EURASIP Journal on Applied Signal Processing 2005:3, 359–368

c 2005 Hindawi Publishing Corporation

Comparison between Coherent and Noncoherent

Receivers for UWB Communications

Giuseppe Durisi
Istituto Superiore Mario Boella, 10138 Torino, Italy
Email: [email protected]

Sergio Benedetto
Center for Multimedia Radio Communications (CERCOM), Politecnico di Torino, 10129 Torino, Italy
Email: [email protected]

Received 1 October 2003; Revised 1 March 2004

We present a comparison between coherent and noncoherent UWB receivers, under a realistic propagation environment, that takes
into account also the effect of path-dependent pulse distortion. As far as coherent receivers are concerned, both maximal ratio
combining (MRC) and equal gain combining (EGC) techniques are analyzed, considering a limited number of estimated paths.
Furthermore, two classical noncoherent schemes, a differential detector, and a transmitted-reference receiver, together with two
iterative solutions, recently proposed in the literature, are considered. Finally, we extend the multisymbol approach to the UWB
case and we propose a decision-feedback receiver that reduces the complexity of the previous strategy, thus still maintaining good
performance. While traditional noncoherent receivers exhibit performance loss, if compared to coherent detectors, the iterative
and the decision-feedback ones are able to guarantee error probability close to the one obtained employing an ideal RAKE, without
requiring channel estimation, in the presence of static indoor channel and limited multiuser interference.
Keywords and phrases: ultra-wideband communications, noncoherent detection, multisymbol differential detection.

1. INTRODUCTION whose characteristics vary from path to path. Therefore, the

Ultra-wideband (UWB) systems are based on the transmis- received signal can be seen as a train of distorted waveforms
sion of subnanosecond pulses, typically obtained by directly that often show little resemblance with the transmitted pulse
driving an antenna with short electrical pulses. According [6, 7].
to the FCC regulation of February 2002, signals belonging Due to complexity constraints, only a small subset of
to this category are required to possess a −10 dB bandwidth the received replicas is expected to be selected and com-
which exceeds 500 MHz or 20% of its fractional bandwidth bined, a fact that justifies the performance loss illustrated in
[1]. [4, 8, 9, 10] for various selection combining methods. Fur-
Recently, this technology has been considered for both thermore, the presence of pulse distortion increases the com-
adhoc [2] and indoor wireless personal area networks (IEEE plexity of the channel estimation algorithm [11, 12], a topic
802.15.3a). UWB characteristics are claimed to meet the re- that has not been fully analyzed in the literature yet. In gen-
quirements of these applications, in particular, low com- eral, it can be expected that complexity constraints will im-
plexity, low cost, low power consumption, and high data pose suboptimal solutions and determine a further perfor-
rate connectivity [3]. Furthermore, the fine delay resolution, mance loss.
guaranteed by the large signal bandwidth, provides a high ro- A different approach to overcome all the above-
bustness in dense multipath environments [4]. mentioned disadvantages is based on the use of noncoher-
On the other hand, to fully exploit the channel diversity, a ent reception techniques. These techniques do not require
conventional coherent RAKE receiver must be able to capture channel estimation and allow to capture a large amount of
and track the energy associated with a high number of mul- the received energy, despite distortions and multipath prop-
tipath replicas. In [5], it is shown that the number of paths agation. They represent, however, a suboptimal solution, if
to be considered to reach the 85% of the overall energy can compared to coherent receivers, because of the adoption of
sometimes exceed 100. In addition, the radiation and prop- a noisy signal as a reference waveform for the demodulation
agation processes can act on the transmitted pulse as a filter process.
360 EURASIP Journal on Applied Signal Processing

A technique belonging to this category is based on the the transmitter, the channel, and the receiver. In Section 3,
transmitted-reference (TR) or the autocorrelation principle the coherent and noncoherent schemes are presented, ana-
(see [10, 13, 14]). According to this technique, the reference lyzing, in particular, the architectural complexity of each so-
waveform is obtained by averaging over a preamble of un- lution. In Section 4, the simulation results are presented and,
modulated signals. finally, some concluding remarks are given in Section 5.
The same principle is employed by differential receivers
(DRs) [15]. In this case, since the data is differentially modu-
2. SYSTEM DESCRIPTION
lated, the signal associated to the information transmitted at
time n − 1 represents a valid template for the demodulation 2.1. Introduction
of the signal at time n. We consider a UWB system employing binary pulse ampli-
Finally, in [2], some reception schemes, based on the tude modulation (2PAM). The signal transmitted by the user
adoption of energy detectors and orthogonal modulations, k, s̃(k) (t), is divided into blocks of length T seconds, each one
are presented. carrying Nd 2PAM symbols {α(k) j }. In formulae,
All those techniques lead to low-complexity receivers,
able, in principle, to capture a large portion of the transmit-

s̃(k) (t) = s(k)
i (t − iT). (1)
ted energy and less sensible than coherent demodulators to
i
channel variations and synchronization mismatch [15].
Some strategies have been recently proposed to minimize The expression of s(k)
i (t) is related to the transmission tech-
the suboptimality of noncoherent detectors [13]. Assuming nique and it will be detailed in the next sections. From now
the channel static over a block of N, N > 1, transmitted sig- on, we assume that T is chosen such that the propagation
nals, the premise of these strategies is the consideration that channel can be assumed static over this interval.
each received signal contains information that can be used to If Nu users are active and denoting by h(k)i (t) the channel
improve the estimation of the reference waveform. In [13], impulse response associated to the ith signal block transmit-
in particular, TR systems are considered and two maximum ted by the user k, the received signal corresponding to the ith
likelihood (ML) iterative strategies for template estimations block can be written as
are analyzed. The complexity of these techniques is enhanced
by the fact that the iteration process involves the correlation

