measurement techniques of emissions from ultra · pdf filemeasurement techniques of emissions...

2252IEICE TRANS. FUNDAMENTALS, VOL.E88–A, NO.9 SEPTEMBER 2005

INVITED PAPER Special Section on Ultra Wideband Systems

Measurement Techniques of Emissions from Ultra WidebandDevices

Jun-ichi TAKADA†,††a), Shinobu ISHIGAMI†††, Juichi NAKADA††††, Eishin NAKAGAWA†††††,Masaharu UCHINO††††††, and Tetsuya YASUI†, Members

SUMMARY This paper describes the measurement techniques of emis-sions from UWB devices discussed in ITU-R task group (TG) 1/8 to studythe compatibility between ultra-wideband (UWB) devices and radiocom-munication services. This paper also provides the background idea behindthe measurement methods, as the final output of the discussion, i.e. ITU-RRecommendation, will not contain any citations to the references, nor any“educational” description of the theoretical background.key words: UWB devices, measurement, ITU-R

1. Introduction

Various new applications are considered by utilizing the ul-tra wideband (UWB) radio taking the advantages of lowpower consumption, high data rate, and fine position accu-racy. However, UWB signals are usually transmitted overthe frequency spectrum already in use by other systems andapplications, such as the radio communications and remotesensing. For the compatibility between the conventional ra-dio systems and the UWB devices, the emission limit of theUWB devices shall be restricted. The permissible emissionlevel has been discussed extensively between the proponentsof UWB systems and the potential victims. Therefore, it isessential to define the emission level of the UWB devices bythe measurements. However, it is not so straightforward tomeasure the emission characteristics of the UWB devices asthe ordinary carrier modulated signals.

In the International Telecommunication Union Radio-communication sector (ITU-R), task group (TG) 1/8 hasbeen established in July 2002 to study the compatibility be-tween ultra-wideband devices and radiocommunication ser-

Manuscript received May 10, 2005.Final manuscript received May 18, 2005.†The authors are with the UWB Technology Institute, Na-

tional Institute of Information and Communications Technology,Yokosuka-shi, 239-0847 Japan.††The author is with the Department of International Develop-

ment Engineering, Graduate School of Engineering, Tokyo Insti-tute of Technology, Tokyo, 152-8550 Japan.†††The author is with the Communication System EMC Group,

National Institute of Information and Communications Technol-ogy, Yokosuka-shi, 239-0847 Japan.††††The author is with the Advantest Corporation, Meiwa-machi,

Gunma-ken, 370-0718 Japan.†††††The author is with the Telecom Engineering Center, Tokyo,

140-0003 Japan.††††††The author is with the Anritsu Corporation, Atsugi-shi, 243-8555 Japan.

a) E-mail: [email protected]: 10.1093/ietfec/e88–a.9.2252

vices. TG 1/8 is divided into four working groups (WGs),i.e. characteristics in WG1, compatibility in WG2, spec-trum management framework in WG3, and measurementsin WG4.

The standard measurement techniques of emissionsfrom UWB devices have been discussed in TG 1/8 WG4.Presently∗, the preliminary draft new recommendation(PDNR) on the measurement to be entitled “Measurementtechniques of emissions from systems using ultra-widebandtechnology” (ITU-R SM.[UWB.MES]) is still under discus-sion. It should be finalized by the last meeting of TG 1/8 inOctober 2005.

This paper describes the measurement techniques pre-sented in the PDNR ITU-R SM.[UWB.MES]. Since the fi-nal recommendation has not been approved, the authors re-fer the Japanese contribution submitted to May 2005 meet-ing of TG 1/8 [1]. ITU-R recommendation does not con-tain any citations to the references, nor any “educational”description of the theoretical background. Therefore, thepaper provides the background idea behind the measure-ment methods, which will not be available on the PDNR.Although the techniques proposed by the authors are moreemphasized, all the techniques are reviewed. Numerical ex-amples are limited to the microwave frequency, although thePDNR includes those for the quasi-millimeter and millime-ter waves.

Section 2 presents two different philosophies of themeasurements. Section 3 describes a concept of spectrumemission mask. If the emission mask is deployed, some pa-rameters normally applied to emissions, such as occupiedbandwidth and unwanted emissions, do not need to be spec-ified. Section 4 compares the frequency domain and timedomain measurement approaches. Section 5 lists the pa-rameters and their units to be measured. Section 6 describesthe measurement conditions and environments. Section 7presents the detailed measurement techniques of the emis-sion in the frequency domain, which Sect. 8 present those inthe time domain. Finally, Sect. 9 summarize the paper.

1.1 Glossaries

CAF : complex antenna factorDFT : discrete Fourier transformDUT : device under test

∗This paper is written in April 2005.

Copyright c© 2005 The Institute of Electronics, Information and Communication Engineers

TAKADA et al.: MEASUREMENT TECHNIQUES OF EMISSIONS FROM ULTRA WIDEBAND DEVICES2253

EIRP : equivalent isotropic radiated powerEMC : electromagnetic compatibilityFCC : Federal Communications Commission, USAFFT : fast Fourier transformIBW : impulse bandwidthITU-R : International Telecommunication Union Radio-

communication sectorLNA : low noise amplifierLO : local oscillatorPAPR : peak to average power ratioPDNR : preliminary draft new recommendationPRF : pulse repetition frequencyQP : quasi-peakPDF : probability density functionPSD : power spectral densityRBW : resolution bandwidthRMS : root-mean-squareTG : Task GroupTRP : total radiated powerVBW : video bandwidthWG : Working Group

2. Philosophies of Measurements

It seems that there are two different philosophies to considerthe measurement techniques of the radio signals.

One is the very precise measurement of the physicalquantity. The main purpose is to get as accurate physi-cal quantities as possible. To accomplish the purpose, it isthe case that very time consuming measurement proceduresshall be necessary by using very expensive instruments withhigh accuracy and precision.

