Evaluation of Multi-band Carrier-less Amplitude and Phase
Modulation Performance for VLC under Various Pulse Shaping
Filter Parameters
Petr Chvojka
1
, Paul Anthony Haigh
2
, Stanislav Zvanovec
1
, Petr Pesek
1
and Zabih Ghassemlooy
3
1
Department of Electromagnetic Field, Faculty of Electrical Engineering, Czech Technical University in Prague,
Technicka 2, 16627, Prague, Czech Republic
2
High Performance Research Group, Faculty of Engineering, University of Bristol, BS8 1TH, Bristol, U.K.
3
Optical Communications Research Group, NCRLab, Faculty of Engineering and Environment, Northumbria University,
NE1 8ST, Newcastle upon Tyne, U.K.
Keywords: Carrier-less Am4plitude and Phase Modulation, Roll-off Factor, Filters, Visible Light Communications.
Abstract: In multi-band carrier-less amplitude and phase modulation (m-CAP), the transmitter and receiver pulse
shaping filters have a significant impact on the signal performance. Since m-CAP is an emerging and
promising modulation format for visible light communications (VLC), it is necessary to balance the system
performance and the filter length, due to limited integration density in field programmable gate arrays
(FPGAs). In this paper we investigate the m-CAP VLC system performance for different number of sub-bands
m = {2, 5, 10} under varying finite impulse response (FIR) filter parameters, such as the filter length L
f
and
the roll-off factor β. We show that increasing both β and L
f
improves the bit error rate (BER) performance of
the system substantially. We demonstrate that a BER target of 10
-4
is achieved using L
f
≤ 12 symbols for
m = 5 and 10 for low order subcarriers. Moreover, the BER limit is attained for β = 0.2 by all subcarriers,
except the last two for m = 5 and 10, which is a significant improvement, even considering the slightly
increased excess bandwidth in comparison to the literature.
1 INTRODUCTION
As the number of end user devices connected to the
internet will increase, the demand for high speed
internet connection and the transmission capacity
required will grow exponentially (Zvanovec et al.,
2015). Thus, the next generation networks (5G) face
several challenges such as high spectral and energy
efficiency and high capacity. Visible light
communications (VLC) is an emerging technology,
which is able to meet the mentioned requirements by
utilizing existing solid state lighting (SSL) structures
based on light-emitting diodes (LEDs). VLC provides
users with both illumination and data transmission at
the same time and could be used for localization as
well (Armstrong et al., 2013).
Despite the advantage in using existing SSL
infrastructures, the bottleneck of the VLC system is
introduced by LEDs behaving as a first order low pass
filter (LPF), with a 3 dB cut-off frequency in the MHz
region (Haigh et al., 2014). Thus, spectrally efficient
modulation formats such as orthogonal frequency
division multiplexing (OFDM) are frequently
proposed due to such bandwidth limitations in
conjunction with high capacity requirements. OFDM
supports bit- and power-allocation algorithms that
assign bits and electrical power to each subcarrier
(Bykhovsky et al., 2014). However, OFDM suffers
from a high peak-to-average power ratio (PAPR) that
can lead to signal clipping, due to the LED nonlinear
electro-optic characteristic resulting in the link
performance degradation (Mesleh et al., 2012).
Recently, the research community has turned
attention to carrier-less amplitude and phase (CAP)
modulation format as an alternative to OFDM (Wu et
al., 2012; Wei et al., 2012; Wu et al., 2013). CAP is
similar to quadrature amplitude modulation (QAM).
The main difference is that CAP uses finite impulse
response (FIR) filters to generate carrier frequencies
unlike QAM, which utilizes a local oscillator. This
results in a simpler solution for CAP receivers, since
time-reversed matched filters are deployed. In
previous research it was experimentally shown that
CAP outperforms OFDM in VLC when using the sa-
Chvojka, P., Haigh, P., Zvanovec, S., Pesek, P. and Ghassemlooy, Z.
Evaluation of Multi-band Carrier-less Amplitude and Phase Modulation Performance for VLC under Various Pulse Shaping Filter Parameters.