Nu
 
operation, so that the samples of the received signals must be ri (t) = s(k) (k)
i (t) ∗ hi t − τ (k) + ni (t), (2)
memorized and reprocessed. k =1

As far as traditional differential demodulator systems are

where τ (k) is a random variable modelling the delay between
concerned, a well-known technique to reduce its suboptimal-
the transmitter k and the reference transmitter 1, for which
ity is the multisymbol detection, developed in [16] for nar-
τ (1) is assumed equal to zero; ni (t) is a white Gaussian noise
rowband systems. This method does not require iterations;
process with a two-side power spectral density N0 /2.
its complexity, however, is exponential with the block length
and quadratic with the number of correlation operations. An 2.2. Channel model
established strategy to reduce the complexity of this tech-
nique is based on feeding back to the demodulator the es- A fair general expression to describe a block-static UWB
timates on a certain number of previously received symbols. channel is the following one [7]:
This drastically simplifies the demodulator operations. −1
L

In this paper, we present a comparison in terms of com-  
h(t) = al pl t − τl , (3)
plexity and performance of coherent and noncoherent re- l =0
ceivers for UWB communications. To this aim, we propose
a simple channel model, based on [5], able to take into ac- where pl (t) is the impulse response of the filter associated to
count the pulse-based distortions effect due to propagations. the lth propagation path. This permits to take into account
After briefly describing the principal coherent and noncoher- the distortions caused by the physical phenomena related to
ent strategies available in the literature and discussing their the propagation of the pulses. In our analysis, the set of de-
complexity, we theoretically analyze how to extend the mul- lays and amplitudes {al , τl }Nl=−01 is generated according to the
tisymbol concept to the UWB case. In addition, we propose a model proposed by the IEEE 802.15.3a Working Group [5].
decision-feedback (DF) strategy to overcome the complexity This model is based on a modification of Saleh-Valenzuela’s
issue shown by the above-mentioned receiver. [17] one and is able to reproduce the clustering phenomena
Finally, we compare the performance of these systems in observed in several UWB channel measurements. In partic-
terms of error probability. In order to fulfill this task, we ob- ular, in [5], it is underlined that the number of paths with
tained by simulations the bit error rate (BER) curves for both nonvanishing energy can exceed 50.
coherent and noncoherent detectors. In particular, both a The exact characterization of the propagation distortions
single-user and a multiple-access time-hopping scenarios are is a complex task; it is analyzed for example in [7], in which
considered. the shape of the received pulse is obtained via numerical in-
The paper is organized as follows. In Section 2, the ana- tegration for different propagation conditions. Following the
lyzed UWB system is described in its three main components, final remarks of [7], we adopt in this paper a rather simplified
Coherent and Noncoherent UWB Receivers 361

0.4
r(t) 
Tw DEC
0.3 v(t)
Normalized amplitude

Channel Template Bit & code

0.2 estimation generator synchr.

0.1 Figure 2: General structure for an UWB correlation receiver.

0
characterize them according to four complexity parameters:
−0.1 (i) the length of the buffer at the receiver used to memo-
rize the information necessary for the computation;
−0.2 (ii) the operations needed to construct and update the
−1.5 −1 −0.5 0 0.5 1 1.5 template signal;
t (ns) (iii) the number of correlation operations required to de-
x(t) x(t) ∗ p(t), Bw = 3.5 GHz
modulate the data block;
x(t) ∗ p(t), Bw = 5 GHz x(t) ∗ p(t), Bw = 2.5 GHz (iv) the decision rule.

Figure 1: Effect of an ideal lowpass filtering distortion on a sec-

ond derivative of a Gaussian pulse. Three different filter bandwidths
3. RECEIVER STRUCTURES
(Bw ) are considered.
3.1. Pseudocoherent RAKE receivers
model, according to which the path impulse response pl (t) is In case of perfect channel estimation and absence of inter-
approximated by an ideal lowpass filter with bandwidth Bw,l . symbol and multiuser interference, it is well known [18] that
Figure 1 illustrates the effect of the lowpass filtering op- a RAKE receiver is the optimal detection scheme in the sense
eration over the transmitted pulse x(t), modelled as the sec- that it minimizes the probability of error.
ond derivative of a Gaussian pulse and with time duration However, if the knowledge of the channel is not ideal but,
equal to 0.7 nanosecond. It is worth noting that a reduction rather, is acquired through a suitable estimation algorithm,
of the filter bandwidth causes an enlargement of the time du- this structure reduces to a heuristic approximation of the op-
ration of the received pulse. This translates in an increase of timal detection scheme. Adopting the same terminology as
the interpulse interference and a reduction of the correlation in [19], we will refer to this class of receivers as pseudocoher-
between the transmitted and received waveforms. ent RAKE receivers. With a different notation, this particular
structure was analyzed also in [4, 10].
2.3. Correlation receivers We assume that the transmitted signal s(k) i (t) is consti-
tuted by a train of Nd amplitude modulated waveforms
A general symbol-by-symbol correlation receiver structure
for UWB is shown in Figure 2. The received signal is mul- 
d −1
N