The other is the regulatory or the compliance measure-ments. The measurement procedure shall be simple enoughto complete within a feasible period by using the off-the-shelf instruments. The measured value shall be reasonablymeaningful from the view points of compatibility and/or co-existence with other radio systems, but it is not necessarythat the measured value is completely identical to the realphysical quantity.

The latter philosophy is applied through the paper, butthe efforts were made to get more accurate values.

3. Spectral Emission Mask

The UWB emission is spread over a very large frequencyrange and overlaps several frequency bands allocated to ra-dio services. Therefore, it is a reasonable approach that theemission level shall be restricted by the spectral emissionmask, just like the unwanted emission limit of unintentionalemitters for electromagnetic compatibility (EMC).

For example, Fig. 1 shows spectral masks of indoor andhandheld UWB systems approved by the Federal Communi-cations Commission (FCC) in USA [3]. Note that no otheradministration than USA has approved the unlicensed use ofUWB technology yet for commercial applications.

Fig. 1 Emission limits of indoor and hand held UWB systems by FCC.

Although the emission limit itself is not the matter ofthe measurement, the value has the impact on the specifica-tion and the configuration of the instruments.

4. Frequency Domain vs. Time Domain MeasurementApproaches

Chapter 2 of PDNR ITU-R SM.[UWB.MES] [1] describesabout two alternative approaches for the measurements.

One approach involves a measurement of the time do-main characteristics of the signal. The time domain data aretransformed into the frequency domain data via the discreteFourier transform (DFT) or the fast Fourier transform (FFT).The second approach involves a direct measurement of theUWB spectral characteristics in the frequency domain.

There are equipment limitations that must be con-sidered with respect to either of these measurement ap-proaches. The state-of-art single-end oscilloscope can mea-sure a single-event waveform up to 12 GHz, while the quan-tization bits of A/D converter is limited to 8 bits. The sam-pling oscilloscope can measure a periodic waveform up to50 GHz, with the quantization bits of A/D converter 14 bits.It is suitable for the peak power emission measurementswith wider resolution bandwidth (RBW) that can not beachieved by the spectrum analyzer. Contrary, the relativelynarrow dynamic range prevents the measurements of the lowlevel emission outside the UWB bandwidth.

The spectrum analyzers are typically designed to mea-sure conventional narrow-band signals. The IF bandwidthof the spectrum analyzer is significantly smaller than thebandwidth of the UWB waveform, and is even smaller thanthe 50 MHz reference bandwidth for the peak power mea-surement. It is necessary to convert the narrow band mea-surement results into the estimate of the power within widerRBW. Note that this value is not the true power value butonly the estimate of the power. The conventional spectrumanalyzer (Fig. 2) uses a super heterodyne architecture and asweep frequency local oscillator (LO), and the sample tim-ing for each frequency bin is different. Therefore, the mea-surement results shall be influenced by the sweep time andthe measurement time window of the spectrum analyzer, aswell as by the duty ratio and the pulse repetition frequencyof the UWB signal. In addition, conventional spectrum ana-lyzer is not so sensitive as to detect UWB signals at the verylow emission limits applicable in specific frequency bands.


Fig. 2 Block diagram of super heterodyne spectrum analyzer.

5. UWB Parameters

Three power values are to be measured; peak power, averagepower, and quasi-peak power.

Peak power mainly influences the RF front end ofthe receiver, e.g. the saturation of the low noise amplifier.Therefore, it is not necessary to evaluate the total peakpower in the whole bandwidth, but the peak power withinthe sufficiently wide bandwidth that corresponds to the RFfilter of the victim receiver.

Average power is a measure of the interference to thevictim system. To roughly estimate the effect of the inter-ference, the interference signal power can be treated as theincrease of the noise level.

UWB technology is expected to be used in imagingapplications in the frequency below 1000 MHz. They uti-lize an unmodulated pulse train with low pulse repetitionfrequency (PRF). Typically, PRF is below 1 MHz and thespecrum has multiple line peaks. The peak power of a spec-tral line can be an intereference to a tuned narrowband vic-tim receiver. Therefore, emissions below 1000 MHz aremeasured by a quasi-peak signal detector, as this type ofUWB waveform should be peak-power limited. Note thatquasi-peak represents a “weighted” peak.

Equivalent isotropic radiated power (EIRP) is used toevaluate the emission level. It is the product of the powersupplied to the antenna and the antenna gain in a given direc-tion relative to an isotropic antenna. EIRP is a direct mea-sure of the emission, but it is a function of both direction andfrequency, and it is rather difficult to find the highest value.

Instead of the total power in the full frequency band,power spectral density (PSD) is used as a measure of thepower contained within a specified segment of spectrum.PSD is expressed as dBm/MHz.

UWB bandwidth is another parameter to be determinedby the measurement. It is typically defined as −10 dB band-width [2].

6. Measurement Conditions

Chapter 4 of PDNR ITU-R SM.[UWB.MES] [1] describesthe measurement conditions.

Figure 3 shows a block diagram of a general UWBmeasurement system. Measuring receiver can be a spec-trum analyzer and/or an EMI receiver for frequency domainmeasurements, and an oscilloscope for time domain mea-surements.

Fig. 3 General UWB measurement system.

It is recommended to measure the emission in a high-performance anechoic chamber, as other licensed servicesmay use the same frequency and the measurements may becorrupted due to the saturation of the high-gain low noiseamplifier (LNA). A semi-anechoic chamber may be usedat the frequency below 1000 MHz, with the correction of4.7 dB from the maximum value of the receiver antennaheight pattern. In addition, a reverberation chamber maybe employed for the average power measurements.

The UWB device under test (DUT) must be orientedwith respect to the measurement system to ensure recep-tion of the maximum radiated signal. The use of a non-conductive turntable is suggested to systematically searchfor the orientation that provides the maximum responsewithin the measurement system. The spherical wave theory[4] suggests that the maximum number of the ripples of thedirective emission pattern in 360 degrees rotation N is re-lated to the radius of the sphere enclosing the UWB devicer0 and the maximum frequency fH as

N 2π fHr0

c, (1)

where c = 3 × 108 m/s is the velocity of the electromagneticwave in free space. It is a rule of thumb that five to tensamples per ripple is sufficient to find the peak.

