principles and limitations of ultra-wideband fm communications systems

EURASIP Journal on Applied Signal Processing 2005:3, 382–396c© 2005 Hindawi Publishing Corporation

Principles and Limitations of Ultra-WidebandFM Communications Systems

John F. M. GerritsCentre Suisse d’Electronique et de Microtechnique SA, Jaquet-Droz 1, CH-2007 Neuchatel, SwitzerlandEmail: [email protected]

Michiel H. L. KouwenhovenNational Semiconductor BV, Amplifier Design Europe, Delftech Park 19, 2628 XJ Delft, The NetherlandsEmail: [email protected]

Paul R. van der MeerElectronics Research Laboratory/DIMES, Faculty of Electrical Engineering, Mathematics and Computer Science (EEMCS),Delft University of Technology, Mekelweg 4, 2628 CD Delft, The NetherlandsEmail: [email protected]

John R. FarserotuCentre Suisse d’Electronique et de Microtechnique SA, Jaquet-Droz 1, CH-2007 Neuchatel, SwitzerlandEmail: [email protected]

John R. LongElectronics Research Laboratory/DIMES, Faculty of Electrical Engineering, Mathematics and Computer Science (EEMCS),Delft University of Technology, Mekelweg 4, 2628 CD Delft, The NetherlandsEmail: [email protected]

Received 10 October 2003; Revised 7 March 2004

This paper presents a novel UWB communications system using double FM: a low-modulation index digital FSK followed bya high-modulation index analog FM to create a constant-envelope UWB signal. FDMA techniques at the subcarrier level areexploited to accommodate multiple users. The system is intended for low (1–10 kbps) and medium (100–1000 kbps) bit rate, andshort-range WPAN systems. A wideband delay-line FM demodulator that is not preceded by any limiting amplifier constitutesthe key component of the UWBFM receiver. This unusual approach permits multiple users to share the same RF bandwidth.Multipath, however, may limit the useful subcarrier bandwidth to one octave. This paper addresses the performance with AWGNand multipath, the resistance to narrowband interference, as well as the simultaneous detection of multiple FM signals at the samecarrier frequency. SPICE and Matlab simulation results illustrate the principles and limitations of this new technology. A hardwaredemonstrator has been realized and has allowed the confirmation of theory with practical results.

Keywords and phrases: UWB, FM, FDMA, WPAN, subcarrier, multipath.

1. INTRODUCTION

Ultra-wideband (UWB) communications systems are poisedto play an increasingly important role in today’s short-rangecommunications systems, especially personal area network(PAN) applications. By definition, the −10 dB RF bandwidth

This is an open-access article distributed under the Creative CommonsAttribution License, which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited.

BRF of a UWB signal centered at a frequency fc should be atleast 20% of this central frequency or at least 500 MHz foroperation above 3.1 GHz [1]. Since the definition of a UWBsignal does not specify a particular air interface or modula-tion scheme, many different techniques may be applicable toa UWB signal.

Originally, UWB started as an impulse radio, using atime-domain approach [2]. Instead of a continuous sinu-soidal carrier, a sequence of short-duration pulses is usedas the information carrier. The spectrum of such a pulse

mailto:[email protected]





Ultra-Wideband FM Communications Systems 383

sequence (usually Gaussian) has a single broad main lobewith slow spectral roll-off. These pulsed systems were origi-nally intended for radar applications where short pulse du-ration translates into a high resolution. When used in acommunications system, the pulse sequence can be mod-ulated using, for example, pulse-amplitude modulation(PAM) or pulse-position modulation (PPM) techniques.

UWB communications technology was originally in-tended to provide robust, easy-to-implement, low-cost, andlow-power consumption solutions. An impulse radio or mul-ticarrier OFDM system, as proposed recently [3], may be ableto provide a robust high data rate solution, but this comes atthe expense of circuit complexity and power consumption.

The goal of the authors was to search for a complemen-tary low and medium data rate (LDR and MDR) UWB sys-tem that is easy to implement in silicon, provides robustnessto interference compared to narrowband ISM solutions, andis competitive in terms of power consumption [4].

The proposed solution is a constant-envelope frequency-domain approach called UWB frequency modulation(UWBFM) [5]. This double FM scheme uses low-modulation index FSK followed by high-modulationindex analog FM to achieve the wide bandwidth. Differentusers distinguish themselves by different subcarrier frequen-cies. This approach has a number of attractive properties foruse in short-range wireless personal area network (WPAN)systems, where the dynamic range of the RF signals is limited.

The paper starts by presenting the principles of UWBFMtechnology in Section 2. Block diagrams of both the trans-mitter and the receiver are presented and an example of amultiuser system is given. Section 3 discusses the operationof the wideband FM delay-line demodulator in the pres-ence of a single FM signal. In Section 4, the overall receiverperformance under additive white Gaussian noise (AWGN)conditions is examined. The nonlinear relationship betweenthe input and output signal-to-noise ratio (SNR) of thewideband FM demodulator strongly influences the receiverperformance. Section 5 presents a multiuser UWBFM sys-tem based upon frequency division multiple access (FDMA)subcarrier technology, and shows mathematically why thisworks. It also addresses the closely related issue of the ro-bustness of UWBFM to narrowband interference. Section 6addresses the effect of multipath on an UWBFM system.Section 7 presents conclusions and topics for further inves-tigations.

2. PRINCIPLES OF UWBFM

UWBFM can be seen as an analog implementation of aspread-spectrum system with a spreading gain equal to themodulation index β. FM has the unique property that the RFbandwidth BRF is not only related to the bandwidth fm of themodulating signal, but also to the modulation index β thatcan be chosen freely. This yields either a bandwidth-efficientnarrowband FM signal (β < 1) or a (ultra-) wideband signal(β 1) that can occupy any required bandwidth compatiblewith the RF oscillator’s tuning range.

0−20−4050

40

30

20

10

0

β

0 10 20 30 40 50 60

n

Figure 1: Amplitude of the Bessel functions (20 log10(Jn(β)) forvarious values of the modulation index β.