tiplied by a locally available template waveform vn (t), 0 < t < s(k) α(k) (k)
i (t) = iNd + j Ex x t − jTb − c j Tc , (5)
Tb , where Tb is the bit time, with Tb < T, and n is the symbol j =0
index. The template signal is generated according to the in-
formation acquired by the bit and multiple-access code syn- where x(t) is the transmitted waveform, normalized such
 +∞
chronization algorithms and, possibly, channel estimation. that −∞ [x(t)]2 dt = 1, Tb is the bit time, and Ex is the energy
The result of the multiplication is finally passed through an per pulse, equal in this case to the energy per bit Eb . Finally,
integrator with integration time Tw and through a decision the block repetition time T is equal to Nd Tb . The sequence
block. (k) N −1
{c j } j =h 0 contains the TH code employed by the user k for
Without loss of generality, we focus on the first transmit- multiple-access purpose. The codes considered in this paper
ted block and drop, for notation simplicity, the block index are the sequences based on quadratic congruence described
i. Assuming that user 1 is the user of interest, the correlator
in [20]. In our analysis, we will impose that 0 ≤ c(k)
j ≤ Nh − 1,
output α(1)
n corresponding to the nth transmitted symbol in-
with Tb = Nh Tc .
side the block is given by We assume that a suitable channel estimation algorithm
 (see, e.g., [11]) is able to provide the template generator with
α̂(1) a reliable information about the delays and amplitudes of the
n = r(t)vn (t)dt. (4)
Tw N p strongest path. However, for complexity reasons, it does
not attempt to estimate the received waveform shape, assum-
In order to compare the complexity of the correlation ing no distortions during the propagation. This estimation
structures that will be analyzed in the next sections, we will algorithm may require the transmission of pilot symbols that
362 EURASIP Journal on Applied Signal Processing

would modify the structure of the transmitted signal, as de- The description of the demodulation process can be sim-
scribed by (5). However, we will not take into account the plified if an equivalent discrete model of the received signal is
presence of this training sequence, in order to keep low the considered. We denote by r̃n the vector containing the sam-
complexity of our model. ples of the received signal, associated to the nth transmitted
If a maximal ratio combining (MRC) technique is em- pulse, inside the integration window Tw :
ployed, the template signal vMRC (t) for the user 1 can be writ-  
ten as r̃n = r̃ nTb + cn(1) Tc , r̃ nTb + cn(1) Tc + Ts , . . . ,

  (10)
vMRC (t) = a(1) (1)  
l x t − τl , (6) r̃ nTb + cn(1) Tc + Nw − 1 Ts ,
l∈BN p

where Tw = Nw Ts and Ts is the inverse of the Nyquist sam-

where BN p = {l0 , l1 , . . . , lN p −1 } is the ordered set of the N p
pling rate. The template signal v(t) can then be expressed as
strongest paths indices. A soft estimate of the nth transmitted
symbol is then given by
 Tw  1

N r −1
α̂(1)
n = r t + nTb + cn(1) Tc vMRC (t)dt. (7) v= r̃i . (11)
0 Nr i=0

The integration time must be chosen such that Tw ≥ τlN p −1 ,

Finally, the soft output of the demodulator is equal to
lN p −1 ∈ BN p . Assuming, instead, that the information passed
by the channel estimator is only partially used by the tem-
plate generator, then simpler combining techniques can be α̂(1) T
n = v r̃n+Nr . (12)
employed. It is possible, for example, to adopt EGC, exploit-
ing only the information about the delays and the sign of the From an implementation complexity point of view, this
amplitudes of the received strongest paths. In this case, receiver requires a buffer capable of containing at least Nw Nr

  samples. The template waveform is then calculated averag-
vEGC (t) = sgn a(1)
l x t − τl(1) . (8) ing on the data contained in the buffer and it is updated on
l∈BN p a blockbase. The number of correlation operations for each
block is equal to Nd . Finally, a symbol-by-symbol threshold
3.2. Transmitted-reference receiver comparison is adopted as decision strategy.
A different approach is based on the transmit-reference prin-
ciple. This concept avoids the channel estimation step, as a 3.3. Differential receiver
previous received signal is employed as template waveform. In the DR, the template waveform employed in the demod-
This approach is clearly suboptimal, if compared to the co- ulation process consists of a delayed replica of a previously
herent reception; however, it possesses an inherent architec- received signal. The correlation operation reveals the am-
tural simplicity that makes it an ideal candidate for low cost plitude variations from one pulse to the other, carrying the
implementations [21]. With this strategy, the transmitted transmitted information. Given the following transmitted
signal within a block consists of N transmitted waveforms, signal
grouped into a preamble of Nr reference signals followed by
Nd data signals. Using the same notation as in [13], the signal N 


transmitted by the user k can be expressed as s(k) (t) = Ex β(k) (k)
j x t − jTb − c j Tc , (13)
j =0
r −1 
N

s(k) (t) = Ex x t − jTb − c(k)
j Tc
j =0 where β0(k) is an arbitrary phase and β(k)
j
(k) (k)
= α j β j −1 are Nd
(9)
d −1
N
  differential 2PAM symbols, with Eb = ((N − 1)/N)Ex , the
 