A separation distance of 3 m is used for measuring con-ventional wireless devices. However, it may not be possibleto measure low level UWB signals. In the case, the distanceshall be reduced to 1 m or less, taking care to maintain thefar field condition. In fact, there are two far field condi-tions to be considered. First one is the boundary betweenthe reactive region and the radiative region. The radius ofthe boundary rr is

rr =λ

π, (2)

where λ = cf is the wavelength for frequency f . Within

rr, the reactive energy is bigger than the radiated energy,and the measured result does not coincide with the emissionpower. Another one is the boundary between the Fresnel re-gion and the Fraunhoefer region. The radius of the boundaryr f is

r f =8(ra + r0)2

λ, (3)

where ra is the radius of the sphere enclosing the measur-ing antenna. r f is defined to limit the phase error due to thesizes of the antenna and the DUT within π8 rad. Therefore,


the influence of r0 may be negligible as far as the antennaof DUT is not designed with high directivity. Note, how-ever, that r f is bigger than 3 m at 10 GHz if ra is bigger than0.1 m. For the DUT with a high gain antenna, such as ashort range radar, the conducted measurement presented inSect. 7.6 shall be considered alternatively.

The radiated emissions from a UWB device are oftentoo weak to overcome the noise figure of a conventionalspectrum analyzer. The EIRP limit due to the noise is dis-cussed in Appendix A.

Therefore, it becomes necessary to utilize an LNA atthe output of the measurement antenna to reduce the effec-tive noise figure of the overall measurement system.

7. Frequency Domain Measurements

Chapter 6 of PDNR ITU-R SM.[UWB.MES] [1] describesseveral alternative techniques of frequency domain mea-surements.

Three signal detectors are to be used in measuringUWB waveforms. For measuring signal characteristics inthe radio frequency spectrum below 1000 MHz, a CISPR16-1-1 quasi-peak (QP) detector [5] is specified. A root-mean-square (RMS) average detector is specified for mea-suring the average UWB radiated signal amplitude in thefrequency spectrum above 1000 MHz. A peak detector isalso necessary for determining the peak power amplitudeassociated with UWB waveforms in the spectrum above1000 MHz.

7.1 Measurement Uncertainty of Spectrum Analyzers

In general, frequency characteristic of spectrum analyzer isnot flat. For example, when the level measurement of un-modulated signal in the frequency range below 40 GHz isperformed by the spectrum analyzers, the measurement un-certainty is ±3 dB or ±5.5 dB for the high performance spec-trum analyzers or the off-the-shelf general purpose spectrumanalyzers, respectively. In this case, the measurement uncer-tainty includes errors caused by non-linearity of the logarith-mic amplifier and by the gain deviation which depends onthe RBW value of the spectrum analyzer. This uncertaintycan be reduced to 0.49 dB after the calibration by using apower meter and a 4-port coaxial transfer switch. The de-tail of the uncertainty budget and the calibration method ispresented in Appendix B.

7.2 Determination of UWB Bandwidth

A peak PSD with the reference bandwidth of 1 MHz is usedas an emission level to determine the bandwidth in the fol-lowing manner. The measurement shall be made using aspectrum analyzer with a 1 MHz RBW and a 3 MHz videobandwidth (VBW). Choice of VBW is discussed in Ap-pendix C. The analyzer shall deploy a peak detector anda maximum-hold trace mode. The frequency point for thehighest radiated emission evaluated by the peak PSD in a

1 MHz segment is designated as fM . The 1 MHz segments,below and above fM, where the emission level falls 10 dB,are designated fL and fH , respectively. The two recordedfrequencies represent the highest fH and lowest fL bounds ofthe emission. The centre frequency fC and fractional band-width µ−10 can be calculated from fL and fH as

fC =fL + fH

2(4)

µ−10 =fH − fL

fC(5)

7.3 QP Detector for PSD at Frequency below 1000 MHz

A QP detector specified in CISPR 16-1-1 [5] is used to mea-sure PSD below 1000 MHz. However, the drawback of theuse of the CISPR QP detector is its long response time. Infact, the output of QP detector does not exceed the outputof peak detector described in Sect. 7.5. Therefore, it is sug-gested that the peak power shall be measured preliminarily.The QP detector is then used only when the peak power ex-ceeds the emission limit.

7.4 Average PSD

Average PSD is defined as the maximum EIRP within 1 MHzbandwidth averaged over 1 ms. Three alternative measure-ment methods are presented. Table 1 compares these threemethods. First and second methods use a sample detectorand the power sum shall be taken as the post-processing,while third method uses an RMS detector to take the averagein the spectrum analyzer. First and third methods use the IFfilter with RBW of 1 MHz, while the second method sweepswithin 1 MHz bandwidth with narrower RBW to coincidewith the rectangular spectral mask with the bandwidth of1 MHz.

7.4.1 Radiometric Measurements for Low EIRP

In certain frequency ranges, the emission limit of UWB de-vices may be very low. Appendix A describes the noisepower of the spectrum analyzer. It is difficult to measurevery low EIRP by using a conventional spectrum analyzer,even by using the LNA. Radiometric techniques provide aviable method for measuring such a low level of EIRP. Anexample setup of the measurement is shown in Fig. 4. In

Table 1 Comparison of three measurement methods for average PSD.

Sect. 6.4.2.1 Sect. 6.4.2.2 Sect. 6.4.2.3

RBW* 1 MHz 10 kHz 1 MHzVBW** 3× RBW 3× RBW 3× RBWFrequency span 0 1 MHz N/ADetector sample sample RMSSweep time 1 ms coupled function (number of bins)×1 msAverage PSD power sum power sum each binNoteRBW*: Resolution bandwidthVBW**: Video bandwidth, see Appendix C


Fig. 4 Example of radiometric technique for EIRP measurement in 1–2 GHz.

the radiometric techniques, EIRP of the DUT and EIRP ofbackground are measured. After subtracting the latter fromthe former, the true value of EIRP emitted from DUT canbe obtained properly. The detailed scheme is described inRef. [6] and Appendix 5 of PDNR ITU-R SM.[UWB.MES][1].

