sequel for reasons of conciseness.

Finally, Fig. 4 depicts a third scenario in which a high-power out-of-band transmitter (frequency f2) is illuminatingthe DUT. Owing to its high power, the LNA does nolonger operate as expected. The LNA gets desensitised, thatis, gain and linearity decrease, resulting in a desired signalthat is less amplified and hence more difficult to detect.Also this scenario should be accounted for in EMC awaredesign and suitable tests must be established to ensure EMcompatibility of the DUT.

At this point we should make two important remarks. First,from Fig. 4 it is seen that the undesired signal that desensitisesthe amplifier may also result in important in-the-band IMproducts, as studied in scenario 2. In such a case, thedesensitisation will further decrease the ratio of the powerof the desired signal over the power of the IM distortion, asthe wanted signal gets less amplified whereas the unwantedsignal increases more than expected with rising signalslevels. Second, a strong out-of-band signal may desensitisethe DUT without producing IM products in the desiredfrequency band. This process may, however, make theDUT more sensitive to IM products produced by other out-of-band signals. This, rather complex, scenario is notfurther investigated for reasons of conciseness.

2.2 Design and precompliance/compliancetesting strategies

To be able to include IM distortion during the EMC awaredesign phase, the interfering multi-frequency signals willbe applied as sources in a full-wave/circuit co-simulation,

Fig. 3 Response of the DUT to an in-the-band tone f1 and anout-of-band tone f2 (scenario 2)

Fig. 4 Desensitisation because of an out-of-band tone f2(scenario 3)


www.ietdl.org

whereas during precompliance tests these simulated resultsare verified at the subsystem and module levels byperforming direct multi-tone injection by means of anAgilent N5242A PNA-X network analyser. Testing allsubsystems/modules separately provides a quick way forearly diagnostics during design and precompliance testing,but it does not provide a complete picture, since input and/or output of the subsystem/module under test act asfrequency-dependent loads over the broad frequency rangeover which unwanted signals will impinge, and this is nottaken into account when using a network analyser, whichtests the subsystem/module in a constant 50 V environment.

Therefore in the final compliance tests the multi-frequencysignals are transmitted by a standard gain horn lettingthem impinge in the far-field on the complete DUT. InFig. 5 we propose the setup for such a multi-frequency testto be performed in an anechoic chamber. The transmissionsource in the experiment is a standard gain horn, connectedto a highly-linear power amplifier, and finally connectedto an Agilent N5242A PNA-X network analyser used asmulti-frequency generator. The choice of the poweramplifier is crucial as will be discussed in Section 4. As weare dealing with non-linear phenomena, varying powerlevels must be considered. Instead of varying the powerlevel we may make the distance between the transmitter andthe DUT variable, where the lower bound of that distancemust still respect the far-field condition. In the sequel, thein-the-band frequency range is defined by its lower fL andits upper frequency fH. In our experiments we focus on theISM-band extending from 2.4 GHz up to 2.4835 GHz, butfor the ease of the discussion we choose fL ¼ 2.4 GHz to bethe lower bound and fH ¼ 2.5 GHz to be the upper boundof the frequency band of interest.

In the remaining part of this paper, by way of example, wewill illustrate the aforementioned strategy to model and testthe three different IM distortion scenarios by considering asimple susceptible passive component, combined with asimple active component that causes IMD. The DUT(Fig. 6) we consider is an active textile antenna [8] wherethe susceptible component is the passive antenna, and theactive component is a LNA. This example is of practicalimportance since in many such designs the presence of afrequency selective filter placed between the antenna andthe LNA is omitted to reduce cost and to obtain a morecompact and more efficient device. The LNA is directly fedby the antenna signal through a via without any matching

Fig. 5 Anechoic chamber setup


network [8], choosing as antenna impedance the optimalimpedance (35.05 2 j17.644)V for minimum noise figure[8]. No filtering operation is provided by intermediatestages in the DUT.

In the next section, we first consider the EMC aware designand precompliance testing phase, which focuses on thebehaviour of the LNA as a stand-alone subsystem/module.The final compliance testing of the complete DUT, beingthe active antenna, is discussed in Section 4.

3 EMC aware design and precompliancetesting of the LNA

In contrast to the compliance testing phase, during EMCaware design and precompliance testing, the designengineers have the possibility to model, measure andmanipulate the generation and propagation of IM productsin isolated subsystems and modules. Hence, including IMdistortion in the EMC aware design and precompliancetesting cycles offers a cheap and efficient way to avoid in-the-band leakage of undesired signals in the final product.Let us now discuss our strategy by focussing on the LNAmodule of the active antenna. Models and measurements inthis section are based on an LNA that is terminated by50 V at both input and output ports.

