The IEEE 802.11n wireless LAN for real-time industrial

advertisement
The IEEE 802.11n wireless LAN for real-time
industrial communication
Federico Tramarin and Stefano Vitturi
Michele Luvisotto and Andrea Zanella
Institute of Electronics, Computer and Telecommunication Engineering
National Research Council of Italy, CNR–IEIIT
Via Gradenigo 6/B, 35131 Padova, Italy
{tramarin, vitturi}@dei.unipd.it
Department of Information Engineering
University of Padova
Via Gradenigo 6/B, 35131 Padova, Italy
{michele.luvisotto, zanella}@dei.unipd.it
Abstract—In the last years, IEEE 802.11 Wireless LANs
(WLANs) have proved their effectiveness for a wide range of
real–time industrial communication applications. Nonetheless,
the enhancements at the PHY and MAC layers introduced by
the IEEE 802.11n amendment have not yet been adequately
addressed in the context of industrial communication. In this
paper we investigate the impact of some IEEE 802.11n new
features on some important performance figures for industrial
applications, such as timeliness and reliability.
I. Introduction
The IEEE 802.11 wireless LANs (WLANs) [1] represent an
interesting opportunity for real–time industrial communication
since, besides the known advantages of wireless networks [2],
they can provide satisfactory performance for a wide range of
applications. In particular, the IEEE 802.11n standards have
proved to be an effective solution to the communication problems typical of Industrial Wireless Networks (IWNs) where
tight constraints in terms of both timeliness and reliability are
often encountered [3].
The IEEE 802.11, actually, is a family of progressively
defined standards. In 2009 the IEEE 802.11n amendment
was released, providing several improvements to the previous
versions. In particular, it supports Multiple Input Multiple
Output (MIMO) features, which allow for increased reliability,
longer communication distances and higher transmission rates,
while maintaining operations in the unlicensed 2.4 GHz and
5 GHz Instrumentation, Scientific and Medical (ISM) bands.
Today, IEEE 802.11n networks are widely deployed in
general purpose home/office communication systems and, indeed, several off–the–shelf available devices are equipped with
IEEE 802.11n compliant interfaces. Conversely, this is not
the case for industrial environments where these networks
are still rarely deployed. Also, in the scientific literature, the
introduction of IEEE 802.11n in the industrial communication
scenario has not been deeply addressed and only a few
contributions explicitly relevant to this standard are available
[4] and [5].
Thus, in this paper, we start from an analysis of the most
promising features of IEEE 802.11n and determine which
could be useful in a real–time industrial context. Then, we
present the results of some experimental sessions, which
provide useful insights about the proper network configuration
for industrial environments.
978-1-4799-8244-8/15/$31.00 ©2015 IEEE
II. Overview of IEEE 802.11n
The IEEE 802.11n standard provides new and interesting
features at both the physical and data link layers.
1) IEEE 802.11n PHY layer: the introduced enhancements
required, basically, a substantial re-design of the whole layer.
a) Modulation and coding schemes (MCS): The set of
available modulations has been slightly modified with respect
to older versions of the standard, achieving an 11% increase
in raw transmission rate, and also the number of subcarriers
of the Orthogonal Frequency-Division Multiplexing (OFDM)
modulation for 20 MHz channels has been increased from
48 to 52, yielding a further 8% rate improvement. Moreover,
IEEE 802.11n makes available 40 MHz transmission channels,
in both the 2.4 GHz and the 5 GHz frequency bands, as an
alternative to basic 20 MHz ones, roughly allowing to double
the transmission rate, reaching 130 Mbit/s. Finally, two other
appealing features are worth mentioning, namely the reduction
of the guard interval (GI) between two consecutive OFDM
symbols from 800 ns to 400 ns (which raises the transmission
rate to 150 Mbit/s), and the possibility of replacing classic
convolutional codes with the more robust low–density parity–
check (LDPC) ones.
b) MIMO capabilities: the baseline scheme for the exploitation of a multi–antenna MIMO system is represented
by Spatial Division Multiplexing (SDM), which consists in
subdividing the payload in independent data streams, each
assigned to a different transmit antenna. In this case the raw
transmission rate of the system at PHY layer increases with the
number of independent data streams. The amendment defines
at most a 4×4 system, with 4 transmitting and 4 receiving
antennas, to reach the raw transmission speed of 600 Mbps.
