Sittakul, V., & Cryan, MJ. (2009). A 2.4-GHz wireless-over-fibre
system using photonic active integrated antennas (PhAIAs) and
lossless matching circuits.
Journal of Lightwave Technology
,
27
(14),
2724 - 2731. https://doi.org/10.1109/JLT.2009.2015587
Peer reviewed version
Link to published version (if available):
10.1109/JLT.2009.2015587
Link to publication record in Explore Bristol Research
PDF-document
University of Bristol - Explore Bristol Research
General rights
This document is made available in accordance with publisher policies. Please cite only the
published version using the reference above. Full terms of use are available:
http://www.bristol.ac.uk/red/research-policy/pure/user-guides/ebr-terms/
2724 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 27, NO. 14, JULY 15, 2009
A 2.4-GHz Wireless-Over-Fibre System Using
Photonic Active Integrated Antennas (PhAIAs) and
Lossless Matching Circuits
Vitawat Sittakul, Student Member, IEEE, and Martin J. Cryan, Senior Member, IEEE
Abstract—This paper demonstrates that very low cost wireless-
over-fibre systems can be implemented without RF amplification
by using lossless matching techniques. Single stub matching cir-
cuits were designed on low loss substrates and integrated with pho-
todiode chips and results show that the link gain can be theoret-
ically improved by more than 8 dB. A 2-port link gain model is
developed using Agilent ADS which allows simple wideband mod-
eling of such systems. Finally the system is tested using a “live”
access point both with and without matching and results show that
a maximum RF range of 4 m can be obtained at 2 Mbps data rate
over 317 m of installed MMF with no RF amplification.
Index Terms—Radio-over-fibre, vertical cavity surface emitting
lasers (VCSELs), WLAN.
I. INTRODUCTION
T
HERE is much interest in wireless-over-fibre (WoF) links
operating at 2.4 GHz for the distribution of WiFi signals
within buildings, hotels, campuses and airports [1]–[6]. These
Distributed Antenna Systems (DASs) often employ high spec-
ification lasers and RF amplification in order to achieve good
link gain and large RF ranges from remote antenna to the user.
However, there are opportunities for very low cost modules with
reduced RF ranges giving room-scale coverage of 5–10 m [7],
[8]. In [8] a range of 5 m was shown with a carrier frequency
of 60 GHz using an electroabsorption modulator (EAM) based
scheme which requires no laser at the remote end of the link.
However, EAMs are not low cost and can suffer from low re-
sponsivity and be polarization sensitive. The approach outlined
in [7] shows that low cost vertical cavity surface emitting lasers
(VCSELs) can be used within the photonic active integrated an-
tenna (PhAIA) concept [7], [9] which enables integrated mod-
ules with no RF amplification to be used at the remote end.
The use of VCSELs with low bias current requirements and
the removal of RF amplification and associated DC biasing cir-
cuitry results in a very low DC power budget, in the order of
10 mW, at the remote end which opens possibilities for DC
power-over-fibre operation [10], further reducing implementa-
tion costs. In [7] peer-to-peer operation was shown in combina-
tion with simple matching based on variation in antenna input
Manuscript received April 21, 2008; revised August 15, 2008. First published
April 24, 2009; current version published July 09, 2009.
The authors are with the Photonics Research Group, Electrical and Electrical
Engineering Department, University of Bristol, Bristol BS8 1UB, U.K. (e-mail:
m.cryan@bristol.ac.uk).
Digital Object Identifier 10.1109/JLT.2009.2015587
impedance with contact position on the non-radiating edge of
the antenna. Here, a number of further developments are pre-
sented. First, the noise figure (NF) of different length links is
estimated and the link gain of the in-building fibre is measured
directly. Second, a lossless matching circuit is designed and
implemented for the photodiode to improve the link gain. Fi-
nally, live trials using the University of Bristol WiFi network
and impedance matched PhAIAs are presented.
Fig. 1 shows the generic scenario under consideration. It is
a bidirectional system using two MMF links. The system uses
low cost VCSELs and photodiodes (PDs) operating at 850 nm.
The wireless router is a dual antenna module, one of which
is removed to enable direct coaxial tapping-off of the RF sig-
nals. This approach ensures that local users can still access the
router. The tapped-off signal is then fed to a coaxial splitter
which enables bidirectional operation of the system, albeit with
a reduction in link gain due to splitter loss. These losses could
be removed or reduced with the use of amplifiers or circula-
tors, however this paper is addressing the very simplest, lowest
possible cost scenario as a benchmark. The RF signals are then
fed in to the transceiver module consisting of matching circuits
and, VCSEL and PD mounted on planar substrates. Low cost
fibre alignment fixtures have been used to produce portable units
amenable to testing. At the remote end, PhAIA modules are used
whereby the optical devices are mounted directly on the back-
side of 2.4 GHz planar antennas along with matching circuits. A
remote user is then placed a distance,
, from the PhAIA access
point and signal strength and throughput can be monitored for
different wireless distances and VCSEL bias currents. The effect
of long fibre lengths has also been studied in other recent work
[1], but here we are looking at unamplified system performance.