DOI: 10.5220/0005956900250031
In Proceedings of the 13th International Joint Conference on e-Business and Telecommunications (ICETE 2016) - Volume 3: OPTICS, pages 25-31
ISBN: 978-989-758-196-0
Copyright
c
2016 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
25
Figure 1: Schematic diagram of the VLC system. Note, that ‘UP’ and ‘DOWN’ block represent upsampling and
downsampling, respectively.
me experimental setup. The improvement in achieved
transmission speed was ~ 22%, which is significant
(Wu et al., 2013). Nevertheless, CAP requires a flat
channel frequency response, which is a rare
commodity in VLC networks, due to the LEDs LPF
behaviour. To overcome this, a new approach called
multi-band CAP (m-CAP) was proposed for short
range optical fibre links (Olmedo et al., 2014). The
available bandwidth was split into 6 sub-bands (or
subcarriers for compatibility with OFDM
nomenclature) and a 6-CAP system was compared to
the traditional single CAP (1-CAP) system. Splitting
the bandwidth into m sub-bands has two key
advantages over 1-CAP: i) relaxing the flat frequency
response requirement and ii) allowing to adjust
number of bits-per-symbol for each sub-band. Only a
slight improvement in data rate (100 Gb/s and
102.4 Gb/s for 1-CAP and 6-CAP, respectively) was
shown in (Olmedo et al., 2014), however m-CAP
outperformed the conventional 1-CAP system in
bandwidth efficiency and dispersion.
Thus, we adopted m-CAP modulation for VLC
systems in our previous work both experimentally
(Haigh et al., 2015; Haigh et al., 2015) and
theoretically using numerical simulations (Haigh et
al., 2015; Werfli et al., 2015; Haigh et al., 2015). For
instance, in (Haigh et al., 2015) we showed that for a
higher number of sub-bands, a higher transmission
capacity can be supported. The highest data rate we
achieved was ~ 31.5 Mb/s in the 10-CAP system,
using an LED with a low 4.5 MHz modulation
bandwidth. Increasing the number of sub-bands
results in a lower bandwidth occupied by each
subcarrier. Thus, they are less prone to the frequency
dependent attenuation caused by the first order LPF
behaviour of an LED and hence can support higher
throughput, as mentioned. A highly bandlimited VLC
link was investigated in (Haigh et al., 2015). The LPF
cut-off frequency was set to 0.1 of the signal
bandwidth and it was shown, that 10-CAP system can
support up to 40% improvement in the bit rate
compared to the traditional 1-CAP for the same bit
error rate (BER) target.
As mentioned before, the carrier frequencies are
generated by FIR filters, which are crucial in
determining the system performance and complexity.
Thus, in this paper we investigate performance of the
m-CAP system using different filter parameters such
as filter length L
f
and roll-off factor β through
numerical simulations. We show that a 2-CAP system
does not meet the BER target for any filter length,
while usage of filter lengths in the range of
8 < L
f
< 12 is sufficient for the most of the subcarriers
except two highest for m = 5 and 10. Moreover, just a
slight increment of the roll-off factor β results in
significant improvement in BER performance when a
higher number of sub-bands is deployed. The rest of
the paper is organized as follows: in Section 2 the
main principles of m-CAP and filter parameters are
discussed. In Section 3 and 4 the results from m-CAP
filter analyses are given and the conclusions are
drawn, respectively.
2 SYSTEM DESCRIPTION
2.1 Multi-band CAP Principle
The schematic block diagram of the tested system is
illustrated in Fig. 1. Firstly, m independent data
streams of the length of 10
6
are generated. The data is
mapped into M-QAM constellation symbols,
upsampled by means of zero padding and split into
the real (in-phase) and imaginary (quadrature)
branches. The order of QAM modulation M is set to
16 to stay in consistency with recent literature
(Haigh et al., 2015). The upsampling factor is given
as (Olmedo et al., 2014):
2
1
(1)
where
∙
is the ceiling function and is the roll-off
factor of a square root raised cosine (SRRC) filter.
The real and imaginary parts of the signal are then
passed through the in-phase and quadrature SRRC
filters, respectively. The setting of FIR filters will be
discussed in detail later with respect to particular
parameters under investigation. Finally, the filter
outputs are summed up and modulate the LED
intensity. The output signal is described by (Haigh et
al., 2015; Junwen et al., 2013):
OPTICS 2016 - International Conference on Optical Communication Systems
26
⊗
⊗
(2)
where
and
are the real and imaginary QAM
symbols for the
subcarrier and
and
are the
in-phase and quadrature transmit filter impulse
responses, respectively, and ⊗ denotes time-domain
convolution.