Assume an FM signal V(t) with amplitude A and carrierfrequency fc (ωc = 2π fc) modulated by a sinusoidal signalm(t) of frequency fm(ωm = 2π fm) such that

m(t) = Vm sin(ωmt

). (1)

An RF oscillator sensitivity of KO [rad/Vs] yields a deviation∆ω = 2π∆ f equal to

∆ω = KOVm, (2)

resulting in an FM signal V(t):

V(t) = A sin(ωct + ϕ(t)

)

= A sin(ωct + KO

∫ t

−∞Vm sin

(ωmτ

)dτ)

= A sin(ωct − KOVm

ωmcos

(ωmt

)+ ϕ0

)

= A sin(ωct − β cos

(ωmt

)+ ϕ0

),

(3)

where ϕ(t) is the instantaneous phase excursion due to theFM, ϕ0 is an arbitrary but time-independent constant, and βis the modulation index defined by

β = ∆ f

fm= ∆ω

ωm. (4)

Equation (3) can be expressed as a sum of Bessel functionsJn(β) of the first kind (order n) of the argument β:

V(t) = A∞∑

n=−∞Jn(β) sin

(ωc + nωm

)t. (5)

Theoretically, the spectrum of an FM signal is infinitelylarge. In practice, the higher-order Bessel functions Jn(β)decay rapidly for n > β. Figure 1 shows the value of theBessel functions Jn(β) for various values of modulation in-dex β. The bandwidth of an FM signal is well approximated

384 EURASIP Journal on Applied Signal Processing

d(t)

100 kbps

Subcarrieroscillator

m(t)

1 MHzRF

oscillator

V(t)

4 GHz

Figure 2: Block diagram of the UWBFM transmitter.

by Carson’s rule:

BRF ≈ 2(β + 1) fm = 2(∆ f + fm

). (6)

As a result of the fast decay of the Bessel functions for n > β,the bandwidth of a wideband FM signal can be controlledby adapting the modulation index β. When the modulationindex β 1, a wideband spectrum is obtained in whichno carrier can be distinguished. The spectral roll-off of thisUWBFM signal is very steep. This strongly improves the co-existence of UWBFM systems with other RF systems op-erating in adjacent frequency bands. Analog FM can thusbe used as a spreading mechanism to generate an unmod-ulated constant-envelope UWB signal of appropriate band-width. An additional modulation mechanism is still requiredto modulate data upon this UWB signal.

Figure 2 shows the block diagram of the UWBFM trans-mitter. The additional modulation mechanism is digital FMby a raw data signal d(t) of the low-frequency subcarrierusing FSK techniques with modulation index βSUB(0.5 <βSUB < 4).

Individual users are assigned separate subcarrier fre-quencies. The approximate bandwidth BSUB of a subcarriermodulated by a lowpass filtered digital signal of bit rateR (bps) equals

BSUB = R(βSUB + 1

). (7)

Figure 3 shows the data, the subcarrier, and the UWB sig-nal in the time domain for a data transition at t = 0 and sub-carrier frequency of 1 MHz; the center frequency of the UWBsignal V(t) was chosen to be 10 MHz for the sake of visibility.

The choice of the subcarrier frequencies fSUB i and mod-ulation indices βSUB i is determined by the data rate(s) andthe number of users in the UWBFM communications sys-tem. A low-subcarrier modulation index yields a lower sub-carrier bandwidth allowing more users. On the other hand,it requires steeper subcarrier filtering in the receiver. This isespecially true since the wideband FM demodulator has aquadratic transfer function resulting in an expanded subcar-rier dynamic range. In the context of a short-range WPAN,this limited dynamic range (typically 30 dB) can be accepted.We will show in Section 6 that when no equalization is used,multipath may limit the useful subcarrier frequency range toone octave.

An example of a hybrid system providing both LDR andMDR subcarriers are shown in Figure 4. The 3 LDR users(1–3) operate at 10 kbps with a modulation index βSUB = 1yielding a bandwidth of 20 kHz. They are spaced 150 kHzapart. The single MDR user (4) operates at 100 kbps anduses a modulation index βSUB = 2, yielding a bandwidth of300 kHz.

−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

t (µs)

Am

plit

ude

d(t)

m(t)

V(t)

Figure 3: Time-domain view of data d(t), subcarrier m(t), andUWB signal V(t).

10

5

0

−5

−10

−15

−20

−25

−30

−35

−401 1.2 1.4 1.6 1.8 2

Frequency (MHz)

Pow

ersp

ectr

um

(dB

)

1 2 3 4

Figure 4: Spectrum S( f ) after the wideband FM demodulator inthe receiver for a hybrid LDR-MDR 1–2 MHz subcarrier system.

The UWB signal V(t) is obtained by feeding the modu-lated subcarrier signal m(t) into the FM input of the RF os-cillator operating at the desired center frequency ( fc) of theUWB signal.

Figure 5 shows an example of the spectral density of suchan UWBFM signal. The signal power is−13 dBm (140 mVp-p

in a 50Ω load). The subcarrier frequency is 1 MHz and thedeviation ∆ f is 600 MHz, yielding a modulation index β of600. The−10 dB bandwidth is almost equal to the bandwidthpredicted by Carson’s rule in (6):

B−10 dB ≈ 2∆ f . (8)

The spectral density is lowered by a factor of 10 log10(β) =28 dB. This UWB signal is FCC compliant. The flat spectrumis a result of the triangular subcarrier waveform. The powerspectral density of a wideband FM signal is determined by


0

−10

−20

−30

−40

−50

−60

−70

−80

−90

−1003 3.5 4 4.5 5

f (GHz)

FCC spectral mask

Spec

tral

den

sity

(dB

m/M

Hz)

Unmodulated RF carrier

Figure 5: Spectral density S (dBm/MHz) of the unmodulated car-rier at 4 GHz and the UWBFM signal obtained with fSUB = 1 MHzand β = 600.

4 GHzLNA

WidebandFM

demodulator

Subcarrier 1filter

& demodulator

d1(t)

1–2 MHzSubcarrier 2

filter& demodulator

d2(t)

Subcarrier nfilter

& demodulator

dn(t)

......

...

Figure 6: Block diagram of the UWBFM receiver.

and has the shape of the probability density function (pdf)of the modulating signal m(t) [6].

Triangular subcarrier waveforms have a uniform pdf andtherefore yield a flat RF spectrum. From a realization pointof view, the triangular waveform is relatively straightforwardto generate using integrated circuits.

The receiver demodulates the UWBFM signal withoutfrequency translation. No local oscillator and no carrier syn-chronization are required. Figure 6 shows the block diagramof the UWBFM receiver. The receiver in its basic form com-prises a wideband FM demodulator, one or several low-frequency subcarrier filtering and amplification stages, andsubcarrier demodulators. One possible implementation usesa bandpass filter to filter out the wanted subcarrier signal fol-lowed by a phase-locked loop (PLL) to perform the FSK de-modulation.

3. WIDEBAND FM DEMODULATOR,SNR CONVERSION

The wideband FM demodulator is implemented as a delay-line demodulator as shown in Figure 7, where τ = N/(4 fc)with N = 1, 3, 5, . . .. The operation of the delay-line demod-

VRF

fRFAmplifier

τ

Delay element

Multiplier

Vdemod

fsub

Figure 7: Delay-line FM demodulator.

ulator is that of FM-to-PM conversion in the delay line fol-lowed by a phase detector [7].

