Radio over Fiber for Broadband Wireless Access

X. N. Fernando and S. Z. Pinter

ADROIT Group
Department of Electrical and Computer Engineering, Ryerson University, Toronto, ON M5B 2K3, Canada

ABSTRACT - Fiber based wireless access systems support high-speed multimedia in real-time by
combining the capacity of optical fiber with the flexibility of wireless networks. For example, single sub-
carrier multiplexed radio-over-fiber (ROF) link can support wireless LAN, cellular radio, and CATV
simultaneously. We investigate various issues in this scenario in order to provide a cost-effective, high
performance solution for high-speed fiber based wireless access. We devised a system identification
technique for a concatenated fiber-wireless channel (for single and multiuser), and have proposed various
compensation schemes to equalize the time varying linear plus static nonlinear channel. We have also done
various experimental studies on the performance of an ROF link.

1. INTRODUCTION Wi) system. This can be done either in a sub-carrier

Wireless communications is entering a new phase
where the focus is shifting from voice to multimedia
services. Present consumers are no longer interested in
the underlying technology; they simply need reliable
and cost effective communication systems that can
support anytime, anywhere, any media they want.
Furthermore, new wireless subscribers are signing up
at an increasing rate demanding more capacity while
the radio spectrum is limited. As a result, wideband
radio links will become more prevalent in today's
communication systems. To satisfy this increasing
demand, the high capacity of optical networks should
be integrated with the flexibility of radio networks.
This is the focus of the Advanced Radio and Optics
Integrated Technology (ADROIT) group.
multiplexed or baseband plus ROF manner. Although
the fiber has several GHz of bandwidth, the cross
coupling due to nonlinearity impairs the system
performance and our focus is to quantify it and to
come up with solutions. In this paper we provide an
overview of the ROF research done by ADROIT.
2. METHODOLOGY
The ADROIT Fi-Wi solution is shown in Fig. 1. Our
solution increases the frequency reuse and enables
wideband access by providing a micro/pico cell
scenario.

The aforementioned wideband demands can be met

with micro/pico cellular architectures. Specifically, by
using fiber based wireless access schemes such as
radio-over-fiber (ROF), high-speed multimedia can be
easily supported. ROF refers to a fiber optic link
where the optical signal is modulated at radio
frequencies and transmitted via the optical fiber.

The focus of our group ADROIT is to investigate

various signal processing strategies to provide a cost-
effective, high performance solution for high-speed
fiber based wireless access. We have researched key
aspects of the ROF such as: estimation and
equalization strategies of the concatenated fiber-
wireless channel in both single and multiuser
environments [1], [2], [3], nonlinearity compensation
by predistortion and asymmetric linearization of the
ROF link [6]. Currently, we focus on the different
issues of supporting both IS-95 cellular and IEEE
802.11 type Wi-Fi signals over the fiber-wireless (Fi-
Figure 1. ADROIT fiber-wireless solution for wideband
multimedia access.
Here, inexpensive low power radio access points
(RAP) provide wireless access. These robust RAPs are
connected to the central base station via the ROF
links. Several observations can be made from Fig. 1.
First, signal processing should not be done at the RAP
for cost considerations. Therefore channel
compensation should be done at the portable unit or at
the central base station. The concatenated fiber-
wireless channel should be handled jointly either way.
This is a challenging task because of the time varying
multipath wireless channel and nonlinear optical
channel. Furthermore, the uplink and downlink require
different solutions. Secondly, it is desirable not to
modify the portable units in order to allow for
seamless roaming. Thirdly, by performing most of the
signal processing at the central base station, or by
asymmetric distribution of the complexity, the cost
can be shared among many users and helps reduce
overall system cost.
Figure 2. ROF experimental setup. In this photo we have a
Rohde & Schwarz signal generator sending a 1.5 GHz
WCDMA signal through the ROF link and into the Tektronix
WCA380 wireless communication analyzer.
2.1 Estimation
ROF provides an excellent link allowing for high
bandwidth communication of several channels.
However, when the wireless link is in series with the
ROF link–especially in a multiuser environment–
nonlinear distortion of the ROF link, due mainly to the
laser diode (and partly to the high-gain RF amplifier at
the optical receiver), becomes the primary concern.
Also, the dynamic range of the input signal is
adversely affected in the uplink where the received
signal first travels through the wireless channel
(resulting in path losses, fading and shadowing) before
entering the ROF link. In order to limit the effect of
these distortions, estimation, and subsequently
equalization, of the concatenated fiber-wireless system
should be done.
We have performed a complete identification of the
ROF uplink in [1] and [2] in a single and multiuser
environment, respectively. The uplink is a Wiener
system, which consists of a linear part (wireless
channel) followed by a mildly nonlinear part (ROF
link). The wireless channel is identified using
correlation analysis and the nonlinear link is identified
using a least squares polynomial fit.
2.2 Equalization
Once the channel is estimated, an appropriate
equalizer must be devised for the compensation of the
linear and nonlinear parts. We have proposed an
efficient Hammerstein type decision feedback
equalizer (DFE) for the Wiener type uplink in [3].
This equalizer compensates for the linear and
nonlinear distortions separately. This modular
architecture is attractive for commercial
implementation. The receiver consists of a polynomial
filter, which inverse models the optical link, and a
linear DFE arrangement that compensates for the
wireless channel dispersion. The filter parameters
have been optimized and performance analysis has
been carried out in [3]. Currently, the Hammerstein
type equalizer is implemented for a single user,
however, we are working on expanding the
equalization to a multiuser environment. In addition,
innovative polynomial compensation methods are also
being investigated.
2.3 Predistortion and asymmetric linearization
Several approaches have been proposed to
characterize and solve the problem of nonlinear
distortion. Some of these approaches attempt to use a
post nonlinearity recovery block as in [4] or try to
mitigate the nonlinear distortion in the network layer
as in [5]. However, we developed an adaptive
baseband model for the ROF link, and proposed two
different predistortion schemes for the nonlinearity.
One is being implemented now in a field
programmable gate-array. Our proposed baseband
predistortion scheme trains higher order adaptive
filters to inverse model the ROF link [6]. Simulation
results show very good performance improvement for
the phase compensation. For the amplitude
compensation, however, some backoff is required to
protect the laser diode from excessive input power.
2.4 Relative intensity noise (RIN)
We came up with a new expression for the RIN noise
in ROF links. This is a fundamental contribution and
will most likely change the way people analyze ROF
links.
3. RESULTS Figure 4 shows relationships between fiber length,
BER, RF bandwidth and radio cell size. It can be seen
For brevity's sake, we can only discuss some of our that for IS-95 systems (1.25 MHz RF bandwidth),
relevant research results. Regarding the estimation of when the fiber is short (< 5 km) relatively large radio
the Wiener type fiber-wireless channel, we show our cell size is possible. However, this cell size
simulation results in Fig. 3. Here we see a good significantly decreases with fiber length and RF
estimation of the linear wireless channel when using bandwidth.
an input of 16 independent PN sequences and an SNR
of 22 dB.
4. CONCLUSIONS
1.5