LNAs are traditionally characterised by their OIP3, OIP2and P1dB levels. Table 1 compares the results obtainedthrough simulations with Agilent’s Advanced DesignSystem to the values measured with the Agilent N5242APNA-X Network Analyzer at 2.45 GHz. A good agreementis obtained between simulated and measured data. However,the interpretation of these numbers requires awareness ofthe specific modelling/measurement setup that led to eachof these results. We note that P1dB was obtained byinjecting a single tone at 2.45 GHz, whereas two in-the-band tones spaced by 10 MHz and centred about 2.45 GHzwere injected to find OIP3. Yet, to study OIP2, a singletone, now at 1.225 GHz, is again considered, observing thecircuit’s response at 2.45 GHz. In reality, first, the inputpower levels will never be that high as to, for example,reach the third-order interception point. Second, the actualnon-linear circuit will exhibit a frequency-dependentbehaviour, given that all components possess frequency-dependent characteristics, specifically when considering theeffect of unwanted out-of-band components that may be

Fig. 6 Passive antenna and LNA

Table 1 Main tones power, OIP3, OIP2 measured and simulated

Value, GHz Simulated, dBm Measured, dBm

P1dB@f0 ¼ 2.45 –0.601 –0.498

OIP3@f1 ¼ 2.445 f2 ¼ 2.455 +27.76 +25.526

OIP2@f0 ¼ 1.225 +72.394 +61.763

107


www.ietdl.org

present in a very wide frequency band, much wider thanintended by the design engineer. Hence, in EMC awaredesign and testing, we must carefully analyse the problemfor different tone spacings and power levels.

Therefore to characterise the effect of out-of-band signals,the spurious free dynamic range (SFDR) is put forward as asuitable measure. In literature [9, 10], several definitions ofthe SFDR can be found. For our purposes, we define SFDRas the ratio between the desired component and the highestspurious one leaking into the band(s) of operation, asschematically shown in Fig. 7. In EMC aware design anupper limit for this SFDR should be set and modelled/testedfor different combinations of desired and/or unwantedsignals. In this paper, by way of example, we will considera spurious component to be acceptable as long as the SFDRis higher than 40 dB. However, it must be kept in mind thatthe optimum threshold is strictly application-dependent.

We first discuss the relevancy in terms of EMC awaredesign of the modelling/measurement data that led to theresults of Table 1. These data are the gain at 2.45 GHz, theSFDR for two tones centred at 2.45 GHz ( f1 ¼ 2.445 GHzand f2 ¼ 2.455 GHz) and the power of the second-orderharmonic modelled/measured for varying power levels, asshown in Fig. 8. A good agreement is found betweensimulated and measured gain and for the power ofthe second-order harmonic but less so for the SFDR. TheSFDR results refer to the ratio between the in-the-band toneand the highest in-the-band intermodulation product. Themeasured SFDR value stays above 40 dB for an inputpower lower than –3 dBm. For the sake of completeness,PNF (power noise floor) and PMDS (power minimumdetectable signal) were also measured by means of the

Fig. 7 Worst-case identification for the calculation of the SFDR

Fig. 8 Gain at f0 ¼ 2.45 GHz, SFDR for f1 ¼ 2.445 GHz andf2 ¼ 2.455 GHz, and second-order harmonic power for f0 ¼ 1.225 GHz,simulated and measured

108


Agilent N5242A PNA-X Network Analyzer and turn out tobe equal to –90 dBm and –52 dBm, respectively.

Although the gain plot demonstrates densensitisation athigher input levels, this result is not relevant for ourpurposes as it is only caused by in-the-band signals. Hence,extra filtering will not affect this curve. In the same way,the SFDR result, produced by two in-the-band components,will not change when introducing filters before thenon-linear component. Hence, for an EMC engineerstudying in-the band leakage of undesired components, onlydata related to the second-order harmonic offers relevantinformation. To that end an SFDR must be introducedcomparing the power of the second-order harmonic leakinginto the band to the typical power levels of the desiredsignals encountered in the application.