We will now analyze this demodulator for a single-inputsignal V1 of amplitude A1 as given below:

V1(t) = A1 sin(ωc1t + ϕ1(t)

)= A1 sin

(ωc1t − β cos

(ωmt

)).

(9)

The multiplier output signal VDEMOD equals

VDEMOD(t) = A21 sin

(ωc1t + ϕ1(t)

)× sin

(ωc1(t − τ) + ϕ1(t − τ)

).

(10)

If we ignore the high-frequency term at 2ωc1—a compo-nent that can be easily filtered out in a practical circuitrealization—the lowpass filtered output signal VDEMODLP canbe written as

VDEMODLP(t) = A21

2cos

(ωc1τ + ϕ1(t)− ϕ1(t − τ)

). (11)

By choosing the delay time τ equal to an odd multiple of aquarter period (T) for the carrier frequency fc of the FM sig-nal

τ = NT

4= N

π

2ωc, N = 1, 3, 5, . . . , (12)

equation (11) can be rewritten as

VDEMODLP(t)=(−1)((N+1)/2) A21

2sin

(ϕ1(t)−ϕ1(t − τ)

).

(13)

When the delay τ is much smaller than the period Tm of themodulating waveform of frequency fm (i.e., requiring that fm fc1), this can be written as

VDEMODLP(t) = (−1)((N+1)/2) A21

2sin

(τ∂ϕ1(t)∂t

). (14)

Substituting (3) and (12) in (14) yields

VDEMODLP(t)=(−1)((N+1)/2) A21

2sin

(Nπ

2∆ω

ωcsin

(ωmt

)). (15)

This corresponds to a constant times the sine of the originalmodulating signal m(t).


1.5

1

0.5

0

−0.5

−1

0 0.5 1 1.5 2

f / fc

Am

plit

ude

UWBFM

N = 1

N = 3

N = 5

Figure 8: Normalized relation between delay-line demodulator in-put frequency VFMDEMOD( f ) and output voltage for various valuesof N .

The demodulator output voltage as a function of the in-put frequency VFMDEMOD( f ) is given by

VFMDEMOD( f ) = A21

2cos

(Nπ

2f

fc

). (16)

Figure 8 illustrates this relation for various values of the pa-rameter N .

The demodulator sensitivity is proportional to N . Theuseful RF bandwidth BDEMOD of the FM demodulator is in-versely proportional to N and given by

BDEMOD = 2N

fc. (17)

The useful bandwidth is defined as the maximum frequencyrange over which the static demodulator transfer function ismonotonic.

We define the FM demodulator overdrive O as

O = 2∆ f

BDEMOD= N

∆ f

fc. (18)

Equation (15) can now be rewritten as

VDEMODLP(t) = (−1)((N+1)/2) A21

2sin

(Oπ

2sin

(ωmt

)). (19)

An overdrive O = 1 corresponds to a deviation of the FMinput signal equal to one half of the FM demodulator band-width.

It is important to note that the demodulator output sig-nal amplitude is proportional to the square of the input sig-nal amplitude. As a result, the dynamic range of the demodu-lated signal is expanded. A 20 dB variation at RF yields 40 dBafter the demodulator. This strongly impacts the subcarrierfiltering. A direct-conversion architecture for the subcarrierfiltering and demodulation relaxes the baseband filter speci-fications.

Table 1: Harmonic distortion as a function of the overdrive O.

Overdrive O HD3 (dBc) HD5 (dBc)

0.05 −72 −154

0.10 −60 −130

0.20 −48 −106

0.50 −31 −74

1.00 −18 −48

Table 2: Harmonic distortion as a function of the relative frequencyoffset o for an overdrive O = 0.50.

Offset o HD2 (dBc) HD3 (dBc)

0.05 −36 −31

0.10 −30 −31

0.20 −24 −31

0.50 −7 −21

The sinusoidal transfer function of the demodulatoryields odd harmonic distortion in the output signal. This dis-tortion is a function of the overdrive O. A frequency offset inthe transmitter carrier frequency fc, with respect to the de-modulator center frequency, results in even-order harmonicdistortion. We define the relative offset o as

o = 2∆ fCBDEMOD

. (20)

Numerical values can be obtained by examining the behaviorof the function y:

y = sin(o + O

π

2sin(x)

). (21)

Table 1 presents values for the third and fifth harmonic dis-tortions for various values of the overdrive O and zero-frequency offset. Table 2 illustrates the effect of the frequencyoffset on the harmonic distortion for a fixed overdrive O =0.50.

It appears that in a practical UWBFM system, multipathis a major source of distortion due to the envelope variationsit introduces. As can be seen from (15), the delay-line de-modulator is sensitive to both AM and FM.

Consider the AM input signal VAM(t) of frequency fc andtime dependent amplitude A(t) given by

VAM(t) = A(t) sin(ωct). (22)

The multiplier output signal VDEMOD can be written as

VDEMOD(t) = A(t)A(t − τ) sin(ωct)

sin(ωc(t − τ)

). (23)

If we ignore the high-frequency term at 2ωc, the lowpassfiltered output signal VDEMODLP can be written as

VDEMODLP(t) = 12A(t)A(t − τ) cos

(ωcτ

)≈ 1

2A2(t) cos

(ωcτ

).

(24)


SNRRF1

Ψ

SNRDEMOD2

BSUB

SNRSUB3 Pb

Figure 9: UWBFM receiver block diagram for determining the out-put probability of error Pb.

The approximation is valid for amplitude variations whosebandwidth is much smaller than 1/τ. An amplitude-modulated signal with sinusoidal modulation at modulationfrequency fm has a time-varying envelope A(t) given by

A(t) = A1(1 + m cos

(ωmt

)). (25)

The lowpass filtered demodulated signal VDEMODLP is equalto

VDEMODLP(t) = A21

[1 +

m2

2+ 2m cos

(ωmt

)

+m2

2cos

(2ωmt

)]cos

(ωcτ

) (26)

which for low-modulation depth m can be approximated by

VDEMODLP(t) ≈ A21

[1 + 2m cos

(ωmt

)]cos

(ωct). (27)

This implies that the AM sensitivity equals zero at the operat-ing points chosen for FM demodulation as given by (12) andhas its maximum values in between, where the term cos(ωcτ)has its extreme values. As a result, the delay-line demodulatorprovides strong AM rejection for narrowband signals cen-tered on those operating points. This fact can be exploitedto lower the demodulator output voltage for out-of-band in-terfering signals with a strong AM component (like OFDMWLAN signals at 5.25 GHz).

