Content uploaded by Heikki Karvonen
Author content
All content in this area was uploaded by Heikki Karvonen on Oct 05, 2018
Content may be subject to copyright.
Performance Evaluation of Bluetooth Low Energy
Technology under Interference
Heikki Karvonen, Konstantin Mikhaylov,
Dinesh Acharya and Md. Moklesur Rahman
Centre for Wireless Communications, University of Oulu, Finland,
heikki.karvonen@oulu.fi
Abstract. This paper focuses on analytical and experimental performance eval-
uation of the Bluetooth Low Energy (BLE) technology. Studies have been con-
ducted in indoors case relevant to healthcare and medical scenarios. Performance
of the recently published BLE 5 coded technique is compared to BLE 4 which is
currently the most used technology in commercial wireless healthcare and medi-
cal devices. This new improved BLE version may continue fostering the success
of BLE use in those application scenarios as well as enable novel Internet of
Things (IoT) solutions for Smart Healthcare. The main goal of this work was to
evaluate the packet error rate (PER) performance of BLE under ZigBee interfer-
ence, since it is envisaged, that coexistence problems may arise with the further
growth of number of the different IoT devices deployed. In the paper we first
develop and analytical model to characterize the PER of BLE link with varying
distance to interfering nodes. Then we conduct a series of practical measurements
using the Nordic Semiconductor nRF52840 chipset which includes support for
new BLE 5 coded feature and compare the results of the experiments with that of
the analytical model. Our results show that ZigBee interference is very harmful
for BLE communication when operating at the same frequency band, i.e., assum-
ing worst-case scenario. The proposed model can be used to evaluate PER of
BLE in various interference scenarios to get insight of communication reliability
which is very important specifically for healthcare and medical applications.
Key words: wireless coexistence, BLE 5, ZigBee, packet error rate (PER), In-
ternet of Things, healthcare and medical applications.
1 Introduction
The use of wireless sensor devices has been continuously increasing during recent years
thanks to rising success of Internet of Things (IoT) applications. Sensor devices can be
used in many different scenarios, e.g., smart factories and homes, environmental mon-
itoring, autonomous traffic, medical and healthcare applications. Wireless body area
2 Heikki Karvonen et al.
networks (WBANs) are used in the context of smart healthcare applications, operating
in hospitals or homes, as well as for versatile sport and fitness activities. WBAN sensors
can be also connected to Internet, being a one specific IoT use case which is gaining an
increasing business interest [1], [2].
IoT applications require low-power wireless communication solutions since most of
the use cases needs a long lifetime for the sensor nodes without battery replacement, or
even using only the energy scavenged from the operation environment. There are vari-
ous low-power technologies that have been proposed for wireless sensor nodes. The
most well-known low-power technologies are Bluetooth Low Energy (BLE) [3] and
IEEE Std. 802.15.4 [4] (ZigBee [5]). Specifically for WBAN purposes has been defined
IEEE Std. 802.15.6 [6] and ETSI SmartBAN [7]. In [8] it was found that BLE stands
out as the most widely used in current commercially available products in healthcare
and medical applications. Above mentioned technologies operate in the industrial, sci-
entific and medical (ISM) 2.4 GHz band, which is available worldwide enabling in-
teroperability in different regions. IEEE Std. 802.15.6 defines solutions also to sub-
Gigahertz bands as well as for ultra wideband (UWB) up to 10.6 GHz. Today, the 2.4
GHz band is rapidly becoming congested due to the presence of several other wireless
technologies such as IEEE Std. 802.11 (Wi-Fi), and most recently, the upcoming unli-
censed LTE solutions (LTE-U) [9]. Therefore, the coexistence issues may arise as the
number of IoT devices operating at that band increases rapidly.
It is important to evaluate the wireless communication performance in the congested
scenario at 2.4 GHz ISM band especially for applications which require reliable com-
munication as is the case in many healthcare and medical scenarios. In [10] authors
studied analytically the packet error rate of BLE under interference of ZigBee, Wi-Fi
and BLE 5 in hospital scenario. Here will be used experimental measurements to verify
the analytical model findings in case of BLE under ZigBee interference. We conduct
our measurements and report the results not for BLE 4 only, but also for recently pub-
lished BLE 5 coded (S=8) mode.
The structure of the rest of the paper is as follows. Section 2 briefly describes the
specifics of the BLE technology. Analytical model for PER calculation is introduced
ion Section 3. Measurement devices are described in Section 4. Section 5 introduces
the analytical and experimental results. Conclusions are given in Section 6.