In addition, multiple antennas could alternatively be used
to send redundant information with the aim of increasing
communication reliability. To this purpose, a useful technique
is Space–Time Block Coding (STBC), according to which
consecutive OFDM symbols are opportunely encoded in time
and sent over different antennas in order to maximize the
decoding probability at the receiver, at the expense of the
bitrate.
The various options made available by the IEEE 802.11n
PHY, are simply denoted as MCSs, where MCS 0 to MCS 7
refere to configurations with a single spatial stream, whereas
schemes from MCS 8 to MCS 15 are used for two spatial
streams.
2) IEEE 802.11n MAC layer: the new MAC layer strongly
builds on the IEEE 802.11e foundatthe ones which could
actuallyions, which introduced the Quality of Service (QoS)
concept and defined the possibility for a station to obtain a
transmit opportunity (TXOP) period during which it can send
multiple consecutive frames avoiding contention and backoff
procedures. IEEE 802.11n enhances this feature allowing to
aggregate more frames into a single data unit to be transmitted
during a TXOP, thus reducing the overhead due to interframe
spaces and headers.
In addition, the Block ACK (BA) mechanism allows the
receiver to acknowledge the transmission of multiple data units
with a single frame, further improving channel utilization. As
a difference from the original IEEE 802.11e procedure, in
the new standard the BA is implicitly sent by the receiver
in response to an aggregated frame.
III. IEEE 802.11n for Industrial Wireless Networks
In this section we focus on the aforementioned features and
investigate those that can potentially bring improvements to the
performance of industrial communication systems. Before going into details, it is worth recalling that the typical real–time
industrial traffic consists in data exchanged by controllers and
field devices at low levels of factory automation. Both cyclic
and acyclic transmissions take place and are characterized by
limited payloads, critical timing and reliability constraints. As
an example, cyclic operations could require the transmission
of a few bytes with periods as low as some hundreds of
microseconds and very low jitter.
A. Packet transmission time
In the assessment of the IEEE 802.11n performance, one of
the principal components is the transmission time of a packet
at the PHY layer. According to the protocol specification, it
can be derived as
&
'
l · 8 + L MH + LPH
T T X (l) = T preamble +T S IG +T S Y M ·
+T S E (1)
NDBPS
where T preamble and T S IG are the durations of PHY preamble
and SIGNAL, whereas T S Y M is the duration of an OFDM
symbol. The ceiling function defines the number of transmitted
symbols: indeed, the term l is the number of bytes contained
in the payload, whereas the terms L MH and LPH account for
the length of the MAC and PHY layers headers, respectively.
NDBPS is the number of data bits contained in an OFDM
symbol and, finally, T S E assumes the fixed value of 6 µsfor
transmissions in the 2.4 GHz band.
The time T preamble deserves particular attention since, for
backward compatibility purposes, a frame begins with an
IEEE 802.11a/g legacy preamble, followed by a new High–
Throughput (HT) preamble. This structure introduces a considerable overhead and can be replaced by the Greenfield (GF)
preamble, in which the legacy fields are removed, leading to
a fixed 12 µs reduction of the transmission time, at the cost of
losing compatibility with older devices.
Eq.1 makes it possible to determine the impact of PHY
layer features on packet transmission time for different payload
lengths and configuration options, which affect the NDBPS
parameter. The following considerations can be drawn:
•
•
•
•
The use of 40 MHz channels decreases the transmission
time even for short payloads. The benefits of using wider
channels obviously increase for greater payloads and are
more evident for lower MCSs.
The 2×2 configuration with 2 spatial streams generally
allows for a reduction of the transmission time with
respect to 1×1 configurations. However, while for lower
MCSs such a reduction takes place for any payload size,
this is not the case for higher MCSs. For example, with
MCS 7, a 2×2 configuration performs better only for
payloads greater than approximately 170 bytes.
Considering 20 MHz channels, IEEE 802.11g has lower
transmission times for small payloads than IEEE 802.11n
thanks to a lower preamble overhead. For example, with
MCS 0, IEEE 802.11n is slightly advantageous only for
payload greater than 224 bytes. This difference may be
partially mitigated by the use of the GF preamble.
Analogously, with MCS 0 and payloads shorter than 140
bytes, IEEE 802.11g is faster than IEEE 802.11n also
when using 40 MHz channels. In this case, however,
the use of the GF preamble completely removes this
difference.