The output signal is passed through an ideal
analogue LPF channel and the signal bandwidth is set
to B = 1 Hz without any loss of generality. Thus, we
can expect the same results for any signal bandwidth.
The cut-off frequency of the LED is set as f
c
= 0.5 Hz
as in (Haigh et al., 2015), which results in an out of
band transmission, where the signal attenuation of
20 dB/decade is assumed. After passing the
zero-mean additive white Gaussian noise (AWGN)
channel, the signal is passed through time reversed
real and imaginary receiver filters, which are matched
to the transmit filters. Next, both the real and
imaginary parts of the signal are downsampled
according to n
s
and de-mapped for QAM
constellation symbols estimation. Finally, the
transmitted bits are compared with the received bits
for BER estimation.
As in our previous work we set the BER target
limit to 10
-4
(Haigh et al., 2015), which is below the
International Telecommunication Union’s (ITU’s)
recommendation error floor (3.8×10
-3
) for forward
error correction (FEC) with an overhead of 7%.
2.2 FIR Filters
The FIR filters at both the transmitter and receiver are
critical for the final m-CAP VLC system
performance, since their number is scaled by a factor
of 2m (two filters for each sub-band placed at both
transmitter and receiver). Thus, one must consider the
trade-off between performance and system
complexity when implementing m-CAP in e.g. field
programmable gate array (FPGA). However, to the
best of author’s knowledge filter parameters for
m-CAP VLC systems have never been tested. There
is a strong requirement to test such filter performance,
as the computational complexity of the FIR filters
increases with 2L
f
/symbol.
The pulse shaping SRRC filters have two key
parameters: i) roll-off factor β and ii) filter length L
f
.
The value of β varies in the range of 0 ≤ β ≤ 1 and
determines the excess of bandwidth. A larger β results
in the greater bandwidth requirements (see Fig. 2(a)),
which is (1 + β) times the symbol rate. Clearly, using
β = 0 means ideal spectrum utilization. The impulse
responses of the SRRC filters are orthogonal in the
time domain with a 90° phase shift forming a Hilbert
pair. They are generated as the product of a cosine and
sine wave with SRRC filter impulse response for the
real and imaginary parts of the signal, respectively.
Obviously, the cosine/sine wave frequency gives the
carrier frequency of each sub-band. The impulse
responses of the transmit filters are given by (Haigh
et al., 2015; Junwen et al., 2013):
sin
1
4
cos
14
⋅cos
21
(3)
for the in-phase filter and
sin
1
4
cos
14
⋅sin
21
(4)
for the quadrature filter, where T
s
is the symbol
duration, γ = πt/T
s
and δ = 1+β.
Figure 2: (a) shows frequency responses of the SRRC filters
for m = 2. The impulse responses for each sub-band for the
real and imaginary part of the signal are in Fig. 2(b) and
2(c), respectively. Note the excess bandwidth for β = 0.15.
Fig. 2(a) illustrates filters’ frequency responses
for 2-CAP system for each sub-band, the time domain
equivalents for the real and imaginary signal are
depicted in their respective colours in Fig. 2(b) and
2(c), respectively.
3 RESULTS
In this paper, we investigate the m-CAP VLC system
Evaluation of Multi-band Carrier-less Amplitude and Phase Modulation Performance for VLC under Various Pulse Shaping Filter
Parameters
27
performance under various filters parameters for the
first time. The order of QAM modulation format is set
to M = 16, while the number of sub-bands is
m = {2, 5, 10}. As mentioned before, the signal
bandwidth B and the LED 3-dB modulation
bandwidth f
c
are 1 Hz and 0.5 Hz, respectively.
We adopted two different approaches and set the
parameters to stay in consistency with recent
literature (Olmedo et al., 2014; Haigh et al., 2015;
Werfli et al., 2015):
Fixing the roll-off factor β = 0.15 and varying the
filters length in the range of
2 ≤ L
f
≤ 16 symbols
Fixing the filter length L
f
= 10 symbols and
varying the roll-off factor in the range of
0 ≤ β ≤ 1
3.1 Fixed Roll-off Factor
The number of filter taps is crucial for the system
performance, since the implementation of the
m-CAP modulation on a digital signal processor
(DSP) such as an FPGA is limited by the available
hardware resources. Dividing the signal bandwidth
into m = 10 sub-bands requires 20 FIR filters at the
transmitter part and a further 20 at the receiver.