In this work, a live access point and matching circuits are em-
ployed for both uplink and down link PDs. The VCSELs used
happen to have input impedances that are almost purely real and
close to 50
at 2.4 GHz [7] and therefore simple matching can
be used in this case. The PDs on the other hand are very capac-
itive in nature and thus the use of lossless matching circuits can
remove the resulting mismatch loss and can theoretically im-
prove the link gain by up to 8 dB in both up and down links.
The paper is organized as follows: Section II show results for
a basic link. Section III presents the design and performance
for lossless matching of the photodiodes. Section IV gives de-
sign details of the photonic active integrated antennas. Finally
Section V shows the live access point results in infrastructure
mode.
0733-8724/$25.00 © 2009 IEEE
SITTAKUL AND CRYAN: A 2.4-GHZ WoF SYSTEM USING PhAIAs 2725
Fig. 1. The WoF concept with matching circuits in infrastructure mode.
Fig. 2. The basic link diagram.
II. BASIC LINK
CHARACTERIZATION
A. Link Gain
Two low cost 850 nm VCSELs, (VCSEL1 and VCSEL2)
(TSB-8B12–017) from Truelight (www.truelight.com.tw) were
used in this work. The slope efficiencies of VCSEL 1 and
VCSEL 2 were measured to be 0.40 and 0.42 W/A respectively.
In addition, the link consists of two low cost 850 nm GaAs PDs
from Truelight (TPD-8D12–006) and their responsivities have
been measured to be 0.6 A/W and PD2 to be 0.4 A/W. This
difference is due to variations in the coupling from MMF into
the PD. The input impedances of the PDs and VCSELs have
already been measured in [7].
It is important to initially characterize a short one way link
in order to obtain a reference for longer length in-building links
and to obtain data for use in the calculation of other link pa-
rameters such as noise figure. Therefore, a single VCSEL and
PD link was measured using the set up shown in Fig. 2. The
measured results are shown in Fig. 3 for different VCSEL1 bias
currents.
The results show that a link gain of
can be achieved
at 2.4 GHz. It is useful to confirm that the measured link gain
is close to that predicted theoretically. To predict the high fre-
quency link gain is quite challenging [2] and will not be ad-
dressed here, the low frequency link gain is more straightfor-
ward to predict. When assuming identical source and load im-
pedances in the system, the link gain is given by (1) [7], [11].
(1)
Fig. 3. Link gain at different VCSEL1 bias currents at 2.4 GHz.
where is the slope efficiency of the laser, is the respon-
sivity of the PD,
and are the coupling efficiencies
from laser and PD to MMF respectively,
and are the
impedance of the PD and VCSEL respectively and
is the
loss in the fibre.
Since the slope efficiencies are directly measured into the
MMF, they can be included together with the coupling efficiency
terms and the loss of the short length of MMF is assumed to be
negligible. Substituting a VCSEL1 slope efficiency of 0.4 W/A,
a PD responsivity of 0.6 A/W,
and
(measured VCSEL input resistance [7]), link gain can be calcu-
lated to be
. Fig. 3 shows a measured low frequency
link gain (at 45 MHz) of between
to which is close
to the calculated value. The variation can be explained by the
change in the VCSEL slope efficiency over the range of mea-
sured VCSEL bias currents.
B. Noise Figure of the Link
The noise performance of the link can be expressed in terms
of noise figure. The noise figure of an analog link is a measure of
the degradation of the signal to noise ratio (SNR) between input
and the output of the link. The noise figure, NF, of the link is
defined as [11]
(2)
2726 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 27, NO. 14, JULY 15, 2009
TABLE I
N
OISE
FIGURE AT
2.4 GHZ FOR A
1 m MMF L
INK
(3)
where
is input signal, is the output signal, is input noise
power,
is the output noise power and is the link gain.
The input noise is the thermal noise from a matched resistive
load and can be written as [11]
(4)
where
is Boltzmann’s constant which has a value of
1.38
, is the measured temperature and is
the measurement bandwidth.
The output noise of the link (
) has three sources which are
thermal noise,
, shot noise, and relative intensity noise,
. In a 1 Hz bandwidth they can be defined as follows [11]:
(5)
(6)
(7)
where
is electron charge which has a value of 1.602 ,
is the average photodiode current, RIN is the laser relative
intensity noise in dB,
is the output load resistance and
is the additional noise representing the effects of all the internal
link noise sources at the link output.
In this work,
is considered to be the thermal noise of the
VCSEL. We can calculate
at the link output in the same way
as
which results in an additional term. Substituting (4)
– (7) into (3), the expression for the noise figure becomes
(8)
(9)
Taking
provided by the manufacturer,
, from Fig. 3 at 2.4 GHz, and monitored
PD current,
, into (9), the Noise Figure of the basic link for 1
meter of fibre can be calculated as shown in the Table I.