B. Improving reliability
One of the most critical aspects in IWNs is the possibly high packet error rate (PER) that may yield several
retransmissions before successful packet delivery, with the
consequent introduction of random fluctuations of the service
time. The support for multi–antenna systems introduced by
IEEE 802.11n, however, can be profitably exploited to mitigate
this issue.
In this framework, additional antennas can be used to
improve reliability through STBC techniques. Several coding
schemes have been proposed to this regard, the most popular being the Alamouti method [6]. A detailed explanation
of how this strategy increases reliability in several system
configurations, including the 2×2 system studied in this paper,
is provided in [6].
C. MAC layer options
While the QoS–aware features of IEEE 802.11e have been
proven to be helpful in an industrial context [7], the effectiveness of both frame aggregation and Block ACK is
questionable. Indeed, these techniques perform at their best
when a station sends big chunks of data in a row. In most
industrial applications, instead, packets are transmitted singularly, either on a periodic basis or triggered by a specific
event. In these scenarios, hence, aggregation and BA may
lead to larger overheads [8], thus increasing the delay of the
whole data transmission, also because the aggregated data
units need to be completely received before being forwarded to
upper layers. Furthermore, the use of BA techniques requires
the initial exchange of some specific control frames between
two nodes, causing an additional overhead. Conversely, in
industrial communication networks, efforts are generally made
to avoid any frame exchange that is not strictly related to
data transfer (e.g. RTS/CTS), since they may reduce the
communication efficiency.
From the previous discussion, the following recommendations can be suggested for the profitable deployment of IEEE
802.11n WLANs in industrial communication systems.
For what concerns PHY layer, 40 MHz channels should
be preferred to traditional 20 MHz ones, since transmission
rate is doubled without compromising reliability. If backward
compatibility with older devices is not an issue, then GF
preamble should always be adopted, since it provides a gain
of 12 µs in frame transmission time. The use of short GI
is discouraged. Indeed, numerical simulations (not shown
here due to space limitations) revealed it does not reduce
significantly the packet transmission time. On the other hand,
short GI increases vulnerability to inter–symbol interference,
which may be a problem in industrial environments. The use
of LDPC codes is strongly recommended, since further simulations showed it has no impact on the packet transmission
time while providing increased reliability with respect to convolutional encoding. If a 2×2 system is available, the gain in
terms of packet transmission time obtained by using 2 spatial
streams with respect to the single–stream case is not very
significant, especially for small packets. As a consequence,
the best choice in an industrial context is to exploit additional
antennas to increase robustness through the use of STBC, as
experimentally confirmed by the results shown in next section
will strongly confirm this assertion.
As for MAC layer configurations, native implementations
of frame aggregation and Block ACK are not useful in a real–
time industrial context and therefore should be disabled.
IV. Performance assessment
In this section, we present some experimental measurements
that provide insights on the impact of 802.11n features on the
significant performance figures of a typical industrial network.
All measurements have been carried out in our research
laboratory. The test bench was composed by two desktop PCs,
equipped with wireless network interface cards (WNIC) by
TP-LINK that support operations with two antennas. The cards
are based on the Atheros AR9287 chip, fully compliant with
IEEE 802.11n, and both are ”SoftMAC” devices, i.e. they
enable a fine control of the transmission path by executing
most part of the MAC layer in software. Specifically, they
are managed by the open source ath9k Linux driver. The
WNICs were placed at a distance of roughly 2 meters from
each other, with a line–of–sight (LOS) path always available.
The network setup for the MAC and PHY layers followed,
as much as possible, the guidelines provided in Section III-D.
However these WNICs, did not support neither LDPC codes
nor the GF preamble.
PC #1
ath9k
WNIC
AR9287
PC #1
ath9k
WNIC
AR9287
PCI-E conn.
(a)
Agilent RF Gen
(b)
WNIC
AR9287
PC #2
ath9k
WNIC
AR9287
Agilent RF Gen
Fig. 1. Sketch of the adopted measurement setups.
The adopted experimental setup is shown in Fig. 1. In
order to emulate realistic industrial working conditions, we
introduced an artificial and controlled source of channel impairments, by using an Agilent RF generator. This allowed us
to inject AWGN noise centered on the carrier frequency of
the adopted ISM channel through a directional antenna, that
makes it possible to finely tune the SNR at the receiver side.