+ α(k)
j Ex x t − j + Nr Tb − c(k)
j+Nr Tc , output of the integrator can be expressed as
j =0

where this time Eb = Nd Ex /(Nr + Nd ). In order to limit the α̂(1) T

n = r̃n−1 r̃n . (14)
effect of the noise in the demodulation process, a bandpass
filter z(t), of bandwidth B f , is employed at the receiver. Since It is reasonable to expect that the performance of this re-
the template waveform is a noisy signal, the bandwidth of the ceiver, in presence of random hopping codes, will be similar
filter must be chosen so as to trade signal energy with noise to the one of the TR when Nr = 1.
reduction. The architecture of this receiver consists of an Nw samples
Denoting by r̃(t) the signal at the output of the bandpass buffer, fed by a delay block and with symbol-base update.
filter, r̃(t) = r(t) ∗ z(t), the template signal is calculated as an The number of correlations and the decision strategy is equal
average of the reference pulses. to the TR case.
Coherent and Noncoherent UWB Receivers 363

3.4. Iterative transmitted-reference receivers for each block. In addition, the ML-ITR requires an ex-
We focus on the single-user case, and assume that the system tra buffer of Nw samples to memorize the template wave-
parameters are set such as to avoid intersymbol interference. form. On the contrary, the extra buffer length required by
As described in [13], a strategy to improve the performance the GLRT-ITR is equal to Nd Nw , as the receiver constructs a
of TR receivers is based on the adoption of an ML estimation template waveform for each one of the received symbols. The
of the template signal, given the observed block of N vectors template waveform is obtained combining during each itera-
rn . We will refer to this receiver as ITR-ML. Calculating this tion the information contained in the buffer with the previ-
estimator is equivalent to solving ous step correlators outputs. Denoting by Ni the number of
iterations, the ML-ITR requires Nd Ni correlation operations
v = arg max PR0 ,...,RN −1 (r0 , . . . , rN −1 | z). (15) per block, while the GLRT-ITR requires Nd (Nd − 1)(Ni −
z∈RNw 1) + Nd . This increase is again due the fact that a different
template waveform is associated to each one of the received
This expression can be solved iteratively, leading to the fol- symbols. Finally, both receivers adopt a symbol-by-symbol
lowing recursive equation [13]: threshold comparison decision rule.
r −1 d −1
1
N
N

rTj+Nr v(m−1)
v(m) = rj + r j+Nr tanh 2 , 3.5. Multisymbol receivers
Nr + Nd j =0 j =0
N0
Up to now, we analyzed receivers based on a symbol-by-
(16)
symbol decision strategy. However, as noted by Divsalar and
where m is the iteration counter and
Simon in their milestone work [16], this strategy is not opti-
Nr −1 mal if the random parameter that prevents us from using co-
1

v(0) = rj. (17) herent detection (in [16], the channel phase rotation, in our
Nr j =0
case, the entire channel impulse response) is constant over
an interval in which more than two symbols are transmit-
The interpretation of (16) is straightforward. The template ted. The basic idea in [16] is to exploit this time invariance,
signal is obtained as an average of the received signals, where making a joint decision on several symbols, simultaneously,
the modulated ones are weighted according to their reliabil- through an ML sequence estimator.
ity. In fact, Our approach is similar to the one adopted in [16]; how-
ever, two main differences must be underlined. In our case,
rTn+Nr v(m) the modulation technique is not limited to be differential.
E α̂(1)
n | rn+Nr , v
(m)
= tanh 2 . (18)
N0 Furthermore, as already mentioned, the random parameter,
under which the receiver minimizes the probability of er-
In [13], another estimation strategy, based on the gener- roneously detecting the entire information sequence, is not
alized likelihood ratio test (GLRT), is presented. We will refer the channel phase (whose distribution can be assumed uni-
to this receiver as ITR-GLRT. According to this strategy, the form) but the entire channel impulse response. Therefore,
template waveform for the nth transmitted symbol is given the strategy developed in [16] based on the existence of a least
by favorable a priory distribution for the unknown parameter
 [22] cannot be applied directly to our case. We tackle this
vn(±1) = arg max PR0 ,...,RN −1 r0 , . . . , rN −1 | z, bn = ±1). (19) inconvenience considering the problem of jointly detecting
z∈RNw
the information sequence and the channel impulse response.
As in the previous case, the equation can be solved itera- Before analyzing, in details, the receiver structure, it is worth
tively, leading to noting that, since the decision strategy is not a symbol-by-
symbol one in this case, the model depicted in Figure 2 does
1 not apply. However, we will continue using the same nota-
vn(m) =
Nr + Nd − 1 tion as in the previous sections.
  We focus on the single-user case. Each received vector r̃n
N

r −1 d −1
N

rTj+Nr vn(m−1)  (20) in (10) can be decomposed into a useful signal part s and
 
× rj + r j+Nr tanh 2  a Gaussian noise part ni . In particular, according to (9), the
 N0 
j =0 j =0
j =n
following expression holds for 0 ≤ i < N:

with s + ni if i < Nr ,
r̃i =  (22)
r −1
N
αi s + ni if i ≥ Nr .
1
vn(0) = rj. (21)
Nr j =0
For simplicity of notation, we will term by r the concatena-
Evidently, the architectural complexity of these receivers tion of the vectors r̃i , that is, r = [r̃0 , r̃1 , . . . , r̃N −1 ], and with a
is rather high. In particular, they both require a buffer of the vector containing the Nd 2PAM symbols {αi }Ni=d0−1 trans-
length Nw N samples to memorize all the received sequence mitted inside the block.
364 EURASIP Journal on Applied Signal Processing