7.4.2 Total Radiated Power Measurement by Using Rever-bration Chamber

A reverberation chamber is a metal chamber with rotatingmetallic blades which are called stirrers. The stirrers havebeen used to realize a statistically uniform electromagneticfield distribution by changing the boundary condition insidethe chamber. DUT is set at the testing area of a chamber,and the radiated power is evaluated by an average receivingpower that is measured by carrying out stepping or contin-uous rotation of the stirrers. Figure 5 shows a measurementsetup in the reverbration chamber.

Reverberation chambers offer advantages for UWBmeasurements at higher frequencies:

• Measurement sensitivity is grater than that for freespace measurements, since the energy is “contained”in the chamber.• The probability density function (PDF) of the received

signal is substantially independent of the orientation ofthe test device and receive antenna.• Since the radiation power in each frequency is mea-

sured, the frequency at which the highest radiatedemission occurs fM can be known.

The maximum radiation direction cannot be known by a re-verberation chamber measurement, and the total radiationpower (TRP) is measured instead of EIRP. Upper boundof EIRP can be estimated as the product between TRP andmaximum antenna gain. More accurate value can be mea-sured in the subsequent anechoic measurement at the fre-quency fM.

The requirements for a reverberation chamber to beused for emission measurements are described in Sect. 65.7of the CISPR 16-1-4 [7]. In the measurement process, DUTin the chamber is rotated at intervals of 45 degrees, and theexamination is repeated four times. The receiving antennarotation is set at intervals of 90 degrees. TRP of DUT canbe measured using the substitution method. The input powerPT X is delivered from the reference signal generator to theinput of the transmitting antenna, and the receiving power

Fig. 5 Measurement setup in the reverberation chamber.

PRX is measured by the receiver connected to the receivingantenna. The receiver is a spectrum analyzer setup in thesame manner as the anechoic measurements. In the nextstep, DUT is fixed at the same position as the transmittingantenna. Then the received power PM of DUT can be mea-sured. TRP of DUT PDUT is obtained as

PDUT =〈PM〉PT XηT X

〈PRX〉 , (6)

where ηT X is the radiation efficiency of the transmitting an-tenna.

7.5 Peak PSD

Peak PSD is defined as the maximum EIRP within 50 MHzbandwidth with the observation time window of 0.1 ms.However, most of the spectrum analyzers are not equippedwith the IF filter with RBW of 50 MHz. Therefore, the mea-surement shall be done with narrower RBW, and the resultis converted to the 50 MHz reference bandwidth by somemeans.

Two alternative measurement methods are presented,but only the first method is clearly defined. Table 2 com-pares the parameters for two methods. The first method usespeak detector and the power sum over 50 MHz bandwidthshall be taken as the post-processing. When the additionlaw of the signal is known to satisfy either power sum law(7) or voltage sum law (8), the appropriate formula is used.Otherwise, voltage sum law (8) is used as it always giveslarger value than power sum law (7).

P = 10 log10

S pN · IBW

N∑n=1

10P(n)10

, (7)

P = 20 log10

S pN · IBW

N∑n=1

10P(n)20

, (8)

where P dBm is the peak PSD in the span, S p MHz is thefrequency span, N is the number of measurement bins in


Table 2 Comparison of two measurement methods for peak PSD.

Sect. 6.7.2 Sect. 6.7.3

RBW* 1 MHz 1 MHzVBW** 3 MHz ≥ 1 MHz (3 MHz recommended)Frequency span 50 MHz not specifiedDetector peak peakSweep time 0.1 ms per bin not specifiedPeak PSD power sum or voltage sum scaling; rule not specifiedIBW*** corrected no consideration

NoteRBW*: Resolution bandwidthVBW**: Video bandwidth, see Appendix CIBW***: Impulse bandwidth, see Appendix D

the span, P(n) dBm is the power value of the n-th bin mea-sured by the spactrum analyzer, and IBW MHz is the im-pulse bandwidth of the spectrum analyzer. The definitionand the meaning of the impulse bandwidth is described inAppendix D. The second method has not yet been clearlydefined. Instead of measuring the spectrum within 50 MHzreference bandwidth, the maximum peak PSD for 1 MHzRBW is scaled to 50 MHz, but the conversion law is notspecified.

7.6 Conducted Measurement

The radiation measurement takes a long time to calibratethe measurement environment, as well as to measure it un-der the condition where directivity and gain of antenna areaffected by the chassis, cables, etc. Contrary, the conductedmeasurement is effective to measure TRP in a short time toknow whether the highest emission level exceeds the emis-sion limit. In addition, the measurement sensitivity may notbe the major problem in the conducted measurements sincethe power is fed dorectly from DUT to the spectrum ana-lyzer. EIRP can be obtained as the product between TRPand maximum antenna gain.

The conducted measurement is possible only when theantenna terminal is available on the transmitter (DUT). Thespectrum analyzer is directly connected to DUT via thecoaxial cable to measure the spectral characteristics of TRP.The antenna gain shall be separately measured by varyingthe direction to find the maximum gain.

8. Time Domain Measurements

Chapter 7 of PDNR ITU-R SM.[UWB.MES] [1] describesthe time domain measurements and their limitations.

8.1 Desirable Specifications for an Oscilloscope

There are two categories of oscilloscopes. One is a single-event oscilloscope to acquire the waveform by the realtimesampling. The other is a sampling oscilloscope to mea-sure only the periodic waveform by shifting the sample tim-ing. Sampling oscilloscopes exceeds the single-event oscil-loscopes in the performance. However, it is necessary tostop time gating of a UWB signal waveform to generate a

Table 3 Specifications of state-of-the-art single-event and sampling os-cilloscopes.