It can be seen that the noise figure can be reduced quite dra-
matically by decreasing the VCSEL bias current. However, we
have to trade this off with the linearity of the link which can be
worse at low VCSEL current [7].
In reality much longer fibre links will be used and this will
have two competing effects. Firstly there will be decreased link
gain and as (9) shows this will lead to increased noise figure.
However, the extra fibre loss will reduce the PD current and thus
TABLE II
N
OISE FIGURE AT 2.4 GHZ FOR 300 m FIBRE REEL MMF LINK
Fig. 4. 634 m in-building and 600 m fibre-reel MMF losses.
reduce both the shot noise and RIN noise as shown in (6) and (7).
To quantify these effects a 300 m length has been characterized
and the results are shown in Table II.
The results show that longer links have worse noise figure,
but for example at 4 mA it is increased by
, whereas at
6 mA it increases by
. This highlights the number of
parameters which need to be optimized to get good performance
in these types of systems.
C. Multimode Fibre Characterization
Prior to sending data over a link, it is helpful if the fibre being
used is characterized in terms of its frequency response. The
setup diagram is similar to Fig. 2 but with different fibre lengths,
the in-building fibre was looped back on itself to enable VNA
measurements to be performed this results in the length of
.
In this configuration, the link gain of the 634 m In-building
fibre (62.5/125
) was measured and normalized to that of the
1 m MMF link at the same set of VCSEL1 bias currents and is
shown in Fig. 4 along with 62.5/125
fibre-reel MMF links
of a similar length 600 m.
The results show that there is a large amount of extra loss
associated with the in-building fibre. There are a number of
possible reasons for this difference, firstly, the in-building fibre
could have a number of very tight bends which will alter the
differential mode delay (DMD) of the fibre and hence affect
the frequency response. Secondly, MMFs which are nominally
identical can have very different frequency responses due to fab-
rication imperfections.
The results also show that the RF fibre loss is not constant
with VCSEL bias current and with appropriate choice of bias
the loss at working frequency of 2.4 GHz can be improved
SITTAKUL AND CRYAN: A 2.4-GHZ WoF SYSTEM USING PhAIAs 2727
Fig. 5. Layout of the PD on the microstrip carrier with matching circuit.
Fig. 6. The measured and modelled return loss of the PD with and without
(measured only) matching.
by 10 dB. This has been observed previously [12], [13] and is
caused by the VCSEL beam spot and therefore launch position
offset changing as a function of the VCSEL current.
III. L
OSSLESS MATCHING
CIRCUITS IN THE
BASIC LINK
A. Matching Circuit Modeling
Initially simple single stub has been used and the design has
been implemented using Agilent’s Advanced Design System
(ADS). For low cost implementation the matching circuit and
optical device mounting circuit would be implemented on the
same PCB, however, for ease of prototyping, this has not been
implemented here. Thus two PCBs are used, one with a built-in
via hole which carries the optical chip (microstrip carrier) and
one which implements the matching circuit. Since both boards
need the same ground plane the two boards are placed back-to-
back. This also enables easy alignment of the fibre to the optical
devices, whilst allowing access to the matching circuit for fine
tuning. A connection between the microstrip lines on the two
boards is achieved with 3 mm long wire, the inductance of which
is included in the matching circuit. The input impedance of the
PD is measured as
at 2.4 GHz [7] and for
matching purposes this is represented as a model of 10
series
resistor and capacitor,
. The PD circuit model
at 2.4 GHz and a single stub matching is designed on ADS on
low loss substrate (
, ). The results
for this are shown in Fig. 6, where very good matching is seen
at 2.4 GHz.
Fig. 7. The simulated and measured link gain for the basic link versus VCSEL
bias current.
The matching circuit was fabricated and was then fine tuned
by altering the dimensions of the strip lines by adding copper
tape. The new matching circuit can be seen in Fig. 5, the length
of the stub has been increased from 14.14 mm as designed to
15 mm. The results are shown in Fig. 6 and it can be seen that
very good matching is achieved.
B. Link Gain Improvement Using Lossless Matching Circuit
In order to confirm the improvement in link gain obtained by
matching, the simple link was remeasured this time including
the PD matching circuit. This setup is similar to Fig. 2 except
for PD matching circuit included with results shown in Fig. 7.
Fig. 7 shows a measured peak link gain of
at 2.4 GHz
which is increased by 11 dB from the unmatched case. The mis-
match loss (
) which is being removed by matching can be
defined by
(10)
where
, the reflection coefficient of the PD, is given by
(11)
Substituting PD impedance (
) and load impedance
(50
) in the (10) and (11), we can calculate the mismatch loss
to be 8.33 dB which is close to the measured improvement. The
discrepancy here will be due partly to the accuracy of the mea-
surement of the PD input impedance and also to alignment re-
peatability between these two sets of measurements. There is a