4. BER PERFORMANCE WITH AWGN

This section illustrates how the double FM system performsunder AWGN conditions. Figure 9 shows a receiver block di-agram useful for calculating the probability of error Pb of thedigital output signal.

It consists of a cascade of the following blocks:

(1) wideband FM demodulator;(2) subcarrier filter;(3) subcarrier demodulator.

The wideband FM demodulator acts as an SNR converter.The SNR at the wideband demodulator output is a nonlinearfunction of the input SNR:

SNRDEMOD = Ψ(SNRRF

). (28)

Next, the bandwidth of the demodulated signal BDEMOD islimited to the bandwidth BSUB of the FSK subcarrier signal in

VS

VN

ΣVI

τ

VO

Figure 10: FM demodulator model for calculating the SNR transferΨ.

the subcarrier bandpass filter. The subcarrier SNR (SNRSUB)is given by

SNRSUB = BDEMOD

BSUBSNRDEMOD . (29)

This SNRSUB determines the probability of error at the sub-carrier demodulator output. Assuming binary FSK with co-herent detection and a modulation index βSUB = 1 for thesubcarrier modulation scheme, the probability of error Pbequals [8]

Pb = 12

erfc

√SNRSUB

2. (30)

The calculation of the SNR transfer function Ψ of the FMdemodulator is based upon the model presented in Figure 10.

The exact mathematical calculation is rather tedious [9];it appears that the noise conversion depends on the offset o.The following intuitive calculation assumes a frequency off-set o = 0 and autocorrelation of the noise RN (τ) = 1 for val-ues of the delay τ according to (12). Although a rather coarseapproximation, it yields results that correspond well to mea-surements made on a hardware prototype of the widebanddemodulator.

The FM demodulator’s input voltage VI consists of thesum of signal voltage VS and noise voltage VN . By definition,the signal power S and noise power N are equal to S = V 2

S

and N = V 2N . The bandwidth of both signals is BRF = 2(β +

1) fSUB.The FM demodulator’s output voltage VO equals

VO = V 2I =

(VS + VN

)2

= V 2S + V 2

N + 2VSVN

= S + N + 2√S√N

= VSO + VN1O + VN2O.

(31)

The output voltage VO is the sum of one signal term VSO andtwo independent noise terms VN1O and VN2O. The outputsignal power SO and the output noise power NO are as fol-lows:

SO = S2,

NO = N2 + 4SN.(32)


60

50

40

30

20

10

0

−10

−20−25 −20 −15 −10 −5 0 5 10 15

SNRRF (dB)

SNR

SUB

(dB

)

1 kbps

10 kbps

100 kbps

1000 kbps

Figure 11: SNR conversion at various bit rates with AWGN for R =1–1000 kbps, BRF = 1 GHz, and βsub = 1.

The output SNR is then given by

SONO

= Ψ(S

N

)= S2

N2 + 4SN=(S

N

)2( 11 + 4(S/N)

). (33)

This expression can be approximated by

SONO

= Ψ(S

N

)≈(S

N

)2

forS

N 1, (34)

SONO

= Ψ(S

N

)≈ 1

4S

Nfor

S

N 1. (35)

Equation (34) corresponds to the operation below thresh-old, while (35) corresponds to the operation above thresh-old. Assuming a flat spectrum for the noise terms VN1O

(noise × noise) and VN2O (noise × noise) of BRF, and re-ferring to Figure 9, we can now calculate the SNRSUB at theinput of the FSK demodulator as follows:

SNRSUB= BRF

BSUB

SDEMOD

NDEMOD= BRF

BSUBSNR2

RF

(1

1 + 4 SNRRF

).

(36)

Figures 11 and 12 illustrate results for data rates R from 1 to1000 kbps. The subcarrier modulation index βSUB is constantand equal to 1, resulting in a subcarrier bandwidth BSUB =2R. Figure 11 shows the SNRSUB as a function of the RF SNR(SNRRF). A 10-fold increase in data rate results in a 10-foldincrease of the subcarrier bandwidth and gives a 10 dB shiftdownwards in the SNR curve of Figure 11.

Figure 12 shows the probability of error Pb as a functionof the SNRRF for the UWBFM system with constant 1 GHzbandwidth for various data rates. For comparison, the figurealso shows the probability of error for a narrowband binaryFSK system occupying an RF bandwidth BSUB.

In a narrowband frequency modulation (NBFM) system,no SNR conversion occurs and therefore, the four curves co-incide.

100

10−1

10−2

10−3

10−4

10−5

10−6

10−7

−30 −25 −20 −15 −10 5 0 5 10 15

SNRRF (dB)

Pb

1 kbps

10 kbps

100 kbps1000 kbps

UWBFM2-FSK

Figure 12: Probability of error Pb for various bit rates for an FSKand a 1 GHz bandwidth UWBFM system (R = 1–1000 kbps, BRF =1 GHz, and βsub = 1).

A fair comparison between UWBFM and FSK can bemade for signals having equal signal power (resulting inequal energy per transmitted bit Eb) and equal receiver noisesingle-sided power density N0. Figure 13 shows the results ofsuch a comparison; the probability of error Pb is shown as afunction of Eb/N0. It can be concluded that for LDRs, thereis a considerable penalty in the receiver performance.

The reason is that for LDRs and error probability valueshigher than 1E-6, the wideband FM demodulator is alwaysoperating below threshold. At MDRs, the situation gets bet-ter and the difference between FSK and UWBFM lowers to11 dB at a data rate of 1000 kbps.

At even higher data rates, the difference remains con-stant, since now the wideband FM demodulator operatesabove threshold and the SNRSUB increases linearly with theSNRRF. The performance degradation is also a function ofthe subcarrier modulation index. A smaller value of βSUB

lowers the performance penalty. Figure 14 illustrates thisphenomenon for a fixed data rate of 1000 kbps and subcar-rier modulation index values of 0.5, 1, 2, 4, and 8.

What does this imply for the link budget of a typicalUWBFM communications system operating at 4 GHz with abandwidth of 1 GHz, and data rate R, subcarrier modulationindex βsub = 1, at an error probability Pb of 1E-6?

The answer is relatively straightforward in terms ofpathloss, defined as the difference (dB) between the trans-mitted power PTX and the received power PRX as

PLdB = 10 log10

(PTX

PRX

). (37)

Substituting BSUB = 2R into (36), it is straightforward to cal-culate the required SNRRF and obtain the SNRSUB of 14 dBrequired to obtain Pb = 1E-6. One finds values for SNRRF

between −22 dB(R = 1 kbps) and −5 dB(1000 kbps).


100

10−1

10−2

10−3

10−4

10−5

10−6

10−7

0 10 20 30 40 50

Eb/N0 (dB)

Pb

1 kbps

10 kbps

100 kbps

1000 kbps

UWBFM2-FSK

Figure 13: Probability of error Pb comparison for UWBFM andFSK (R = 1–1000 kbps, BRF = 1 GHz, and βsub = 1).