Starting from the a posteriori probability of r, given s and 3.6. Decision-feedback differential receiver
a and using standard techniques [22], the following expres- The main increase of complexity associated with the MSR is
sion for the log-likelihood function (LLF) of the couple (a, s) due to the maximization of the LLF. A well-known strategy
can be derived: to overcome this problem is based on the DF technique [24].
This concept consists of a symbol-by-symbol decision strat-
r −1
N
d −1
N

    egy, obtained by feeding back the decisions taken on a certain
LLF(a, s) = − r̃i − s2 − r̃i+N − αi s2 . (23)
r number of previous symbols.
i=0 i=0
We start with (27) and assume that an estimate of the
first Nd − 1 symbols of the block is available at the receiver.
The maximization of LLF(a, s) can be carried out in two steps We denote by {α̃i }Ni=d1−1 these estimated values. Substituting
[11]. First, s is varied while a is kept constant. A maximum is into (27), the following decision rule can then be used for
then found for ŝ = ŝ(a). Finally, in the second step, the func- the Nd th symbol αNd :
tion LLF(a, ŝ(a)) is maximized with respect to a. The first step
yields α̂Nd = α̂N −1
 N −2  N −2 j −2

r −1
N
d −1
N


 
N
1 = arg max r̃Ti r̃ j αk +r̃Ti r̃N −1 α̃k α
ŝ(a) = r̃i + αi r̃i+Nr . (24) α∈{±1} i =0 j =i+1
N k =i+1 k=i+1
i=0 i=0
 −2


N

Substituting (24) in (23), it can be shown that = arg max r̃Ti r̃N −1 α̃k α .
α∈{±1} k=i+1
(28)
max LLF(a, ŝ(a))
a Evidently, this is equivalent to employing the same zero
Nr −1 Nd −1 d −2 N

d −1
 threshold decision rule adopted by all the other symbol-by-


 T N

 T
= max α j r̃i r̃Nr + j + αi α j r̃Nr +i r̃Nr + j . symbol techniques analyzed up to now. Generalizing (28),
a
i=0 j =0 i=0 j =i+1 the soft outputs α̂n for the nth symbol in the block can be
(25) calculated as follows:
−1
n
−1
n T
Consequently, the decision rule for the transmitted sequence
a becomes the following one: choose â such that α̂n = r̃i α̃k r̃n . (29)
i =0 k=i+1

â = arg max LLF(a, ŝ(a)). (26) It is interesting to note that, owing to the block structure
a
of the transmitted signal, the demodulator operation in (29)
requires the knowledge of the decisions taken over a progres-
We will term the receiver adopting the above-mentioned de- sively increasing number of previously received symbols. A
cision rule as multisymbol receiver (MSR). slightly different approach, similar to the one employed in
The computation of the LLF of the pair (a, s) can be easily [19], consists of keeping constant for each symbol the dimen-
extended to the case of differential modulation. We will refer sion on the feedback vector. In order to do so, the transmitted
to this receiver as multisymbol differential receiver (MSDR). signal format must be modified such that the block structure
Using the same notation as in Section 3.3, one can show that, is removed, that is,
in this case, the estimate â of the binary input symbols a is
given by +∞ 


s(k) (t) = Ex β(k) (k)
j x t − jTb − c j Tc , (30)
j =0
−2 N

N
−1 j

â = arg max r̃Ti r̃ j αk . (27) with Eb  Ex for an infinitely long transmitted sequence. In
a i=0 j =i+1 k=i+1
this case, after an initial transient period, it is possible to em-
ploy a demodulation rule with DF constant window, equal
As far as the complexity issue is concerned, We focus for to N − 1 symbols. In particular, the decision-feedback differ-
simplicity on the differential case. As in the previous section, ential receiver (DF–DR) turns out to be equivalent to a cor-
the buffer length is equal to Nw N samples, the amount of relation receiver that uses the following template waveform:
samples necessary to memorize all the received signals for
each block. On the contrary, the receiver does not require
this time a template construction or update. The number of −1
N
−1
n

correlation operations for vn = rn−i α̂ j . (31)

 each block can be derived from i=1 n−i+1
(27) and it is equal to N2 . Finally, the joint decision al-
gorithm consists of finding the maximum of the function The complexity of the template update operation in (31) can
LLF(a, ŝ(a)) : {±1} → R. The complexity of this operation be reduced noting that it admits a recursive solution. In other
is exponential with the block length N [23]. words, it is possible to calculate the template signal for the
Coherent and Noncoherent UWB Receivers 365

symbol n+1, starting from the template signal for the symbol As far as coherent RAKE receivers are concerned, we
n. In formulae, it is possible to show that compare MRC and EGC, assuming that the strongest 10
and 5 paths are perfectly estimated. For the noncoherent re-
 −1
n  ceivers, the window amplitude Tw is set to Nh Tc = 28 nano
vn+1 = α̃n vn − rn−N α̃n− j + rn , (32) seconds and a −10 dB bandwidth bandpass filter is consid-
j =n−N+1 ered. This choice was found to guarantee a good compromise
between the amount of captured energy and noise reduction.
where α̃n is the estimated value of the symbol αn . Like in [19], The delays τ (k) are modelled as uniformly distributed
the absence of a block structure has the convenient side effect random variables over [0, T], and a different realization of
of allowing time-varying channel model. the IEEE 802.15.3 channel model, modified as described in
As far as complexity is concerned, this receiver has the at- Section 2.2, is assigned to each user. The BER curves pre-
tractive feature to conserve the same number of correlation sented in this section are obtained by averaging over the re-
operations and the same decision rule as the DR. However, it sults relative to 20 different indoor scenarios. In the single-
requires a buffer capable to contain all the samples belonging user case, the hopping code has not been simulated.
to the feedback window (Nw N samples), plus the memory As far as noncoherent receivers are concerned, it is pos-
necessary for storing the template signal, and the decisions sible to derive an asymptotical BER curve, to which the per-
taken on the previously received N − 1 symbols. The template formance of iterative or DF structures should be compared.
update operation must be performed symbol by symbol, em- In fact, a successful template estimation should lead to a re-
ploying the low-complexity operation described in (32). We constructed waveform equal to the convolution between the
will refer to this as DF-DR. transmitted signal and the channel impulse response of the
desired user. In formulae,