Thus, we investigate the BER performance of the
system (see Fig. 1) for a range of L
f
. The results are
illustrated in Fig. 3(a)-(c) for E
b
/N
0
= 10 dB (solid
line, unfilled markers) and 20 dB (dashed line, filled
markers) for each subcarrier (denoted as s) while β is
fixed to 0.15. It can be seen that increasing L
f
for low
values of E
b
/N
0
improves the BER performance by
one order of magnitude in the best case for m = 5 and
10, while for m = 2 the improvement in BER is almost
negligible. With higher order subcarriers such a
performance enhancement due to the increased filter
lengths is considerably reduced. This is caused by the
attenuation outside of the modulation bandwidth.
On the other hand, when higher E
b
/N
0
is employed,
the BER target is easily met for low order subcarriers
for m = 5 (s = 1, 2, 3) and 10 (s = 1, 2, …, 8). However,
note: the 2-CAP system completely fails to meet the
BER limit in the whole considered range of L
f
due to
a wider bandwidth requirement, which is five times
higher that of 10-CAP. The performance
improvement is further illustrated in Fig. 3(d), where
Figure 3: BER performance for a range of L
f
for different values of m is shown in (a)-(c). Two different values of
E
b
/N
0
= 10 and 20 dB were considered. The performance improvement is most significant for low order subcarriers for
m = 5 and 10. When employing m = 2, BER target is never met. Received constellation diagrams are shown in Fig. 3(d) for
each m for the subcarrier s = 1 and two filter lengths L
f
= 4 and 10 at E
b
/N
0
= 20 dB.
OPTICS 2016 - International Conference on Optical Communication Systems
28
constellation diagrams for the subcarrier s = 1 for
each value of m and filter lengths L
f
= 4 and 10 are
illustrated.
Figure 4: The measured SNR for each subcarrier for a range
of m and L
f
= 4 and 10 at E
b
/N
0
= 20 dB. Increasing the
number of subcarriers increases the differences between
measured SNR when using short (4 symbols) and long (10
symbols) filter.
The highest value of L
f
to attain the BER target is
12, which is reflected in Fig. 3(b) and 3(c) for s = 3
and 8, respectively. This introduces an important
result: utilization of longer filters (i.e., L
f
> 12) for
higher order subcarriers is impractical since no
additional BER improvement is achieved. Fig. 4
illustrates the measured signal-to-noise (SNR) ratio
for the range of m = {2, 5, 10} for L
f
= 4 (solid line)
and 10 (dashed line) at E
b
/N
0
= 20 dB. By utilization
of higher order m-CAP system we can experience
increasing SNR gain when compared short and long
filter. The SNR gain for the last subcarrier is then
~ 1.2 dB, ~ 3.3 dB and ~ 4.1 dB for m = 2, 5 and 10,
respectively.
3.2 Fixed Filter Length
The roll-off factor β determines the excess of
bandwidth, as mentioned. Thus, using lower values of
β is appropriate to save the bandwidth employed in
the system. In contrast, it is also possible to increase
the system BER performance at the cost of the
bandwidth.
Fig. 5(a)-(c) illustrate the BER performance of the
system for a range of β for all values of m under
E
b
/N
0
of 10 dB (solid line, unfilled markers) and
20 dB (dashed line, filled markers) and for L
f
= 10.
Figure 5: BER performance for a range of β for each m is shown in (a)-(c). Two different values of E
b
/N
0
= 10 and 20 dB
were considered. Only slight increment of β improves the BER performance for low order subcarriers for m = 5 and 10. When
employing m = 2, BER target is met for β = 0.3 and 0.8 for the subcarriers s = 1 and s = 2, respectively. Received constellation
diagrams are shown in Fig. 5(d) for each m for the subcarrier s = 1 and two values of β = 0.1 and 0.5 at E
b
/N
0
= 20 dB showing
improved performance while increasing β.
Evaluation of Multi-band Carrier-less Amplitude and Phase Modulation Performance for VLC under Various Pulse Shaping Filter
Parameters
29
The results show that the performance
enhancement of each subcarrier is negligible with
increasing β at E
b
/N
0
= 10 dB (maximum one order of
magnitude for low order subcarriers for m = 5 and 10).