Single-event Sampling

Maximum frequency 12 GHz 50 GHzQuantization bits 8 bits 14 bits

periodic waveform.The relationship between the quantization bits n and

the dynamic range D dB is expressed as

D = 20 log10 2n. (9)

For example, to measure an UWB signal with D = 60 dB, atleast 10 bits quantization is required for the A/D converter ofthe oscilloscope. It is simultaneously required for the analogfrontend of the oscilloscope to have a noise floor below theminimum descritization level. Simulation studies have beenperformed for the measurement error of the noise levels orspurious level, and the results are shown in Appendix E. Forcomparison, specifications of state-of-the-art single-end andsampling oscilloscopes are listed in Table 3. Comparing theresults of Appendix E and Table 3, it is insufficient to usethe oscilloscopes for the purpose to measure low emissionlevels. Therefore, the time domain measurements by us-ing oscilloscopes are only adequate to measure the highestemission limit.

As is defined in Sect. 7.5, the reference bandwidth ofthe peak power is fb =50 MHz, but most of the spectrumanalyzers are not equipped with the IF filter with RBW of50 MHz. Use of an oscilloscope may be one of the effectivemethods to achieve the wider RBW by the post processingof the measured time domain data. Instead of the IF filter, adigital filter with 50 MHz RBW can be implemented offline.The filter can be implemented either in frequency domainor in time domain. Since the Gaussian filter is usually usedas the IF filter of the spectrum analyzer, the Gaussian dig-ital filter is considered in the time domain processing. Thefrequency transfer function of the Gaussian filter G( f ) with−3 dB bandwidth fb is given as

G( f ) = 2−2

(f− fc

fb

)2

, (10)

where fc is the center frequency of the filter.To find the peak PSD, the measurement time window

shall be Tw = 0.1 ms. The number of samples necessary for


the data processing Ns is simply obtained by

Ns = Tw fs, (11)

and Ns = 5 × 105 when the sampling frequency is fs =

50 GHz. For the sampling oscilloscope, this value is the up-per bound because TW is replaced by the period of the signalTp when Tp < Tw.

8.1.1 Measurement Error due to Jitter

Due to the phase noise of the sampling clock, the sampletiming of the sampling oscilloscope has some random errorcalled jitter. It is assumed that the timing error of the sam-pling τe is according to PDF h(τe). When a periodic wave-form g(t) is measured by the sampling oscilloscope with thesample timing jitter expressed as h(τe), and is averaged bythe intrinsic average function of the sampling oscilloscope,the measured waveform s′(t) is

s′(t) =∫ ∞

−∞s(t − τ)h(τ)dτ. (12)

Equation (12) shows that the effect of the jitter is expressedas the convolution of the periodic waveform and the jitterPDF. In the frequency domain, Eq. (12) is rewritten as

S ′( f ) = S ( f )H( f ), (13)

where S ( f ), S ′( f ), and H( f ) are the Fourier transform ofs(t), s′(t), and h(t), respectively. The characteristic func-tion of the jitter behaves like a bandpass filter. For example,when h(τ) is Gaussian distributed with the standard devia-tion σ, i.e.

h(τ) =1√2πσ

exp

(− τ

2

2σ2

), (14)

its Fourier transform is

H( f ) = exp(−2π2σ2 f 2

), (15)

and −3 dB bandwidth is about 0.22/σ. Considering theanalogy to video filter in Appendix C, the acceptable de-viation of the jitter for 50 MHz peak measurement is 1.5 ns.

8.2 Post-Processing of Time Domain Data

8.2.1 Complex Antenna Factor

When an antenna receives a plane wave of frequency f , asshown in Fig. 6, a complex antenna factor (CAF) Fc( f ) isdefined as [9]

Fig. 6 Definition of CAF.

Fc( f ) =E( f )V0( f )

, (16)

where E( f ) and V0( f ) are the complex electric field strengthat a specific at a specific reference point of an antenna ele-ment and the complex matched voltage of the antenna termi-nal for the load of the matched impedance Z0 = 50Ω. Notethat the conventional antenna factor can not be used to per-form the reconstruction of electric-field waveform, as thegroup delay characteristics are not known. Contrary, CAFhas the phase information and the effect of the group delayis considered. CAF must be measured for each individualantenna.

8.2.2 Reconstruction of Electric Field from MeasuredTime Domain Data

Figure 7 shows an example of an apparatus for measuringthe electric field waveform radiated by DUT. The waveformobserved with an oscilloscope vm(t) is obtained as the con-volution of the impulse response of the measuring apparatusfrom the antenna output to the output of the oscilloscopewith the antenna output signal va(t). EIRP can be convertedfrom the electric field intensity by considering distance as-suming the far field condition.

Figure 8 shows the equivalent circuit of the waveformmeasuring apparatus shown in Fig. 7. In the figure, S de-notes the S-matrix of a pre-amplifier and cables, and S S

is the S-matrix of an oscilloscope. Γa and Γs denote thereflection coefficients of a receiving antenna and of the in-put port of the oscilloscope, respectively. Now, S 12 of thepre-amplifier, i.e. a transmitting S-parameter from the out-put port to the input port of the pre-amplifier, is assumedto be zero. S 22S and S 12S can also be assumed as zero be-cause Vm is not a real signal but a digitized numeric outputby the oscilloscope. As a result of S-parameter analysis un-der the above-mentioned conditions, the electric field in thefrequency domain E( f ) is expressed as

E( f ) =(1 − S 11Γa)(1 − S 22S 11s)

S 21S 21sFc( f )F [vm(t)]

Fig. 7 Electric field waveform measuring apparatus.

Fig. 8 Equivalent circuit of the waveform measuring apparatus.


=Fc( f )

KF [vm(t)] (17)

where F denotes the Fourier transform, and is implementedas DFT or FFT. K is the calibration coefficient of the system,and it is unity in the ideal case.