100

10−1

10−2

10−3

10−4

10−5

10−6

10−7

0 5 10 15 20 25 30 35 40

Eb/N0 (dB)

Pb

βSUB = 0.5 βSUB = 8

UWBFM2-FSK

Figure 14: Probability of error Pb comparison for UWBFM andFSK with βSUB values of 0.5, 1, 2, 4, and 8 (R = 1000 kbps).

The equivalent receiver’s input noise power PNRX for anRF bandwidth BRF and a receiver noise figure NF is equal to

PNRX = kT NFBRF. (38)

Expressed in dBm and with the noise figure expressed in dB,(38) yields

PNRX,dBm = 10 log10

(kTBRF

)+ NFdB +30 (dBm). (39)

Assuming a receiver noise figure NF of 3 (5 dB), this yields anequivalent noise power PNRX = −79 dBm at the receiver in-put. Figure 15 shows the received power as a function of the

−40

−50

−60

−70

−80

−90

−100

−110

−120100 101 102 103

Distance (m)

Pow

er(d

Bm

)

PNRX

Figure 15: Received signal PRX (dBm) as a function of distance un-der free-space propagation conditions at 4 GHz and equivalent in-put receiver noise power.

Table 3: Required SNRRF, allowable pathloss, and equivalent free-space range to obtain Pb = 1E-6 for various data rates assumingAWGN conditions, and βSUB = 1.

Data rate R SNRRF PL dFS

(kbps) (dB) (dB) (m)

1 −22 90 183

10 −17 85 106

100 −11 79 52

1000 −5 73 25

distance under free-space propagation conditions, assum-ing isotropic antennas with an efficiency of 100%. Transmitpower equals −11 dBm, the maximum power allowed for aUWB system with 1 GHz bandwidth.

The pathloss can be written as

PLdB = 10 log10

(PTX

SNRRF PNRX

). (40)

Under free-space propagation conditions, using a subcarrierdeviation βsub = 1, and a bit rate varying between 1 and1000 kbps, a pathloss (at the center frequency fc) between 90and 73 dB can be dealt with, as shown in Table 3.

Table 3 also shows the equivalent range dFS that can becovered under free-space propagation conditions. These fig-ures indicate a good link margin for LDR and MDR WPANapplications.

These results reveal that there is room for LDR and MDRUWBFM systems. Their performance is degraded by 10–15 dB with respect to narrowband FM systems, however inreturn, the UWBFM provides robustness against interfer-ence and multipath that is not available in narrowband FSKsystems.


5. MULTIUSER UWBFM SYSTEMS AND ROBUSTNESSAGAINST NARROWBAND INTERFERENCE

In a traditional FM receiver, a limiting device removing AMcomponents precedes an FM demodulator. Simultaneous de-modulation of multiple FM signals is not possible in sucha system. Usually, simultaneous demodulation is consideredundesirable. In an FM broadcast radio, it clearly makes nosense to demodulate all of the FM broadcast stations simul-taneously.

In a WPAN, a limited number of users may need to com-municate together at the same time. This can be accom-plished using time-division multiple-access (TDMA) tech-niques as in the GSM system, for example, which requiressynchronization between the users. In a WPAN application,it would be much more elegant if little or no coordinationbetween the users is required.

In UWBFM, an FDMA technique is used assigning dif-ferent subcarrier frequencies to different users. Using an FMdemodulator without hardlimiting, allows for simultaneousdemodulation of multiple FM signals. We will now showmathematically that this is possible. We assume that the de-modulator input signal is the sum of two UWBFM signalsV1 and V2 of amplitudes A1 and A2, phases ϕ1 and ϕ2, car-rier frequencies fc1 and fc2, and message signals m1 and m2

having a subcarrier frequency of fm1 and fm2:

V(t) = V1(t) + V2(t)

= A1 sin(ωc1t + ϕ1(t)

)+ A2 sin

(ωc2t + ϕ2(t)

).

(41)

The demodulator output signal VDEMOD now equals

VDEMOD(t) = V1(t) + V2(t)V1(t − τ) + V2(t − τ)

= V1(t)V1(t − τ) + V2(t)V2(t − τ)

+ V1(t)V2(t − τ) + V1(t − τ)V2(t).

(42)

The first two terms of (42) represent the useful signal U(t):

U(t) = V1(t)V1(t − τ) + V2(t)V2(t − τ), (43)

which after lowpass filtering yields the sum of the two mod-ulating signals m1 and m2.

The last two terms of (42) constitute the residue W(t):

W(t) = V1(t)V2(t − τ) + V2(t)V1(t − τ). (44)

This residue may corrupt the useful signal U(t).We will now investigate the low-frequency terms of the

residue W(t) (obtained by a lowpass filter at half the carrierfrequency fc) in more detail. Writing out the individual low-frequency terms yields

WLF(t) = A1A2

2cos

(ωc1−ωc2

)t+ωc2τ+ϕ1(t)−ϕ2(t−τ)

+A1A2

2cos

(ωc2−ωc1

)t+ωc1τ+ϕ2(t)−ϕ1(t−τ)

.

(45)

These two terms represent two FM signals centered at fre-quency (ωc1 − ωc2) modulated by (m1(t) − m2(t)) whichis the difference of the modulating signals. In other words,the residue terms are the result of asynchronous down-conversion of signals V1 and V2 by each other, resulting in nodemodulation. The spectrum of this residue is spread over abandwidth of 2(∆ f1 + ∆ f2).

Figure 16 illustrates the resulting signals for the simul-taneous demodulation of the two UWBFM signals of equalamplitude and carrier frequency as shown in Table 4. Thetwo UWBFM spectra are completely overlapping. The FMdemodulator is a delay-line demodulator with N = 3 andfc = 4 GHz. Its useful bandwidth BDEMOD is 2.67 GHz.

Figure 16a shows the spectrum of the RF signal V1 + V2

at the delay-line demodulator input. The two componentsin the demodulator output signal, that is, the useful signalULF(t) and the residue WLF(t), are shown in Figure 16c. Theinstantaneous frequency of the residue is proportional to thedifference of the two modulating signals m1 and m2 as shownin Figure 16b. The envelope of the residue is proportional tothe sum of the two modulating signals m1 and m2.

Figure 17a shows the spectrum of the useful part ULF( f )with peaks at the subcarrier frequencies and their thirdharmonics. Figure 17b shows the spectrum of the residueWLF( f ). Only a small part of the residue power falls withinthe subcarrier bandwidth.