4. SIMULATION RESULTS −1
L

v(t)−→v̂(t) = a(1) (1)
l pl t − τl ∗ x(t), 0 ≤ t < Tw , (33)
In order to compare the receiver structures described in the l =0
previous sections also in terms of performance (BER), we
simulated a UWB system operating in indoor environment. where the limit is for the block length N and number of iter-
The system parameters are chosen so as to obtain a pulse rate ations m that tend to infinity (the latter is obviously only for
of around 36 Mpulse/s per user. In particular, we set Tc = 4 iterative receivers). More precisely, the signal v̂(t) in (33) can
nanoseconds and Nh = 7. With this parameter choice, the ef- be seen as a matched filter to the convolution of the trans-
fect of intersymbol interference should be in average reduced mitted signal and the channel impulse response, truncated
if the TX-RX pair is linked by a channel defined in [5] as over Tw . We will refer to the receiver employing v̂(t) as tem-
type 1 or 2. These channels are in fact characterized by an plate waveform as an ideal RAKE receiver (ARAKE, follow-
average delay spread of 5 and 10 nanoseconds, respectively. ing the notation of [4], where “A” stands for “all”). In prac-
For simplicity, we do not consider here the effect of channel tical cases, the convergence process is limited by the effect of
coding. the intersymbol and multiuser interference. For comparison
For the block noncoherent schemes, we considered two purposes, we derived also the performance obtained employ-
settings. In the first one, we fix the block length N equal to ing a filter (MF) matched to the transmitted signal only.
10. That corresponds to assume that the multipath chan- In Figure 3, the BER curves for the single-user case are
nel is static over an interval of 0.28 microseconds. All the plotted for pseudocoherent receivers. The MF shows rather
noncoherent block schemes are characterized by a bit rate poor performance due to the low amount of energy the re-
of around 32 Mbps per user that corresponds to Nr = 1. ceiver is able to collect. As already noted in [9], the chan-
The bit rate is instead equal to 36 Mbps for the coherent nel statistics allow to obtain with EGC nearly the same per-
schemes and the DF-DR. In this setting, the small value of formance as MRC. Compared to the performance obtained
N allows us to evaluate by simulation also the performance employing an ideal receiver, able to perfectly reconstruct the
of the MSDR. In the second situation, we enlarged the value received waveform in the observation window Tw , the sub-
of N to 20, keeping Nr = 1; therefore an improvement of the optimal RAKE receivers present a loss, at 10−4 of BER, equal
performance of iterative and adaptive noncoherent schemes to 5.5 and 8.5 dB, for 10 and 5 fingers, respectively.
is expected. However, we were not able to simulate the per- In Figures 4 and 5, we plot the BER curves for noncoher-
formance of the MSDR because of the high computational ent receivers for N = 10 and N = 20, respectively. In both
complexity. cases, the TR and DR receivers show a loss of approximately
As a multipath channel, we considered the model de- 7.5 dB. Therefore, in our setting, these low-complex receivers
scribed in Section 2.2 to take into account the effect of pulse are already able to outperform a pseudocoherent RAKE re-
distortions. In particular, the filter pn (t) in (3) is randomly ceiver equipped with 5 fingers.
chosen between a 5, 3.5, and 2.5 GHz bandwidth ideal low- The ITR-ML techniques show a limited improvement in
pass filter. The transmitted waveform x(t) is a second deriva- performance for N = 10, where its loss from the ARAKE
tive of a Gaussian pulse, with time duration equal to 0.7 nano curve at 10−4 is equal to 6.5 dB, while the loss reduces to
second. 4.5 dB when the block length is enlarged to 20. The results
366 EURASIP Journal on Applied Signal Processing

100 100

10−1 10−1

10−2 10−2

P(e)
P(e)

10−3 10−3

10−4 10−4

0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Eb /N0 Eb /N0

AWGN RAKE-EGC, N p = 10 ARAKE ITR-ML

MF RAKE-MRC, N p = 5 TR ITR-GLRT
ARAKE RAKE-EGC, N p = 5 DR DF-DR
RAKE-MRC, N p = 10

Figure 3: BER of pseudocoherent receivers. Single-user case Figure 5: BER of noncoherent receivers. Single-user case (Nu = 1)
Nu = 1. with block length equal to 20 (N = 20).