The output waveform of the Gaussian filter (Eq. (10))e f (t) is expressed as

e f (t) = F −1

[Fc( f )G( f )

KF [vm(t)]

]. (18)

Finally, EIRP p(t) is obtained from the electric field as

p(t) =4πr2e2

f (t)

Zf=

(e f (t)r

)2

30, (19)

where Zf = 120 πΩ is the characteristic impedance of freespace, and r is the distance between DUT and the receivingantenna. The peak PSD is the maximum value of p(t).

Different from the frequency domain measurementSect. 7.5, this time domain measurement gives the true valueof the peak PSD.

More detailed discussion and a measurement exampleare available on Ref. [10].

8.2.3 Use of Spectrum Analyzer as Frequency Converterfor Time Domain Measurement

It is very expensive to execute the baseband sampling of theUWB signal by using the high performance oscilloscope. Infact, the full band measurement is not necessary to obtainthe peak PSD. Therefore, a wideband frequency downcon-verter can help mitigation of the required specification of theoscilloscope. In Fig. 9, a spectrum analyzer is used as a fre-quency downconverter, and the oscilloscope samples the IFoutput of the spectrum analyzer. The first IF in Fig. 2 can beoutput from the spectrum analyzer. The center frequency ofthe first IF is relatively high, ranging from 60 to 500 MHz.Thus, the bandwidth of the signal is considerably wider thanthe reference bandwidth of 50 MHz for peak PSD. It is apractical idea to use this IF signal to measure in the band-width wider than RBW of spectrum analyzers. This first IFsignal is measured by an oscilloscope. Since it samples theIF signal and not RF signal, sampling rate of oscilloscopecan be much lower. The post processing is very similar toSect. 8.2.2 to get the peak PSD. By changing the RF centerfrequency of the spectrum analyzer, the whole bandwidthcan be covered by the appratus.

The following requirements shall be fulfilled by the ap-paratus shown in Fig. 9:

Fig. 9 Time domain measurement apparatus using a spectrum analyzeras a frequency down converter.

• The passband amplitude response of the spectrum ana-lyzer should be flat.• The passband phase response should be linear.• The conversion gain from the RF input level to the IF

output level shall be obtained by the calibration foreach of the center frequencies.• Input bandwidth of the oscilloscope shall be suffi-

ciently high to sample the IF signal.

9. Conclusion

This paper has described the measurement techniques of theUWB emission, based on the discussion in ITU-R TG 1/8.In addition, the paper has provided the background idea be-hind the measurement methods, which will not be presentedon the ITU-R Recommendation. Since the final draft has notbeen approved, some change may occur from this paper. Ifthere are major changes in the PDNR, the authors intend tosubmit a supplement of this paper.

Acknowledgements

The authors have worked together in the microwave ma-surement working group, National Institute of Informationand Communications Technology (NICT) UWB Consor-tium. They have also contributed to the ITU-R TG 1/8 WG4.They submitted 21 contribution documents and dispatchedseveral members to every meeting. Mr. T. Yasui has beenthe chairman of WG4.

The authors thank all the members in NICT UWB In-stitute and UWB Consortium for their helpful discussions,suggestions, and encouragement. Data presented in Ap-pendix E have been provided by Dr. Hitoshi Sekiya of An-ritsu Corp.

References

[1] Japan, “Proposed modification to the PDNR SM.[UWB.MES],”ITU-R, SG 1, TG 1/8, Contribution Document 1-8/284-E, May2005.

[2] “Working document toward a preliminary draft new recommenda-tion ITU-R SM.[UWB.CHAR]—Characteristics of ultra-widebandtechnology,” ITU-R, SG 1, TG 1/8, Contribution Document 1-8/256-E, Annex 1, Dec. 2005.

[3] “Radio frequency devices,” Part 15, Federal Communications Com-mission Rules, Dec. 2003.

[4] J.E. Hansen, Spherical Near-Field Antenna Measurements, PeterPeregrinus, London, 1988.

[5] “Specification for radio disturbance and immunity measuring appa-ratus and methods—Part 1-1: Radio disturbance and immunity mea-suring apparatus— Measuring apparatus,” CISPR 16-1-1, IEC, Nov.2003.

[6] M. Uchino, “Radiometric measurement of equivalent isotropic ra-diated power of UWB devices,” IEICE Trans. Commun. (JapaneseEdition), vol.J87-B, no.6, pp.921–923, June 2004.

[7] “Specification for radio disturbance and immunity measuring ap-paratus and methods—Part 1-4: Radio disturbance and immu-nity measuring apparatus—Ancillary equipment—Radiated distur-bances,” CISPR 16-1-4, IEC, May 2004.


[8] H. Sugama and Y. Yamanaka, “EMI measuring equipment above1 GHz—Characteristic evaluation of spectrum analyzers,” IEICETechnical Report, EMCJ99-86, 1999.

[9] S. Ishigami, H. Iida, and T. Iwasaki, “Measurements of complex an-tenna factor by the near-field 3-antenna method,” IEEE Trans. Elec-tromagn. Compat., vol.38, no.3, pp.424–432, July 1992.

[10] S. Ishigami and Y. Yamanaka, “Reconstruction of electric-fieldwaveform radiated from UWB device by using the complex antennafactor,” Proc. International Symposium on Electromagnetic Compat-ibility (EMC SENDAI ’04), pp.41–44, June 2004.

Appendix A: Noise Level of Spectrum Analyzer

Appendix 3 of PDNR ITU-R SM.[UWB.MES] [1] describesconversion from the noise level of the spectrum analyzer toEIRP. This section describes an alternative way to derive thenoise EIRP.