Therefore we now focus on two tones’ excitations thattypically result in important in-the-band leakage whenomitting a filter before the non-linear component. The firstregions that require our attention are the boundaries of thefrequency band(s) of operation. Fig. 9 shows the outputspectrum of the LNA when excited with two tones: in thetop figure these tones are located at f2 ¼ 2.395 GHzand f1 ¼ 2.405 GHz (i.e. a spacing of 10 MHz centred atthe lower end of the ISM band), whereas in the bottomfigure the two tones are centred around the upper boundof the band, at f1 ¼ 2.495 GHz and f2 ¼ 2.505 GHz. Thevalues given in Fig. 9 are for an input power level ofPin ¼ P1dB, with P1dB ¼ –0.498 dBm (see Table 1). Thishigh-input power level was deliberately selected in order toobtain the worst case results. The results of Fig. 9 clearlyshow the existence of unwanted in-the-band tones.However, the SFDR remains above 40 dB (+43.54 dB).

Next, the power level of the out-of-band tone is kept atP1dB but the in-the-band tone power level is decreased to–10 dBm. This is considered to be relevant for the testphase because in a real transmission context it is very likelythat an out-of-band component can be stronger in powerthan the in-the-band one. The obtained results are shown inFig. 10. We observe a difference of 59.35 dB between thein-the-band tone and the highest spurious component fallingwithin the band. Hence, the SFDR is again acceptable.Besides these example configurations, other two-toneexcitations, even two out-of-band signals, may result in

Fig. 9 Output spectrum with Pin ¼ P1dB at the lower and at theupper side of the considered frequency band, for Df ¼ 10 MHzand excitation by two equal power tones f1 and f2


www.ietdl.org

relevant in-the-band leakage. However, filtering thanks to thefrequency dependency of the susceptible linear componentmay reduce the power of these leaking IM products, asdiscussed in the next section.

4 Compliance testing of the active antenna

Let us now turn to the compliance testing phase of the DUT(Fig. 11), in which all relevant scenarios leading to in-the-band leakage should be verified. To generate a strongenough incident signal, two highly linear power amplifiersare cascaded: the highly linear power amplifier containedin the Agilent N5242A PNA-X Network Analyzer anda Mini–Circuits ZRL–3500 power amplifier providing amaximum gain of GPA ¼ 21 dB. The choice for that type ofamplifier is not arbitrary. Indeed, we must make sure thatits linearity is superior to that of the LNA, which is a partof our DUT. If this would not be the case, our test set-upwould indeed not make sense. This makes clear thatthe choice for a particular additional amplifier in themeasurement chain depends on the specific DUT one isinterested in. The Mini–Circuits ZRL–3500 poweramplifier (further referred to as MC amplifier) has an OIP3of +45 dBm. The maximum allowed input power isspecified to be +10 dBm. However, when using the MC

Fig. 11 DUT measurement setup in the anechoic chamber

Fig. 10 Measurement results as in Fig. 9 but now for two toneswith a power difference of 10 dBm


amplifier in a cascade with the PNA-X amplifier, carefulexperiments have revealed that this input power must befurther restricted in order to drive the second amplificationstage without generating IMD, distortion that could perturbthe measurement results of the DUT. Several tests wereperformed to characterise the amplifier cascade in terms oflinearity, by driving the input of the MC amplifier with avariable power (emitted by the PNA-X) ranging from –20to 0 dBm. Up to 0 dBm no significant IMD turns up andhence this 0 dBm will be the highest input power level atthe MC amplifier used for the DUT measurements.

To define the relevant power range of desired andunwanted signals incident from the standard gain horn ontothe DUT we rely on the equivalent isotropically radiatedpower (EIRP) [11], being the amount of power that thestandard gain horn emits to produce the peak power densityin the direction of maximum antenna gain. Referring toFig. 11, the EIRP is given by

EIRP[dBm] = Pin[PNA−X ][dBm] + GPA[dB] + GHORN[dBi]

− L[cables][dB] − Mt[dB] (1)

with GHORN the gain of the standard gain horn (ScientificAtlanta SGH-1.1 or 1.7, depending on the desired excitationfrequency), Lcables the cable losses and with Mt representingthe transmission mismatch. The relevant numbers at2.45 GHz are summarised in Table 2. We must also takeinto account that the maximum allowed EIRP that may beradiated by emitters in the 2.45 GHz ISM band is limited inEurope to +20 dBm by ETSI EN 300 328 and by the USFCC Rules, Part 15.247, to +36 dBm.