Surprisingly, setting β > 0.2 does not bring any
additional improvement and BER curves have almost
constant character for m = 5 and 10.
On the other hand, only a slight increment of β
improves the BER performance significantly for
lower order subcarriers at higher E
b
/N
0
(i.e., 20 dB).
The BER target is achieved by almost all subcarriers,
except the last subcarriers, i.e., s = 4, 5 and for m = 5
and s = 8, 9 for m = 10, respectively, for β = 0.2. This
is remarkable improvement when compared to E
b
/N
0
of 10 dB even in the cost of a slightly increased
bandwidth excess. Moreover, both subcarriers in the
2-CAP system attain to meet the BER target when
using β = 0.3 and 0.8 for s = 1 and 2, respectively.
This is in contrast with Fig. 3(a) where both
subcarriers failed to achieve the BER target even for
longer FIR filters. Fig. 5(d) illustrates received
constellation diagrams for s = 1 for each value of
m = {2, 5, 10} and β = 0.1 and 0.5 showing the
performance improvement while increasing β.
The measured SNR for each subcarrier for the
range of m for β = 0.1 (dashed line) and 0.5 (solid
line) at E
b
/N
0
= 20 dB is illustrated in Fig. 6. It is clear
that the SNR gain for β = 0.5 is decreased when the
higher order subcarriers are deployed. For instance,
the SNR gain for the last subcarrier is ~ 4.5 dB,
~ 3.5 dB and ~ 2 dB for m = 2, 5 and 10, respectively,
which is in contrast with previous measurements in
Fig. 4. This is caused by the higher bandwidth
requirements for β = 0.5, which results in larger
attenuation acting on the individual subcarriers.
Figure 6: The measured SNR for each subcarrier for a range
of m and β = 0.1 and 0.5 at E
b
/N
0
= 20 dB. Increasing the
number of sub-bands reduces the SNR gain for higher order
subcarriers when employing a higher roll-off factor β.
4 CONCLUSIONS
In this paper we investigated the m-CAP performance
under varying filter parameters such as filter length
and roll-off factor. Increasing both parameters
considerably improves the BER performance of the
system. The main result for this analyses towards real
implementation is that deploying L
f
> 12 symbols is
impractical, since no more performance improvement
can be achieved for m = 5 and 10. The 2-CAP system
utilization is not recommended since it failed to attain
BER target for any value of L
f
. The roll-off factor β
has significant impact on the system performance
especially at higher E
b
/N
0
. The BER target was
achieved for the range of values of β for all the
subcarriers in the tested system except the highest
subcarrier of m = 10.
For future work, we will implement m-CAP
system in FPGA for further investigation of filters’
impact on the system performance for the cases of
limited hardware resources.
ACKNOWLEDGEMENTS
This work was supported by CTU grant
SGS14/190/OHK3/3T/13.
REFERENCES
Zvanovec, S., Chvojka, P., Haigh, P.A., Ghasemlooy, Z.,
Visible light communications towards 5G,
Radioengineering, vol. 24, April 2015.
Armstrong, J.; Sekercioglu, Y.; Neild, A., Visible light
positioning: a roadmap for international
standardization, in Communications Magazine, IEEE,
vol.51, no.12, pp.68-73, December 2013.
Haigh, P. A., Ghassemlooy, Z., Rajbhandari, S.,
Papakonstantinou, I., Popoola W., Visible Light
Communications: 170 Mb/s Using an Artificial Neural
Network Equalizer in a Low Bandwidth White Light
Configuration, Lightwave Technology, Journal of, vol.
32, pp. 1807-1813, 2014.
Bykhovsky, D., Arnon S., An Experimental Comparison of
Different Bit-and-Power-Allocation Algorithms for
DCO-OFDM, Lightwave Technology, Journal of, vol.,
vol. 32, pp. 1559-1564, Apr 15 2014.
Mesleh, R.; Elgala, H.; Haas, H., LED nonlinearity
mitigation techniques in optical wireless OFDM
communication systems, in Optical Communications
and Networking, IEEE/OSA Journal of , vol.4, no.11,
pp.865-875, Nov. 2012.