A. Evaluation of the Packet error rate
In order to evaluate the system PER, an application was
developed for the experimental setup of Fig. 1(a). It performs a periodic transmission of short UDP broadcast packets
(50 Bytes) from one PC to the other. The directional antenna
of the RF generator was oriented as to interfere only with the
receiving WNIC and the interference power was accurately
varied to scan a range of 35 different SNR values, with 1
dB steps. For each SNR value the application sent 1000 UDP
packets. Since we wished to assess the PER at the PHY layer,
we disabled MAC–layer retransmissions, thus eliminating any
kind of error recovery.
The results are reported in Fig. 2, where three different
configurations are considered, namely IEEE 802.11n 2×2 with
2 streams (gray dashed lines), 2×2 with 1 stream and STBC
enabled (gray solid lines), and 1×1 IEEE 802.11g (black lines).
0
10
PER
D. Recommendations for system configuration
−1
10
−10
g − BPSK, 1/2
n − BPSK, 1/2, STBC
n − BPSK, 1/2, no STBC
g − 64−QAM, 3/4
n − 64−QAM, 3/4, STBC
n − 64−QAM, 3/4, no STBC
−5
0
5
10
SNR [dB]
15
20
25
Fig. 2. PER for different system configurations (50 Byte payload)
At a first sight, the leftmost curves, which refer to a BPSK
modulation with code rate 1/2, show a gain in the order
of 2-3 dB of IEEE 802.11n over IEEE 802.11g. Also, the
introduction of STBC further improves PER. The rightmost
curves of Fig. 2 provide a comparison between the highest available transmission rate of IEEE 802.11g (54 Mbit/s)
and the equivalent 2×2 IEEE 802.11n configurations, namely
MCS 6 (2×2 one stream with STBC) and MCS 14 (2×2 two
streams). We selected these two latter MCSs to carry out a
fair comparison, since they use the same modulation as IEEE
802.11g at 54 Mbit/s (64-QAM with code rate 3/4). Looking
at the figure, the beneficial impact of STBC is evident.
Consequently, the adoption of STBC is recommended since it
ensures greater reliability without significantly affecting packet
transmission time, as it will be discussed in Section III-B.
B. Service time evaluation
In an industrial communication context, a very meaningful
performance index, related to several other metrics, is repre-
TABLE I
Assessment of service time (fixed transmission rate)
No Interference
Configuration
IEEE
IEEE
IEEE
IEEE
802.11g
802.11n
802.11g
802.11n
6 Mbit/s
MCS0
54 Mbit/s
MCS7
Interference
Mean
Std. Dev.
Mean
Std. Dev.
434.0 µs
391.1 µs
332.1 µs
344.1 µs
5.8 µs
13.4 µs
3.9 µs
8.9 µs
919.3 µs
489.5 µs
556.4 µs
428.4 µs
961.8 µs
212.6 µs
460.1 µs
184.9 µs
In the interfering scenario, IEEE 802.11n performs definitely better than IEEE 802.11g. This is clear also from the
Empirical Cumulative Distribution Functions (ECDFs) plot
shown in Fig. 3. Indeed, a significant increase of the success
probability at the first transmission attempt (from 60% to
nearly 80%) is recognizable when IEEE 802.11n is adopted.
This is explained by considering that the introduction of
STBC allows for a reduction of the PER, as observed in
Fig. 2. The increased reliability is also confirmed by the strong
reduction of the service time standard deviation highlighted in
Tab. I. Furthermore, the performance improvements are also
appreciable even in terms of average service time values that
are lower for both IEEE 802.11n configurations with respect
1
0.8
Empirical CDF
sented by the service time that is defined as the time required
to deliver a packet from the application point of view. Hence,
we carried out a measurement campaign to characterize this
performance index.
The measurement setup was modified as shown in Fig. 1(b),
with two independent WNICs installed on the same PC. Such
an arrangement allowed to increase the accuracy of service
time measurements, since the timestamps for the various
events can all be retrieved from the same internal Time
Stamp Counter (TSC) processor register, which clearly returns
inherently synchronized readings for both WNICs. In these
experiments, one of the two stations periodically sent a unicast
data frame at the data–link layer by means of a custom
software package, which continuously gathered timestamps
both at the transmitting and receiving sides. In each test, 10000
frames were delivered, with a payload of 50 Bytes and a period
of 5 ms.