The noise power of the receiver N dBW is expressed as

N = 10 log10 kT B + F, (A· 1)

where k = 1.38 × 10−23 J/K is Boltzmann’s constant, T Kis the temperature of the receiver, B Hz is the receiver noisebandwidth, and F dB is the noise figure of the receiver. Fora receiver with 1 MHz bandwidth, i.e. specified RBW, thenoise power is expressed as

N [dBm/MHz] = −114 + F, (A· 2)

where T = 290 K is assumed. Note that typical noise figurevalues of spectrum analyzers are in the range of 18–24 dB.Received power Pr dBm is expressed by Friis’ transmissionformula as

Pr = Pte + 20 log10 λ − 20 log10(4πd) +Gr, (A· 3)

where Pte dBm is EIRP of the DUT, λm is the wavelength,d m is the distance between DUT and the receiver antenna,and Gr dB is the receiver antenna gain. Typical gain valuesof the receiver antennas are in the range of 2 to 4 dBi in-cluding the cable loss. Therefore, the typical noise power isequivalent to the EIRP of −58 to −50 dBm/MHz. If 10 dBSNR is necessary for the measurement, the minimum EIRPto be measured is in the range of −48 to −40 dBm/MHz.Therefore, the use of LNA is necessary to measure the lowlevel EIRP.

Appendix B: Uncertainty Budget of Measurement forSpectrum Analyzers

Table A· 1 shows the estimate of the measurement uncer-tainty in the general purpose and the high-performance spec-trum analyzers that can measure up to 22 GHz. The total un-certainty is obtained as the square root of the squared sumsof the error factors.

Figure A· 1 shows a level calibration system for spec-trum analyzers using a 4-port coaxial transfer switch, a sig-nal generator, an RF power meter, and a spectrum ana-lyzer. In the calibration mode Fig. A· 1(a), the coaxial trans-fer switch links the reference signal generator to the spec-trum analyzer. A CW reference signal with 0 dBm power is

Table A· 1 Uncertainty budget of spectral measurement below 22 GHzwithout calibration.

General purpose High performance

Calibration source1 0 dB 0 dBFrequency response2 ±5 dB ±2.5 dBReference level ±0.4 dB ±0.2 dBLog linearity ±1.0 dB ±0.5 dBRBW switching ±0.3 dB ±0.2 dBTotal uncertainty ±5.5 dB ±3.0 dB1: The uncertainty of the calibration source is included in

the uncertainty of reference level.2: The frequency response is assumed to have a uniform

distribution.

(a) Calibration mode

(b) Measurement mode

Fig. A· 1 Calibration and measurement setup using 4-port coaxial trans-fer switch.

fed from the signal generator to the spectrum analyzer, andthe indicated value PS A is recorded. Next, the coaxial trans-fer switch is turned into the measurement mode to link thesignal generator to the RF power meter and to link the inputport to the spectrum analyzer. Then the spectrum analyzermeasures UWB signal under test, and the indicated value isPM . Simultaneously the RF power meter measures the RFpower of the unmodulated carrier fed from the signal gener-ator, and the indicated value PPM is recorded. The calibratedpower value PDUT is then given as

PDUT =PMPPM

PS A. (A· 4)

Uncertainty budgets after the above calibration aresummarized in Table A· 2. Expanded uncertainty after thecalibration that is defined as 95% confidence interval is0.49 dB below 22 GHz.

Appendix C: Choice of Video Filter

As shown in Fig. 2, a super heterodyne spectrum analyzerhas two filters, i.e. an IF filter and a video filter. Two


Table A· 2 Uncertainty budget of spectral measurement below 22 GHz after calibration.

Sources of uncertainty Xi Estimate Uncertainty U(Xi) PDF Standard uncertainty ui(y)

Lsm 1.000 mW 0.02% Gaussian 0.05%Pm 1.000 mW 0.02% Gaussian 0.05%Pmc 1.000 mW 0.29% rectangular 0.29%Pcal 1.000 mW 0.69% rectangular 0.69%Kb 1.00 1.45% rectangular 1.45%Mu 1.00 4.07% U-shaped 4.07%Muc 1.00 0.45% U-shaped 0.45%Msp 1.00 4.07% U-shaped 4.07%Lsw 1.00 0.40% rectangular 0.40%

Combined standard uncertainty uc(y) 6.01%Expanded uncertainty U = k · uc(y), k = 2 12.02%

Expanded uncertainty in dB 0.49 dBSources of uncertainty

Lsm : Indicated value on the spectrum analyzerPm : Indicated value on the RF power meterPmc : Indicated value on the RF power meter connected to the reference signal generatorPcal : Output power of the reference signal generatorKb : Sensor calibration factor of the RF power meterMu : Mismatch loss between the input port and the RF power meterMuc : Mismatch loss between the reference signal generator and the RF power meterMsp : Mismatch loss between the input port and the spectrum analyzerLsw : Insertion loss of the 4-port coaxial transfer switch

Fig. A· 2 Frequency response of IF Filters (RBW = 1 MHz).

measurement parameters are related to these filters, i.e., thebandwidth of the IF filter is RBW, and that of the video fil-ter is VBW. Figure A· 2 shows the relation between VBWand the peak power recorded by the spectrum analyzer whenRBW is fixed to 100 kHz. The figure suggests that the peakpower measurements should be made with VBW at least3 times of RBW. Otherwise, the video filter attenuates thespectrum component at the band edge, which results in thedecrease of the measured value. Care must be taken in themeasurement, as the default choice of the coupled VBW ina conventional spectrum analyzer usually the same value asRBW.

Appendix D: Impulse Bandwidth

Spectrum analyzers are usually calibrated for sinusoidal in-puts to yield accurate results. However, they are not guaran-teed to yield reproducible results for other type of inputssuch as an impulse waveform generated by an UWB de-vice. It is necessary to know the response to an impulsewaveform. The impulse bandwidth (IBW) represents thisresponse [5].

Fig. A· 3 Experimental setup for measurements of impulse bandwidth.

IBW is not usually indicated as a specification of aspectrum analyzer. Therefore, this value is necessary to bemeasured. Figure A· 3 shows the measured frequency re-sponse curves of the IF filters in the spectrum analyzers ad-justed for a nominal RBW of 1 MHz. The horizontal axisdenotes the frequency normalized by the RBW, and the ver-tical axis does the attenuation of the IF filter. It is shownthat −6 dB or −20 dB bandwidth is different in each modeleven though nominal −3 dB bandwidth is 1 MHz in all theanalyzers.