100 2 dB, respectively. This is an interesting result, as the com-

plexity of this receiver is definitely lower than the one of the
iterative ones. Finally, in Figure 4, we plot the performance
10−1 of the MSDR able to perform close to the ARAKE receiver,
showing a good convergence to this curve for high Eb /N0 at
the cost of a really high computation complexity.
10−2
It is interesting to point out that for N = 20, the DF-
P(e)

DR is able to perform as well as the pseudocoherent receivers

10−3 with 10 fingers also in the low signal-to-noise ratio region
(corresponding to BER of around 10−2 ). This is an impor-
tant result, as it suggests that this receiver is able to guarantee
10−4 performance similar to a pseudocoherent receiver when ade-
quate channel coding techniques are employed. Therefore, it
represents a promising candidate for UWB communications,
0 2 4 6 8 10 12 14 16 18 20 having good performance and limited complexity.
Eb /N0 In Figures 6 and 7, the results for the multiuser case
are illustrated, for both pseudocoherent and noncoherent
ARAKE ITR-GLRT schemes, for a block length of 20. All curves show a BER floor
TR MSDR caused by the multiuser interference. TR, ITR-ML, and DR
DR DF-DR
ITR-ML
are characterized by really poor performance similar to the
MF. On the contrary, both ITR-ML and DF-DR are capable
Figure 4: BER of noncoherent receivers. Single-user case (Nu = 1) of achieving BER similar to the one obtained by employing a
with block length equal to 10 (N = 10). RAKE receiver equipped with 10 fingers. In this case, the con-
vergence to the ideal ARAKE curve is less evident because of
the effects of the stronger interference.
were obtained by stopping the iterative process after 5 iter-
ations (this value will be kept constant for all the iterative
5. CONCLUSIONS
structures that will be analyzed from now on). Moving to the
higher-complexity ITR-GLRT technique, the loss reduces to In this paper, a comparison between coherent and noncoher-
4 dB and 2.5 dB for N = 10 and N = 20, respectively. ent receivers for UWB systems has been presented under a re-
As far as the DF-DR is concerned, its performance is bet- alistic propagation environment that takes into account also
ter than those of all the noncoherent receivers analyzed up the effect of path-dependant pulse distortion. Two classical
till now, its loss from the ARAKE curve being equal to 3 and noncoherent schemes were considered: a differential detector
Coherent and Noncoherent UWB Receivers 367

100 100

10−1 10−1

P(e)
P(e)

10−2 10−2

10−3 10−3

10−4 10−4
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Eb /N0 Eb /N0

AWGN RAKE-EGC, N p = 10 ARAKE ITR-ML

MF RAKE-MRC, N p = 5 TR ITR-GLRT
ARAKE RAKE-EGC, N p = 5 DR DF-DR
RAKE-MRC, N p = 10

Figure 6: BER of pseudocoherent receivers. Multiple-access case Figure 7: BER of noncoherent receivers. Multiple-access case
(Nu = 6, channel distortion). (Nu = 6) with block length equal to 20 (N = 20).

and a transmitted reference receiver. In addition, the iter- transmission systems,” First Report and Order, ET Docket
ative noncoherent schemes recently proposed in [13] were 98-153, FCC 02-48, Adopted: February 2002; Released: April
analyzed and their complexity was discussed. Furthermore, 2002.
[2] Y. Souilmi and R. Knopp, “On the achievable rates of ultra-
the multisymbol principle was theoretically extended to the wideband PPM with non-coherent detection in multipath en-
UWB case and a limited-complexity decision-feed-back dif- vironments,” in Proc. IEEE International Conference on Com-
ferential receiver was considered as a low-complexity subop- munications (ICC ’2003), pp. 3530–3534, Anchorage, Alaska,
timal implementation of the multisymbol technique. USA, May 2003.
Through simulation, we assessed the performance of all [3] G. R. Aiello and G. D. Rogerson, “Ultra-wideband wireless
the analyzed techniques in terms of BER, both in the single- systems,” IEEE Microwave Magazine, vol. 4, no. 2, pp. 36–47,
2003.
user and multiple-access environments. The results show
[4] D. Cassioli, M. Z. Win, F. Vatalaro, and A. F. Molisch, “Per-
that the DF-DR is able to outperform coherent RAKE re- formance of low-complexity rake reception in a realistic UWB
ceivers equipped with up to 10 fingers, obtaining perfor- channel,” in Proc. IEEE International Conference on Communi-
mance similar to the MSDR while maintaining a really low cations (ICC ’2002), vol. 2, pp. 763–767, New York, NY, USA,
complexity. Therefore, it can be considered a promising can- April-May 2002.
didate for low-cost UWB communications. [5] J. Foerster (ed.), “Channel modeling subcommittee report fi-
Future works will be oriented to the design of efficient nal,” IEEE 802.15-02/490, February 2003.
[6] R. J.-M. Cramer, R. A. Scholtz, and M. Z. Win, “Evaluation
synchronization algorithms for the noncoherent receivers, of an ultra-wide-band propagation channel,” IEEE Trans. An-
and to the study of the effect of channel coding at low signal- tennas and Propagation, vol. 50, no. 5, pp. 561–570, 2002.
to-noise ratios. [7] R. C. Qiu, “A study of the ultra-wideband wireless propaga-
tion channel and optimum UWB receiver design,” IEEE Jour-
nal on Selected Areas in Communications, vol. 20, no. 9, pp.
ACKNOWLEDGMENTS 1628–1637, 2002.
[8] B. Mielczarek, M.-O. Wessman, and A. Svensson, “Perfor-
The authors would like to thank Professor Umberto Mengali mance of coherent UWB rake receivers using different chan-
and Dr. Antonio D’Amico for the helpful suggestions. This nel estimators,” in International Workshop on Ultra Wideband
work has been partially sponsored by MIUR (Italian Ministry systems (IWUWBS), Oulu, Finland, June 2003.
of Education and Research) under the Projects CERCOM [9] J. Zhang, R. A. Kennedy, and T. D. Abhayapala, “Performance
of rake reception for ultra wideband signals in a lognormal-
and PRIMO. fading channel,” in International Workshop on Ultra Wideband
systems (IWUWBS), Oulu, Finland, June 2003.
REFERENCES [10] J. D. Choi and W. E. Stark, “Performance of ultra-wideband
communications with suboptimal receivers in multipath
[1] Federal Communications Commission (FCC), “Revision of channels,” IEEE Journal on Selected Areas in Communications,
part 15 of the commission’s rules regarding ultra wideband vol. 20, no. 9, pp. 1754–1766, 2002.
368 EURASIP Journal on Applied Signal Processing