1 The projected impact of implementing ROF schemes

is substantial. The deployment of optical fiber
Amplitude

0.5

0 technology in wireless networks provides great

−0.5 potential for increasing the capacity and QoS without
−1 largely occupying additional radio spectrum. From
0 5 10 15 20 25 30 35 40 45
sample the aforementioned benefits, it is obvious that ROF
1.5
technology will become ubiquitous in today's
1
communication industry. The research performed by
ADROIT will definitely help to provide a cost-
Amplitude

0.5

0
effective, high performance solution for present and
future high-speed fiber based wireless access systems.
−0.5

−1
0 5 10 15 20
sample
25 30 35 40 45 REFERENCES
Figure 3. Estimated (top) and actual (bottom) impulse
response of the wireless channel using 16 independent PN 1. X. N. Fernando and A. B. Sesay, "Nonlinear channel
sequences and an SNR of 22 dB. estimation using correlation properties of PN
sequences," in Proceedings of the IEEE Canadian
Conference on Electrical and Computer Engineering,
Our proposed correlation analysis algorithm is first prize winner, Canada, May 2001, pp. 469-474.
efficient in identifying the Wiener type system. The 2. S. Z. Pinter and X. N. Fernando, "Concatenated fibre-
nonlinear link is also successfully estimated in the wireless channel identification in a multiuser CDMA
presence of noise (not shown here, refer to [2]). environment," 2004, under review for publication in IEE
Proceedings Communications.
x 10
6 BW vs Radio Cell Radius 3. X. N. Fernando and A. B. Sesay, "A Hammerstein type
5
equalizer for the Wiener type fiber-wireless channel," in
4.8
Proceedings of the IEEE Pacific Rim Conference on
Communications, Computers and Signal Processing,
4.6 Victoria, Canada, August 2001, pp. 546-549.
4. P. Raziq and M. Nakagawa, "Semiconductor laser's
4.4
nonlinearity compensation for DS-CDMA optical
4.2 Fiber Length = 5km transmission system by post nonlinearity recovery
block," IEICE Transactions on Communications, vol.
BER = 1E−5
Bandwidth

4 Fiber Length = 7.5km E79-B, no. 3, pp. 424-431, Mar. 1996.

5. W. I. Way, "Optical fiber-based microcellular systems:
BER = 1E−4

3.8 Fiber Length = 20km

BER = 1E−3
an overview," IEICE Transactions on Communications,
3.6
vol. E76-B, no. 9, pp. 1091-1102, Sept. 1993.
6. X. N. Fernando and A. B. Sesay, "Adaptive asymmetric
3.4 linearization of radio over fiber links for wireless
access," IEEE Transactions on Vehicular
3.2 Technology, vol. 51, no. 6, pp. 1576-1586, Nov.
2002.
3
0 5 10 15 20 25 30 35 40 7. http://www.ee.ryerson.ca/wincore/ADROIT/adroit.html
Radio Cell Radius

Figure 4. Radio cell radius versus the RF bandwidth of the

Fi-Wi CDMA system.