It can be seen that the residue is spread across a band-width of 2(∆ f1 + ∆ f2) = BRF = 1200 MHz. Defining signalV1 as the wanted signal and signal V2 as the interference, theRF signal-to-interference ratio SIRRF is as follows:

SIRRF = 20 log10

(A1

A2

)(dB). (46)

For the case where A1 = A2(SIRRF = 0 dB), the total residuepower (between 0 and BRF) is equal to the signal power ofone demodulated subcarrier signal.

Making the approximation that the spectrum of theresidue is flat over its bandwidth, the subcarrier signal-to-interference ratio SIRSUB in the subcarrier filter bandwidthBSUB, can be approximated as

SIRSUB ≈ SIRRF−10 log10

(BRF

BSUB

)(dB). (47)

Consider the case where BRF = 1200 MHz and BSUB =1 MHz; this yields SIRSUB = SIRRF−31 dB, which is sufficientfor demodulation of the subcarrier with an error probabilityPb = 1E-6, down to an SIRRF as low as −17 dB.

For the more general case, where the demodulator inputequals the sum of N input signals V1(t),V2(t), . . . ,VN (t),the output signal of the wideband demodulator will com-prise (according to (42)) N2 terms: N terms of the formVi(t)Vi(t − τ) (i = 1, 2, . . . ,N) for the useful signal U ,and N(N − 1) terms of the form Vi(t)Vj(t − τ) (i =1, 2, . . . ,N ; j = 1, 2, . . . ,N , j = i) for the residue W . Clearly,the residue power increases with the number of users. Thesubcarrier SIR will decrease due to this multiple-access in-terference. Like in a direct-sequence CDMA system, themultiple-access interference will limit the number of users.


0

−20

−40

−60

−803 3.5 4 4.5 5

Frequency (3–5 GHz)

Pow

ersp

ectr

um

(dB

)

(a)

2

1

0

−1

−2

0 100 200 300 400 500

Time (0–500 ns)

Am

plit

ude

(b)

2

1

0

−1

−2

0 100 200 300 400 500

Time (0–500 ns)

Am

plit

ude

ULF(t)WLF(t)

(c)

Figure 16: (a) Spectrum of the RF input signal V1 +V2 (S(V1 +V2)),(b) difference of modulating signals m1 and m2 (m1(t)−m2(t)), and(c) useful signal ULF and residue WLF.

The case of a narrowband or CW interferer is like a sec-ond user with little or no modulation. Figures 18 and 19 illus-trate the case of a UWB signal with a subcarrier frequency of20 MHz, bit rate of 2 Mbps, deviation ∆ f = 600 MHz, and a20 dB stronger CW interferer. The UWB signal and interfer-ence are characterized by the parameters shown in Table 5.The high value for the subcarrier frequency and bit rate werechosen to reduce the simulation time of this SPICE simula-tion. The circuit is simulated over an interval of 8 microsec-onds and the maximum value of the time step used by SPICEequals 20 picoseconds.

Figure 18 shows the spectra of the input and output sig-nals VRF and VDEMODLP of the wideband demodulator.

Figure 19a shows the spectrum of the filtered subcar-rier signal prior to FM demodulation. Figure 19b shows the

Table 4: Parameters of the two UWBFM signals.

UWBFM signal 1 2

RF center frequency 4 GHz 4 GHz

RF voltage 1 V 1 V

Subcarrier frequency 6 MHz 8 MHz

Subcarrier deviation 600 MHz 600 MHz

transmitted and received message signals m and mRX . It canbe appreciated that the UWBFM system can easily cope withthe CW interferer.

The robustness to more realistic interfering signals, likeinband UWB, multiband OFDM signals, and out-of-bandWLAN signals will be addressed in future research.

6. MULTIPATH PERFORMANCE OF UWBFM

As mentioned in Section 3, multipath is a major source ofAM components. The UWBFM signal is inherently scan-ning the frequency-dependent transfer function H( f ) of thechannel. We will first illustrate this for a single-reflectionchannel described by the 2-path model [10] with transferfunction

H2( f ) = 1 + a2ejα2e− j2π f τ2 . (48)

The magnitude of this transfer function equals

∣∣H2( f )∣∣ = √1 + a2

2 + 2a2 cos(2π f τ2 − α2

). (49)

Figure 20 shows the magnitude of this transfer function forparameter values a2 = 0.4, α2 = 0, and τ2 = 1312.5 pi-coseconds. The UWBFM signal is centered at 4 GHz and hasa deviation ∆ f = 600 MHz and a sinusoidal subcarrier signalm(t) of frequency fm = 1 MHz. The instantaneous frequencyof this UWB signal is also sinusoidal with time and varies be-tween 3.4 and 4.6 GHz.

The UWB signal is scanning the frequency-dependentchannel transfer function H( f ) with the rhythm of themodulating signal m(t). The amplitude A(t) of the receivedUWB signal VRX(t) will vary accordingly. Figure 21 showsthe subcarrier signal m(t), the time-varying amplitude A(t)of the receiver’s input signal, as well as the resulting FMdemodulator output voltage VDEMODLP(t); the latter is equalto the output voltage without multipath multiplied by A2(t).

Figure 22 shows the spectrum of the demodulated signal.The harmonics result from the time-varying amplitude ofthe received signal. This example clearly illustrates the pres-ence of the second harmonic of the subcarrier (6 dB belowthe fundamental in this particular example).

In order to illustrate the performance of UWBFM formore complex channels, a case with an 8-path channel modelwith transfer function H8( f ) has been simulated with

H8( f ) =8∑i=1

aiejαi e− j2π f τi . (50)


0

−20

−40

−600 20 40 60 80 100

Frequency (0–100 MHz)

Pow

ersp

ectr

um

(dB

)

(a)

0

−20

−40

−600 0.5 1 1.5 2


Pow

ersp

ectr

um

(dB

)

(b)

Figure 17: (a) Spectrum of the useful part ULF( f ) and (b) spectrum of the residue WLF( f ) of the demodulated signal.

40

20

0

−20

−40

−603 3.5 4 4.5 5


Pow

ersp

ectr

um

(dB

)

(a)

10

0

−10

−20

−30

−40

−500 0.2 0.4 0.6 0.8 1


Pow

ersp

ectr

um

(dB

)

(b)

Figure 18: Spectrum of wideband demodulator input and output signals (a) VRF( f ) and (b) VDEMODLP( f ) for a 20 dB stronger CW interferer.

0

−20

−40

−60

−80

−100

−1200 0.5 1 1.5 2 2.5 3 3.5 4

Frequency (0–100 MHz)

Pow

ersp

ectr

um

(dB

)

(a)

0 50 100 150 200 250 300

Time (0–8 µs)

Am

plit

ude

m(t)

mRX(t)

(b)

Figure 19: (a) Spectrum of the subcarrier demodulator input signal VDEMODLP( f ) and (b) transmitted message m(t) and received messagemRX(t).