[11] V. Lottici, A. D’Andrea, and U. Mengali, “Channel estima- Sergio Benedetto is a Full Professor of
tion for ultra-wideband communications,” IEEE Journal on digital communications at Politecnico di
Selected Areas in Communications, vol. 20, no. 9, pp. 1638– Torino, Italy, since 1981. He has been a
1645, 2002. Visiting Professor at the University of Cal-
[12] C. Carbonelli, U. Mengali, and U. Mitra, “Synchronization ifornia, Los Angeles (UCLA), at the Uni-
and channel estimation for UWBs signals,” in Proc. IEEE versity of Canterbury, New Zealand, and
Global Telecommunications Conference (GLOBECOM 2003), is an Adjoint Professor at École Nationale
vol. 2, pp. 764–768, San Francisco, Calif, USA, December Supérieure de Telecommunications in Paris.
2003. In 1998, he received the Italgas Prize for
[13] S. Franz and U. Mitra, “On optimal data detection for UWB Scientific Research and Innovation. He has
transmitted reference systems,” in Proc. IEEE Global Telecom- coauthored two books on probability and signal theory (in Ital-
munications Conference (GLOBECOM 2003), vol. 2, pp. 744–
ian), the books Digital Transmission Theory (Prentice-Hall, 1987),
748, San Francisco, Calif, USA, December 2003.
Optical Fiber Communications (Artech House, 1996), and Princi-
[14] R. T. Hoctor and H. W. Tomlinson, “Delay-hopped
transmitted-reference RF communications,” in Proc. IEEE ples of Digital Communications with Wireless Applications (Plenum-
Conference on Ultra Wideband Systems and Technologies Kluwer, 1999), and over 250 papers in leading journals and confer-
(UWBST 2002), Digest of papers, pp. 265–269, Baltimore, Md, ences. He has taught several continuing education courses on the
USA, May 2002. subject of channel coding for the UCLA Extension Program and for
[15] M. Ho, V. S. Somayazulu, J. Foerster, and S. Roy, “A differential the CEI Organisation. He has been the Chairman of the Commu-
detector for an ultra-wideband communications system,” in nications Theory Symposium of ICC 2001, and has organized nu-
Proc. IEEE 55th Vehicular Technology Conference (VTC 2002), merous sessions in major conferences worldwide. Sergio Benedetto
vol. 4, pp. 1896–1900, Burmingham, Alabama, USA, Spring is the Area Editor for the IEEE Transactions on Communications
2002. for Modulation and Signal Design and a Distinguished Lecturer
[16] D. Divsalar and M. K. Simon, “Multiple-symbol differential of the IEEE Communications Society. Professor Benedetto is the
detection of MPSK,” IEEE Trans. Communications, vol. 38, no. Chairman of the Communication Theory Committee of IEEE and
3, pp. 300–308, 1990. a Fellow of the IEEE.
[17] A. Saleh and R. Valenzuela, “A statistical model for indoor
multipath propagation,” IEEE Journal on Selected Areas in
Communications, vol. 5, no. 2, pp. 128–137, 1987.
[18] J. G. Proakis and M. Salehi, Communication Systems Engineer-
ing, Prentice-Hall, Upper Saddle River, NJ, USA, 2002.
[19] G. Colavolpe and R. Raheli, “Noncoherent sequence detec-
tion,” IEEE Trans. Communications, vol. 47, no. 9, pp. 1376–
1385, 1999.
[20] T. Erseghe, Ultra wide band pulse communications, Ph.D. dis-
sertation, University of Padova, Padova, Italy, 2002.
[21] Y.-L. Chao and R. A. Scholtz, “Optimal and suboptimal
receivers for ultra-wideband transmitted reference systems,”
in IEEE Global Telecommunications Conference (GLOBECOM
2003), vol. 2, pp. 759–763, San Francisco, Calif, USA, Decem-
ber 2003.
[22] H. L. Van Trees, Detection, Estimation, and Modulation Theory
Part I, John Wiley & Sons, New York, NY, USA, 1968.
[23] P. Hoeher and J. H. Lodge, “Turbo DPSK: Iterative differen-
tial PSK demodulation and channel decoding,” IEEE Trans.
Communications, vol. 47, no. 6, pp. 837–843, 1999.
[24] F. Edbauer, “Bit error rate of binary and quaternary DPSK
signals with multiple differential feedback detection,” IEEE
Trans. Communications, vol. 40, no. 3, pp. 457–460, 1992.

Giuseppe Durisi was born in Torino, Italy,

in 1977. He received the Laurea degree
(summa cum laude) from the Politec-
nico di Torino in 2001. In January 2002,
he joined Istituto Superiore Mario Boella,
where he works on the MIUR-financed Na-
tional Project PRIMO. Since January 2003,
he has been pursuing the Ph.D degree at
the Department of Electronics, Politecnico
di Torino. In 2002, he spent six months as a
Visiting Researcher at IMST, Kamp-Lintfort, Germany, working on
the IST FP5 Project Whyless.com. His fields of interest are in ultra-
wideband physical layer design, channel estimation, and multiuser
detection.