To provide an exhaustive assessment of the service time,
measurements were carried out both with and without interference on the system. In the former case, noise power
was modulated with a stochastic process, characterized by
ON periods (in which the interference is present) alternated
with OFF periods with no interference. The duration of ON
periods was a uniform random variable ranging from 100 µs to
200 µs, while the duration of OFF periods had an exponential
distribution with 200 µs mean. Two system configurations were
compared, namely IEEE 802.11g and 2×2 IEEE 802.11n using
40 MHz channels and STBC, i.e. the optimal configuration
identified in Sec. III-D.
A first set of results is reported in Tab. I. The comparison
of MCS 0 for IEEE 802.11n and 6 Mbit/s for IEEE 802.11g
(both using a BPSK modulation with code rate 1/2), in the
case of no interference, shows a significant reduction of the
service time for IEEE 802.11n, as the average value decreases
from 434 to 391 µs, with only a slight increase of the standard
deviation.
0.6
0.4
0.2
0
0
IEEE 802.11n MCS0
IEEE 802.11n MCS7
IEEE 802.11g 6 Mbit/s
IEEE 802.11g 54 Mbit/s
500
1000
Service time [µs]
1500
Fig. 3. Empirical CDF of service time with interference and adopting both
MCS 7 and rate adaptation with 50 bytes packets.
to those obtained from IEEE 802.11g. As a final remark, the
ECDFs show that all packets are transmitted within 1500 µs for
IEEE 802.11n. Conversely, with an IEEE 802.11g system, in
the same environmental conditions, a significant percentage of
frames (almost 10%) require a longer time to be successfully
delivered.
V. Conclusions
In this paper, we analyzed the impact of the configuration
parameters of a multi–antenna IEEE 802.11n device in real–
time industrial communication was derived. A preliminary
performance comparison with respect to IEEE 802.11g was
experimentally assessed, for a prototype network implemented
in our research laboratory, where the conditions of a typical
industrial environment were emulated. In the next future, we
plan to extend this work by designing suitable rate control
algorithms, able to fully exploit the enhancements provided
by IEEE 802.11n.
References
[1] IEEE Standard for Information technology–Telecommunications and
information exchange between systems–Local and metropolitan area
networks–Specific requirements. Part 11: Wireless LAN Medium Access
Control (MAC) and Physical Layer (PHY) Specifications, IEEE Std.,
March 2012.
[2] T. Sauter, “The Three Generations of Field-Level Networks–Evolution
and Compatibility Issues,” Industrial Electronics, IEEE Transactions on,
vol. 57, no. 11, pp. 3585–3595, November 2010.
[3] S. Ding, P. Zhang, S. Yin, and E. Ding, “An Integrated Design Framework
of Fault-Tolerant Wireless Networked Control Systems for Industrial Automatic Control Applications,” Industrial Informatics, IEEE Transactions
on, vol. 9, no. 1, pp. 462–471, Feb 2013.
[4] S. Santonja-Climent, D. Todoli-Ferrandis, T. Albero-Albero, V. SemperePaya, J. Silvestre-Blanes, and J. Alcober, “Analysis of control and
multimedia real-time traffic over SIP and RTP on 802.11n wireless links
for utilities networks,” in Emerging Technologies and Factory Automation
(ETFA), 2010 IEEE Conference on, Sept 2010.
[5] Y. Jin and F. F. Dai, “Impact of Transceiver RFIC Impairments on
MIMO System Performance,” Industrial Electronics, IEEE Transactions
on, vol. 59, no. 1, pp. 538–549, Jan 2012.
[6] S. Alamouti, “A simple transmit diversity technique for wireless communications,” Selected Areas in Communications, IEEE Journal on, vol. 16,
no. 8, pp. 1451–1458, 1998.
[7] G. Cena, L. Seno, A. Valenzano, and C. Zunino, “On the Performance
of IEEE 802.11e Wireless Infrastructures for Soft-Real-Time Industrial
Applications,” Industrial Informatics, IEEE Transactions on, vol. 6, no. 3,
pp. 425–437, August 2010.
[8] A. Saif, M. Othman, S. Subramaniam, and N. AbdulHamid, “Impact of
aggregation headers on aggregating small MSDUs in 802.11n WLANs,”
in Computer Applications and Industrial Electronics (ICCAIE), 2010
International Conference on, Dec 2010, pp. 630–635.
Download