Table 5: Parameters of the UWBFM and CW jammer signal.

Signal UWBFM CW

RF center frequency 4 GHz 4 GHzRF voltage 1 V 10 VSubcarrier frequency 20 MHz —Subcarrier deviation 600 MHz —

The coefficients ai,αi, τi are given in Table 6. This modelrepresents a short-range communications channel with oneattenuated direct component plus seven delayed compo-nents.

Figure 23 shows the magnitude |H8( f )| of the frequency-domain transfer function of this 8-path channel. Figure 24shows the subcarrier signal m(t), the time-varying amplitudeA(t) of the receiver’s input signal, as well as the resulting FMdemodulator output voltage VDEMODLP(t) for this more com-plex channel.

Clearly, A(t) is changing rapidly with time. Figure 25shows the spectrum of the demodulated signal. The secondharmonic of the subcarrier is still present and the number ofharmonics has significantly increased.

Since the fundamental of the subcarrier frequency isalways present, the demodulation of this signal is not af-fected. The bandpass filtering already present to implement


1.6

1.4

1.2

1

0.8

0.6

0.43 3.5 4 4.5 5

Frequency (GHz)

Mag

nit

ude

Figure 20: Magnitude |H2( f )| of the channel transfer function.

2

1.5

1

0.5

0

−0.5

−1

−1.5

−20 0.2 0.4 0.6 0.8 1

Time (µs)

Am

plit

ude

m(t)A(t)VDEMODLP(t)

Figure 21: Original modulating signal m(t), time-varying ampli-tude A(t), and demodulator output signal VDEMODLP(t) for the 2-path channel.

the subcarrier filtering will remove the subcarrier harmonicsprior to demodulation. UWBFM is therefore robust to mul-tipath.

However, the second and higher harmonic componentswill camouflage any useful signal whose subcarrier is at theirfrequency. Figure 26 illustrates what may happen for theLDR-MDR UWBFM system using subcarrier frequencies be-tween 1 and 2 MHz as shown in Figure 4. The second har-monic components fill up the frequency range between 2 and4 MHz.

20

10

0

−10

−20

−30

−40

−50

−60

−70

−80−40 −30 −20 −10 0 10 20 30 40

Frequency (MHz)

Pow

ersp

ectr

um

(dB

)

Figure 22: Spectrum of the demodulator output signalS(VDEMODLP) in the case of a 2-path channel with a2 = 0.4.

Table 6: Coefficients of the 8-path channel model.

ai αi τi (ns)

0.30 0 0

0.11 0 2.5

0.21 0 3.5

0.33 0 4.0

0.50 0 4.688

0.37 0 5

0.22 0 5.5

0.10 0 6.5

If no form of equalization is used, the useful subcarrierfrequency range is limited to a single octave. Equalization isthe subject of further investigations.

7. CONCLUSION

A novel frequency-domain UWB technology has been pre-sented. UWBFM uses double FM: low-modulation indexdigital FSK followed by high-modulation index analog FMto create a constant-envelope UWB signal whose spec-tral density is lowered by a factor equal to the modula-tion index β. The UWBFM center frequency and band-width can be easily controlled and the spectral roll-off issteep.

The performance degradation compared to NBFM sys-tems is between 10 and 15 dB in terms of probability oferror performance and depends on the subcarrier modu-lation index. Despite this degradation, a 1 GHz bandwidthUWB communications system operating at a center fre-quency of 4 GHz has a range of 25 m under free-space


2.5

2

1.5

1

0.5

03 3.5 4 4.5 5

Frequency (GHz)

Mag

nit

ude

Figure 23: Magnitude |H8( f )| of the channel transfer function.

2

1.5

1

0.5

0

−0.5

−1

−1.5

−20 0.2 0.4 0.6 0.8 1

Time (µs)

Am

plit

ude

m(t)A(t)VDEMODLP(t)

Figure 24: Original modulating signal m(t), time-varying ampli-tude A(t), and resulting demodulator output signal VDEMODLP(t) forthe 8-path channel.

propagation conditions when operating at 1 Mbps and100 m at 10 kbps, yielding a good link margin for LDRand MDR WPAN applications. Moreover, UWBFM pro-vides robustness against interference and multipath, espe-cially for low bit rates, that is not available in NBFM sys-tems.

Characterization, modeling, and subsequent mitigationof in-band interference from other UWB systems like im-pulse radio and multiband OFDM, as well as strong out-of-band signals like WLANs will be addressed in future re-search.

Multiple subcarriers can accommodate multiple users.Simultaneous demodulation of the various UWBFM sig-

20

10

0

−10

−20

−30

−40

−50

−60

−70

−80−40 −30 −20 −10 0 10 20 30 40

Frequency (MHz)

Pow

ersp

ectr

um

(dB

)

Figure 25: The spectrum of the demodulator output signalS(VDEMODLP) for the 8-path channel.

10

5

0

−5

−10

−15

−20

−25

−30

−35

−401 1.5 2 2.5 3 3.5 4

Frequency (MHz)

Pow

ersp

ectr

um

(dB

)

1 2 3 4

HD2(1) HD2(2) HD2(3) HD2(4)

Figure 26: Effect of second harmonic distortion on the spectrumS( f ) after the wideband FM demodulator for a 1–2 MHz subcarriersystem.

nals is achieved in a wideband FM demodulator that isnot preceded by any limiting device. However, due to thesquaring action of the demodulator, the dynamic rangeof the demodulated signal expands, requiring steep sub-carrier filtering. The number of users is also limited bymultiple-access interference, which is a subject for future re-search.

Multipath introduces envelope variations in the ampli-tude of the received signal that result in harmonic distor-tion of the demodulated UWBFM signal. Without equal-ization, the useful subcarrier range is limited to one oc-tave. Equalization techniques will also be addressed in futureresearch.


ACKNOWLEDGMENTS

This work was partially carried out in the context of the SixthFramework IST Project 507102, “MAGNET” (My PersonalAdaptive Global NET). The IST Program is partially fundedby the EC. The authors would like to thank the EC for theirsupport.

REFERENCES

[1] Federal Communication Commission (FCC), “Revision ofpart 15 of the commission’s rules regarding ultra widebandtransmission systems,” First Report and Order, ET Docket98–153, FCC 02–48; Adopted:February 2002; Released:April2002.

[2] M. Z. Win and R. A. Scholtz, “Impulse radio: how it works,”IEEE Commun. Lett., vol. 2, no. 2, pp. 36–38, 1998.