The EIRP is related to the signal at the output of the activetextile antenna by means of

Pout[dBm] = EIRP[dBm] − PL[dB] + GDUT[dB]

− Mr[dB] (2)

with PL[dB] the path loss, depending on the distance dbetween SGA and DUT (Fig. 11), Mr the mismatch factorat the receiver and GDUT[dB] ¼ GAntDUT × GLNA the gainof the complete DUT. GDUT is the product of GAntDUT,the gain of the passive susceptible device and GLNA, thetransducer gain of the non-linear component with theantenna as its source. Although (2) relates to the linearbehaviour of the DUT, the generation of IM products isdirectly related to the output level of the DUT. Hence, theIM distortion will be influenced by the frequency selectivityof the path loss and, more importantly, by GDUT. Thereforeit is important to first characterise the gain of the DUT as afunction of frequency through a separate S-parametermeasurement, by

GDUT[dB] = S21meas[dB] − Mr[dB] − Mt[dB] + GPA[dB]

− GHORN[dBi] + PL[dB] (3)

Table 2 Link budget at 2.45 GHz

Pin[PNA2X ] G[PA] G[HORN] Mt, dB

220:1:0, dBm 21, dB 15.35, dBi 20.1016

L[cables] G[LNA] G[antDUT] Mr, dB

20.06, dB +11, dB +3.5, dBi 20.2174

109


www.ietdl.org

S21meas represents the measured forward transmissioncoefficient. The top graph of Fig. 12 shows the measuredGDUT as a function of frequency. To characterise the effectsof frequency-selective gain, we define the transmissionattenuation (TA) as the absolute value of the differencebetween the maximum value of GDUT and GDUT as afunction of frequency. The resulting TA is shown in thebottom graph of Fig. 12. For these measurements we tookthe distance d to be 10l1, with l1 the wavelength at2.45 GHz. We observe that the DUT already provides 3 dBof attenuation at 2.385 GHz and 2.5 GHz, 10 dB at 2.2 GHzand 2.6 GHz and 15 dB at 1.7 GHz and 2.7 GHz. Asproposed in Section 2, we now investigate the behaviour ofthe DUT using the three proposed scenarios.

4.1 First scenario (Fig. 2)

In our measurement, f0 ¼ 1.225 GHz was used and whilesweeping the input power we observe the occurrence of anin-the-band harmonic at 2f0 ¼ 2.45 GHz. The test wasperformed for different distances between the standard gainhorn and the DUT. Fig. 13 shows the second-orderharmonic’s power over a range of distances from 8l2 downto 5l2 (where l2 is the free-space wavelength at1.225 GHz). The power of the second-order harmonicremains smaller than –52.47 dBm with d ¼ 5l2 for anEIRP of +20 dBm. To reach an SFDR of 40 dB that is ourproposed threshold limit, the desired signal should reach atleast –12.47 dBm for the DUT to pass this EMC test.

Fig. 12 GDUT and TA as a function of frequency, with d ¼ 10l1

Fig. 13 Second harmonic’s power against distance d between thestandard horn and the DUT

110


4.2 Second scenario (Fig. 3)

The two tones radiated by the horn in our experiment aref2 ¼ 2.395 GHz and f1 ¼ 2.405 GHz, with, as for the LNAtesting, a spacing of 10 MHz. They have the same power.The EIRP used for the experiments was varied from aminimum of +10 dBm up to +37 dBm. The distancebetween horn and DUT was varied from 9l1 to 12l1 (free-space wavelength at 2.45 GHz). The results shown inFigs. 14 and 15 refer to a distance of 9l1 and EIRP levels of+27, +30, +32 and +34 dBm. For an EIRP of +20 dBmthe intermodulation products just reach the noise floor(–90 dBm) and the level of the main tones is –27.31 dBm.In Fig. 16, SFDR values are reported, which rely on themeasurements performed at different distances betweenthe standard gain horn and the DUT. Considering an SFDRthreshold of 40 dB, and depending on the distance betweenthe radiating element and the DUT, we can state that theSFDR stays above the threshold for EIRP values smallerthan 32 dBm and for a distance larger than 9l1.

Similar results are obtained at the upper side of thefrequency band and for other distances. Table 3 summarisesthe power level of the largest in-the-band main tone PTone,1

and of the strongest in-the-band third-order intermodulation

Fig. 14 Output spectrum for the excitation by two equal powertones in the proximity of the lower side of the frequency band withEIRP ¼ +27 dBm and +30 dBm, respectively, using d ¼ 9l1

Fig. 15 Output spectrum for the excitation by two equal powertones in the proximity of the lower side of the frequency band withEIRP ¼ +32 dBm and +34 dBm, respectively, using d ¼ 9l1


www.ietdl.org

product PTone,3 for various EIRP values, at a distanced ¼ 9l1. It is seen that the SFDR drops below 40 dB forEIRP ≥ 32 dBm.