IBW is defined as

Bimp =A(t)max

2G0 · IS, (A· 5)

where A(t)max is the peak value of the envelope at the IFoutput of the analyzer with an impulse area

IS =∫ ∞

−∞V(t)dt (A· 6)

applied at the analyzer input, and G0 is the gain of the circuit


Fig. A· 4 Peak power vs. VBW for 100 kHz RBW.

at the center frequency.The experimental setup for measurements of IBW is

shown in Fig. A· 4 [8]. VBW should be set at least threetimes larger than RBW. The waveform of the rectangularpulse with the repetition frequency fp, the pulse width t, andthe pulse voltage Vp is measured by an oscilloscope, and theimpulse area IS is calculated from Eq. (A· 6).

There are two alternative methods to measure IBW.

1. The peak value of the envelope A(t)max is obtained fromthe RMS average value A(t)rms as

A(t)max =√

2A(t)rms. (A· 7)

A(t)rms is measured at a frequency very close to thepeak of the spectrum of the pulse.

2. When PRF is lower than the impulse bandwidth Bimp, apeak-detection spectrum-analyzer indicates a constantpeak value. However, when PRF is higher than the im-pulse bandwidth Bimp, a peak value becomes equal toaverage value, and is proportional to PRF. The PRF ofthe intersection of these two responses is the impulsebandwidth.

Both of the measurement methods almost show the samemeasured values. Note that Bimp/B3 is about 1.5 in a case ofa Gaussian type BPF.

Appendix E: Limitation of the Time Domain Measure-ments by Using Oscilloscope

Simulation studies have been performed for the measure-ment error of the noise level or spurious level.

Either band-limited noise within 960 to 1610 MHz or1200 MHz spurious signal is added to the impulse UWBsignal which has a spectrum of χ2 function. In this study,the peak to average power ratio (PAPR) of a UWB signalis assumed to be 20 dB. Figure A· 5 show the noise andthe spurious levels vs. the number of quantization bits ofan oscilloscope. In Fig. A· 5(a), noise PSD within 960 to1610 MHz is varied from −50 dBm/MHz to −90 dBm/MHz.In Fig. A· 5(b), power of 1200 MHz line spectrum is variedfrom −50 dBm to −90 dBm.

For example, the emission limit between 960 and1610 MHz is −75.3 dBm/MHz in the FCC mask (Fig. 1).Then, 12 bits quantization is necessary to obtain the results

(a) Measurement error of noise power.

(b) Measurement error of spurious power.

Fig. A· 5 Measurement error of noise and spurious measurements in timedomain.

within 1 dB error for the noise measurements, while 17 bitsquantization is necessary for the spurious measurements.

Jun-ichi Takada received B.E., M.E., andD.E. degrees from Tokyo Institute of Technol-ogy (Tokyo Tech), Japan, in 1987, 1989, and1992, respectively. From 1992 to 1994, hehas been a Research Associate at Chiba Uni-versity, Chiba, Japan. Since 1994, he has beenan Associate Professor at Tokyo Tech. Hiscurrent research interests are wireless propaga-tion and channel modeling, ultra-wideband ra-dio, and applied radio instrumentation and mea-surements. Since 2003, he has been a part time

Specialty Researcher in the UWB Institute, National Institute of Communi-cations and Information Technology (NICT), Japan, where he contributesto the propagation measurement and modeling, as well as the standard-ization of the UWB emission measurement in ITU-R. He is a member ofIEEE, ACES, and ECTI Association Thailand.


Shinobu Ishigami received the B.E., M.E.,and D.E. degrees from University of Electro-Communications (UEC), Tokyo, Japan, in 1990,1992, and 1997, respectively. From 1992 to1999, he was a research associate in UEC.He joined the National Institute of Informa-tion, Communications Technology (NICT, for-mer Communications Research Laboratory), in1999. He is now a senior researcher of Com-munication system EMC group in NICT. He is amember of the IEEE, the IEEJ.

Juichi Nakada received the B.S. degreesin Physics from Toyama University in 1987. Hehas been with ADVANTEST CORPORATION,Tokyo, Japan. His interests include wireless de-vice evaluation and test and measurement instru-ments for next generation wireless communica-tion systems.

Eishin Nakagawa received his B.E. degreefrom Chubu Institute of Technology in 1977. Hejoined NEC Radio & Electronics Ltd. in 1977and engaged in development of radio communi-cation equipment. He moved to Telecom Engi-neering Center in 2001 and has been engaged indevelopment of the measuring methods requiredfor the technical regulations conformity certifi-cation of radio equipment.

Masaharu Uchino received his B.E. andM.S. degrees in electrical engineering from To-kyo Denki University in 1978 and 1980 respec-tively. He has been working for Anritsu Cor-poration since 1980. From 1996 to 1998, heworked at the Electromagnetic CompatibilityResearch Laboratories Co., Ltd., Sendai. He re-ceived D. Eng. degree in electromagnetic com-patibility from Tohoku Gakuin University, Sen-dai, in 2000. His current interests are in the areaof precise frequency control and ultra-wideband

measurement.

Tetsuya Yasui received the M.E. degreein electric engineering from Waseda University,Tokyo, Japan, in 1986. He joined the Ministry ofPosts and Telecommunications (MPT, presentlyMIC), Japan in 1986. From 1986 to 2002 he en-gaged in the approval of mobile communicationbusiness, the frequency management, ITU work,the planning of space satellite R & D, the plan-ning of national ICT R & D and the licensingof wireless access systems. Since 2002 he hasbeen with National Institute of Information and

Communications Technology, as Research Center Supervisor engaged ininternational collaboration and standardization for new generation mobilecommunication project and UWB technology. Since 2003 he has been ap-pointed as the chairman of WG4 (measurement method) of TG1/8 (UWB)in ITU-R.

measurement techniques of emissions from ultra · pdf filemeasurement techniques of emissions...

Documents