[3] A. Batra, J. Balakrishnan, and A. Dabak et al., “Multi-bandOFDM Physical Layer Proposal for IEEE 802.15 Task Group3a,” MultiBand OFDM Alliance SIG, September 2004.

[4] J. Farserotu, A. Hutter, F. Platbrood, J. Ayadi, J. Gerrits, andA. Pollini, “UWB transmission and MIMO antenna systemsfor nomadic users and mobile PANs,” Wireless Personal Com-munications, vol. 22, no. 2, pp. 297–317, 2002.

[5] J. Gerrits and J. Farserotu, “Ultra wideband FM: A straight-forward frequency domain approach,” in Proc. 33rd EuropeanMicrowave Conference (EuMC ’03), vol. 2, pp. 853–856, Mu-nich, Germany, October 2003.

[6] H. Taub and D. Schilling, Principles of Communication Sys-tems, McGraw-Hill, New York, NY, USA, 1971.

[7] M. H. L. Kouwenhoven, High-Performance Frequency-Demodulation Systems, Delft University Press, Delft, theNetherlands, 1998.

[8] J. G. Proakis, Digital Communications, McGraw-Hill, NewYork, NY, USA, 3rd edition, 1995.

[9] M. H. L. Kouwenhoven, “An analysis of the quadrature andmath demodulator in the presence of noise,” Internal Rep.,Electronics Research Laboratory, Delft University of Technol-ogy, Delft, the Netherlands, November 1995.

[10] R. Vaughan and J. B. Andersen, Channels, Propagation andAntennas for Mobile Communications, the Institution of Elec-trical Engineers, London, UK, 2003.

John F. M. Gerrits was born in Leiden, TheNetherlands. He received the M.S.E.E. de-gree from Delft University of Technology,The Netherlands, in 1987. His final the-sis was on the design of integrated high-performance harmonic oscillator circuits.In 1988, he joined the Philips T&M Divi-sion in Enschede, The Netherlands, wherehe designed integrated oscillator and data-acquisition systems for oscilloscope appli-cations. In 1991, he joined CSEM, where he has been involvedin both system and circuit design of a single-chip low-powerVHF radio receiver for hearing aid applications and of a single-chip UHF transceiver for ISM applications. His current workinvolves system and circuit design of UWB radio systems, RFand EM simulation techniques, and measurement methodol-ogy. He has recently started a Ph.D. on the fundamental as-pects and practical realizations of UWBFM at Delft Univer-sity of Technology. He is an Editor and coauthor of the book

Low-Power Design Techniques and CAD Tools for Analog and RF In-tegrated Circuits, published by Kluwer in 2001. He holds 3 Euro-pean and 1 US patents.

Michiel H. L. Kouwenhoven was born inDelft, The Netherlands, on July 8, 1971. Hereceived the M.S. degree in electrical engi-neering from Delft University of Technol-ogy in 1993, and the Ph.D. degree from thesame university in 1998. From 1997 until2000, he was an Assistant Professor at theElectronics Research Laboratory, Delft Uni-versity of Technology, were he worked onvarious subjects including design method-ologies for electronic circuits, noise in nonlinear circuits, RF oscil-lator design, and demodulators. In 2000, he joined the Delft De-sign Centre, National Semiconductor Corporation, were he is nowresponsible for the design of RF power detectors, and power con-trol devices. From 1999 until 2001, he served as an Associate Editorfor the IEEE Transactions on Circuits and Systems-II. Dr. Kouwen-hoven received the 1997 Veder Award from the Dutch Foundationfor Radio Science (Stichting Verder) for his Ph.D. work on fre-quency demodulators.

Paul R. van der Meer was born in TheHague, The Netherlands, on July 24, 1970.He received the M.S. and Ph.D. degrees inelectrical engineering from Delft Universityof Technology, Delft, The Netherlands, in1997 and 2003, respectively. In 2001, he co-founded Mirage 3D Simulators, a companyinvolved mainly in car driving simulation.From 2002 till now, he worked as a Post-doc. at the Electronics Research Laboratory,Delft University of Technology, in the RF field. Presently, he iscofounding Seagull Simulation Systems, a company that developscockpit systems and parts for research, professional, and consumerflight simulators.

John R. Farserotu received the B.S.E.E. de-gree from the University of Maryland, Col-lege Park, Md, in 1982 and the M.S.E.E. de-gree in communications engineering fromthe George Washington University, Wash-ington, DC, in 1986. He received hisPh.D. from the Delft University of Technol-ogy, Delft, The Netherlands, in 1998. Dr.Farserotu is currently the Head of the Wire-less Communication Section in the SystemsEngineering Division at CSEM, where he is responsible for lead-ing a team of scientists and engineers working on R&D in wire-less communication and implementation of advanced engineer-ing prototypes. His research interests include mobile wireless per-sonal area networks (WPANs), ultra-wideband (UWB) commu-nication, robust modulation and coding, and HAP/satellite com-munication and networking. Dr. Farserotu has authored or coau-thored over 50 publications in major journals and conferences. Heis the coauthor of IP/ATM Mobile Satellite Networks, published byArtech House in 2001. He teaches a course on this subject at theEcole Polytechnique Federale de Lausanne (EPFL). Dr. Farserotu iscurrently the Vice-Chair of the HERMES Partnership, a network ofmajor European R&D centers in the field of wireless and mobilecommunication bringing together over 1000 engineers and scien-tists.


John R. Long received the B.S. degree inelectrical engineering from the University ofCalgary in 1984, and the M.E. and Ph.D. de-grees in electronics engineering from Car-leton University in 1992 and 1996, respec-tively. He was employed for 10 years by Bell-Northern Research, Ottawa (now NortelNetworks), involved in the design of ASICsfor Gbps fibre-optic transmission systems,and employed for 5 years at the Universityof Toronto. He joined the faculty at Delft University of Technol-ogy in January 2002 as a Chair of the Electronics Research Labora-tory. His current research interests include low-power transceivercircuitry for highly integrated radio applications, and electron-ics design for high-speed data communications systems. Profes-sor Long is currently serving on the Technical Program Commit-tees of the International Solid-State Circuits Conference (ISSCC),the European Solid-State Circuits Conference (ESSCIRC), the IEEEBipolar/BiCMOS Circuits and Technology Meeting (BCTM), andGAAS2004 (EuMW). He is a former Associate Editor of the IEEEJournal of Solid-State Circuits. He received the NSERC DoctoralPrize and Douglas R. Colton and Governor General’s Medals forResearch Excellence, and Best Paper Awards from ISSCC 2000 andIEEE-BCTM 2003.

principles and limitations of ultra-wideband fm communications systems

Documents

cd delft

delft university of

applicable toa uwb signal

constantenvelope uwb

double fm

xj delft

multiple users

computer science eemcs