4.3 Third scenario (Fig. 4)

Let us now focus on how desensitisation affects the SFDR.As already mentioned, any strong undesired signal mayreduce the amplification of the desired signal and therebydecrease the SFDR. In most cases, the most criticalconfiguration will be encountered when a strong interferingsignal is present near the boundaries of the frequency bandof operation, simultaneously causing IM products anddesensitisation in the DUT. Therefore again consider theexperiment that resulted in Figs. 14–16. At the higherEIRP levels we clearly observe that desensitisation startsinfluencing the performance of the DUT. The power ofthe main tones no longer increases at the same rate as forthe lower EIRP values while the intermodulation productsgrow faster. To characterise this phenomenon an importantparameter describing the behaviour of the DUT under thevariation of EIRP is introduced: the tone power variationTPVn, defined as

TPVn =PTone,n

EIRP

∣∣∣∣PL=constant

(4)

In Fig. 17, TPV1 and TPV3 are shown as a function of EIRP,for different distances between transmitter and DUT. Inparticular, by calculating this variation we can identify thepower level at which the amplification starts decreasing,because of the high input power level, for each transmitter-DUT separation. When increasing the EIRP level we expectthe power of the main tones to increase by the same

Fig. 16 SFDR as a function of EIRP for two equal power tonesand varying distances (l ¼ l1)

Table 3 Largest in-the-band main tone PTone,1 and third-order

intermodulation product PTone,3 for varying EIRP values, at a

distance d ¼ 9l1

EIRP, dBm PTone,1, dBm PTone,3, dBm SFDR, dB

+20 227.3 Noise floor 62.7

+27 224.2 279.8 55.6

+30 221.2 267.2 46.0

+32 219.6 249.0 39.4

+34 218.3 239.7 21.4


quantity and the third-order intermodulation productsto grow cubically until the LNA contained in DUTdesensitises because of saturation, exhibiting lower gain,resulting in a higher growth rate for the intermodulationproducts and in a lower growth for the main tones. This isclearly visible in Fig. 17: around +30 dBm of EIRP (forthe closest distance between the test antenna and the DUTequal to 9l1) the LNA starts desensitising.

This effect will be even more pronounced whenconsidering a dual-tone co-located transmitter illuminatingthe DUT, as in the second scenario, but now assuming theout-of-band tone to be much higher in power than thedesired signal. In particular, here we choose this unwantedcomponent to be 10 dB higher in power. EIRP1 representsthe power level applied to the out-of-band tone, whileEIRP2 represents the power level applied to the in-the-bandcomponent. In the top part of Fig. 18 we observe thatthe intermodulation products occur at the same level as thenoise floor, similar to the case with equal power for theout-of-band and in-the-band component. In Fig. 19 we notethat by increasing the EIRP1 to resp. +32 and +34 dBm

Fig. 17 TPV1 and TPV3 as a function of EIRP and for variousdistances l ¼ l1 between horn and DUT

Fig. 18 Output spectrum for the excitation by two tones withdifferent power in the proximity of the lower side of the frequencyband and resp. for EIRP1 ¼ 27 dBm, EIRP2 ¼ 17 dBm and forEIRP1 ¼ 30 dBm, EIRP3 ¼ 20 dBm

111


www.ietdl.org

and EIRP2 to resp. +22 and +24 dBm, the in-the-bandintermodulation product rises to –62.25 and –51.07 dBm,respectively. We observe that the third-order intermodulationproduct grows faster than cubically owing to the influence ofthe out-of-band component. This phenomenon tends to bemore noticeable when increasing the EIRP level. Indeed, forEIRP1 ¼ +32 dBm we observe a difference of 2.39 dBbetween the expected in-the-band intermodulation productand the measured one. For the case EIRP1 ¼ +34 dBmthe difference equals 3.13 dB. This effect quickly reducesthe available SFDR. Indeed, for EIRP1 ¼ +32 dBm, theSFDR level equals 30.85 dB, whereas when going toEIRP1 ¼ +34 dBm, the SFDR drops to 18.98 dB.