Wu, F.M., Lin, C.T., Wei. C.C., Chen, C.W., Huang, H.T.,
Ho, C.H., 1.1-Gb/s White-LED-Based Visible Light
Communication Employing Carrier-Less Amplitude
OPTICS 2016 - International Conference on Optical Communication Systems
30
and Phase Modulation, in Photonics Technology
Letters, IEEE , vol.24, no.19, pp.1730-1732, Oct.1,
2012.
Wei, J.L., Ingham, J.D., Cunningham, D.G., Penty, R.V.,
White, I.H., Performance and Power Dissipation
Comparisons Between 28 Gb/s NRZ, PAM, CAP and
Optical OFDM Systems for Data Communication
Applications, in Lightwave Technology, Journal of ,
vol.30, no.20, pp.3273-3280, Oct.15, 2012.
Wu, F.M., Lin, C.T., Wei, C.C., Chen, C.W., Chen, Z.Y.,
Huang, H.T., Sien Chi, Performance Comparison of
OFDM Signal and CAP Signal Over High Capacity
RGB-LED-Based WDM Visible Light
Communication, in Photonics Journal, IEEE , vol.5,
no.4, pp.7901507-7901507, Aug. 2013.
Olmedo, M.I., Tianjian, Z., Jensen, J.B., Qiwen,
Z.,Xiaogeng X., Popov, S., Monroy, I.T., Multiband
Carrierless Amplitude Phase Modulation for High
Capacity Optical Data Links, in Lightwave Technology,
Journal of , vol.32, no.4, pp.798-804, Feb.15, 2014.
Haigh, P.A., Burton, A., Werfli, K., Le Minh, H., Bentley,
E., Chvojka, P., Popoola, W.O., Papakonstantinou, I.,
Zvanovec, S., A Multi-CAP Visible-Light
Communications System With 4.85-b/s/Hz Spectral
Efficiency, in Selected Areas in Communications, IEEE
Journal on , vol.33, no.9, pp.1771-1779, Sept. 2015.
Haigh, P.A., Chvojka, P., Zvanovec, S., Ghassemlooy, Z.,
Son T.L., Kanesan, T., Giacoumidis, E., Doran, N.J.,
Papakonstantinou, I., Darwazeh, I., "Experimental
verification of visible light communications based on
multi-band CAP modulation," in Optical Fiber
Communications Conference and Exhibition (OFC),
2015 , vol., no., pp.1-3, 22-26 March 2015.
Haigh, P.A., Thai Le. S., Zvanovec, S., Ghassemlooy, Z.,
Pengfei L., Tongyang X., Chvojka, P., Kanesan, T.,
Giacoumidis, E., Canyelles-Pericas, P., Minh H.L.,
Popoola, W., Rajbhandari, S., Papakonstantinou, I.,
Darwazeh, I., Multi-band carrier-less amplitude and
phase modulation for bandlimited visible light
communications systems, in Wireless Communications,
IEEE , vol.22, no.2, pp.46-53, April 2015.
Werfli, K., Haigh, P.A., Ghassemlooy, Z., Chvojka, P.,
Zvanovec, S., Rajbhandari, S., Long, S., Multi-band
carrier-less amplitude and phase modulation with
decision feedback equalization for bandlimited VLC
systems, in Optical Wireless Communications (IWOW),
2015 4th International Workshop on , vol., no., pp.6-
10, 7-8 Sept. 2015.
Haigh, P.A., Aguado, A., Ghassemlooy, Z., Chvojka, P.,
Werfli, K., Zvanovec, S., Ertunc, E., Kanesan, T.,
Multi-band carrier-less amplitude and phase
modulation for highly bandlimited visible light
communications — Invited paper, in Wireless
Communications & Signal Processing (WCSP), 2015
International Conference on , vol., no., pp.1-5, 15-17
Oct. 2015.
Junwen Z., Jianjun Y., Fan L., Nan C., Ze D., Xinying L.,
11 × 5 × 9.3Gb/s WDM-CAP-PON based on optical
single-side band multi-level multi-band carrier-less
amplitude and phase modulation with direct detection,
Opt. Express 21, 18842-18848 (2013).
Evaluation of Multi-band Carrier-less Amplitude and Phase Modulation Performance for VLC under Various Pulse Shaping Filter
Parameters
31