5 Conclusions

Current EMC testing procedures have most often focused onsingle-source/single-frequency illumination of the DUT. Ingeneral there are no conventional test practices that involvemulti-source/multi-frequency testing. Therefore we haveproposed a new modelling and experimental approachdealing with IMD phenomena in a transmission context bymeans of multiple co-located sources. This new approach issuitable for any generic DUT composed of at least oneactive non-linear component and a passive susceptibleantenna. An active textile antenna (consisting of an LNAand a passive antenna) was used as test bench to provethe effectiveness of the new approach with the 2.45 GHzISM-band as the frequency band of interest.

By analysing and measuring the components composingthe DUT (in our case the LNA and the passive antennapart) and by subsequently measuring the entire DUT in

Fig. 19 Output spectrum for the excitation by two tones withdifferent power in the proximity of the lower side of the frequencyband and resp. for EIRP1 ¼ 32 dBm, EIRP2 ¼ 22 dBm and forEIRP1 ¼ 34 dBm, EIRP2 ¼ 24 dBm

112


several transmission scenarios using horn antennas in ananechoic chamber with a PNA-X network analyser asmulti-frequency source, it was shown how a better andmore correct knowledge about leakage of undesired signalsinto the DUT’s frequency band of operation whenilluminating with a single out-of-band frequency (scenario1) or a combination of an in-the-band and an out-of-bandfrequency (scenario 2) can be obtained. To characterise theeffect of harmonics of the out-of-band frequency turning upin the frequency band of interest, the SFDR was putforward as a suitable measure. Furthermore, the influence ofstrong transmitters, desensitising the LNA, on the relativeimportance of intermodulation products in the DUT’soutput spectrum was carefully examined (scenario 3). As ameasure for the desensitisation, we introduced the tonepower variation parameter, that is, the ratio of the power ofa particular tone to the EIRP of the standard gain horn.For these measurements, an additional highly-linear pre-amplifier (in our case the Mini-Circuits ZRL-3500 poweramplifier) was necessary. Concluding, it can be stated thatquantitative testing of active DUT’s with non-linearcomponents illuminated by multiple co-located sourcesemitting both desired, in-the-band, and undesired, out-of-band, signals can be successfully carried out using ananechoic chamber, standard gain horns, a PNA-X networkanalyser and a well-chosen linear preamplifier together witha set of clear measurement scenarios.

6 References

1 Rembovsky, A., Ashikhmin, A., Kozmin, V., Smolskiy, S.: ‘Radiomonitoring: problems, methods and equipment’ (Springer, 2009, 1st edn.)

2 Cruz, K.A.R.P., Carvalho, N.B.: ‘Designing and testing softwaredesigned radios’, IEEE Microw. Mag., 2010, 11, (4), pp. 83–94

3 I. T. 61000-1-2: ‘Electromagnetic compatibility (EMC) - part1–2:general-methodology for the achievement of the functional safety ofelectrical and electronic equipment with regard to electromagneticphenomena’, February 2008. Available: http://iec.ch

4 Mardiguian, M.: ‘Combined effects of several simultaneous, EMIcouplings’. 2000 IEEE Int. Symp. on Electromagnetic Compatibility,Vol. 1, Washington, August 2000, pp. 181–184

5 Nitsch, J., Korovkin, N., Soloveyna, E., Scheibe, H.-J.: ‘Occurrence oflow-frequency noises in electronic systems under action of two-tonehigh frequency electromagnetic excitation’. 2005 IEEE Int. Symp. onElectromagnetic Compatibility, Vol. 2, Magdeburg, August 2005,pp. 618–621

6 Duffy, A., Orlandi, A., Armstrong, K.: ‘Preliminary study of areverberation chamber method for multiple-source testing usingintermodulation’, IET Sci. Meas. Technol., 2010, 4, pp. 21–27

7 Armstrong, K.: ‘Functional safety requires much more than EMCtesting’. EMC-Europe, 6–10 September 2004, pp. 348–353

8 Declercq, F., Rogier, H.: ‘Active integrated wearable textile antennawith optimized noise characteristics’, IEEE Trans. Antennas Propag.,2010, 58, (3), pp. 3050–3054

9 Span Application Note F.: ‘Spurious free dynamic range’ [Online].Available at: http://www.fiber-span.com

10 Lawrence, E.: ‘RF and microwave circuit design for wirelesscommunications’ (Artech House, 1997, 1st edn.)

11 Balanis, C.A.: ‘Antenna theory: analysis and design’ (John Wiley &Sons, New York, 1997, 2nd edn.)


Electromagnetic compatibility aware design and testing of intermodulation distortion under multiple...

Documents

Transcript of Electromagnetic compatibility aware design and testing of intermodulation distortion under multiple...