2 Features of the BLE technology
The low-power version of Bluetooth, BLE 4, has been in use since June 2010 and today
it can be found in almost every smartphone, tablet, and laptop in the market in addition
to a large set of other wireless devices. The most recent version, Bluetooth 5 was intro-
duced in December 2016 [11] with the first commercial development kits becoming
available in early 2017. The long-range and mesh features has made BLE 5 very suita-
ble for versatile IoT applications. The official announcement of BLE 5 states that the
increase in range is up to 4 times compared with BLE 4.2 [12]. In the rest of this section
we will focus on the most important changes of BLE 5 compared to BLE 4.2.
3 Heikki Karvonen et al.
3
The problem of improving the communications range and the maximum throughput
has been addressed in BLE 5 specification by introducing three new physical layer
(PHY) options. In addition to the 1 Mbit/s Gaussian frequency shift keying (GFSK) of
BLE 4 (named in Bluetooth v 5.0 core specification LE 1M), the BLE 5 specifies a 2
Mbit/s GFSK PHY (named LE 2M) for short range high-speed transmission and two
coded PHY (referred to as LE Coded) with payload coded at 500 kbit/s or 125 kbit/s.
The LE coded PHYs are modulated using GFSK at 1 Msym/s rate, but the payload data
are coded in two stages: first by forward error correction convolutional encoder and
then spread by the pattern mapper. In theory, this enables to improve the link budget of
a coded transmission by over 5 dB and 12 dB compared to LE 1M for LE coded at 500
kbit/s and 125 kbit/s, respectively. Note, that only the support of LE 1M PHY is man-
datory.
Another change introduced to improve the communication range is the increase of
the maximum transmit power of a BLE from 10 dBm (10 mW) to 20 dBm (100 mW).
Unfortunately, due to the transmit power restrictions imposed by the frequency regula-
tions, this higher transmit power does not provide any benefit for some regions (namely,
EU, Japan and Korea). The maximum link layer protocol data unit (PDU), increased in
BLE 4.2 from 39 to 257 octets, stayed at this level also in BLE 5. The problem of
coexistence of devices in the 2.4 GHz band has been addressed in BLE 5 by introducing
the special interface proving signaling and messaging mechanisms between collocated
Bluetooth and other mobile wireless standard radios.
In addition to these changes, the functionality of the broadcasting channels in BLE
5 has been substantially enriched by the introduced extended advertising feature. First,
the concept of the secondary advertising channels which are co-allocated with the BLE
data channels was introduced. The format of the advertising packets used in the sec-
ondary channels has been reworked enabling them to carry up to 255 octets of PDU
(compared with 37 octets allowed in the primary channels of BLE 4) and even to sup-
port fragmentation. Another interesting feature enabled in the BLE 5 is the periodic
advertisements. Hopping between the secondary channels in a predefined pseudo-ran-
dom sequence, a periodic advertiser broadcasts the packets, with PDU of up to 255
octets, at regular intervals of time ranging from 7.5 ms to almost 82 s. Importantly, a
scanner device may synchronize with one or even several non-overlapping (in time)
periodic advertisers and get the data from all of them. This equips BLE 5 with a more
efficient and reliable solution for data broadcast than the one possible with BLE 4. Note,
that the support of periodic advertisements and extended advertising features is op-
tional.
Importantly, the BLE 5 is backward-compatible with the earlier versions of BLE –
all the discussed features are optional and are not necessarily needed to be supported.
Nonetheless, as one can easily see, they can substantially increase the communications
range or throughput or enable new modes of operation. Due to this fact, in the marketing
materials of Bluetooth SIG [Error! Reference source not found.], the BLE 5 is
claimed to provide double bandwidth, up to four times higher range and up to 8 times
broadcasting capacity compared to BLE 4.2. However, it must be noted that the im-
proved data rate and communication range cannot be achieved at the same time since
they are provided by different PHY modes.
4 Heikki Karvonen et al.
3 Analytical model
Here will be introduced an analytical model that can be used to compute the PER of
BLE under interference. The developed model takes into account interference of mul-
tiple nodes my aggregating the signal power coming from them.
There are several path loss models (2.4 GHz) proposed for indoor environments. For
line-of-sight (LOS) scenarios the path loss equation is typically defined as
(d)
(1)
where n is the path loss exponent and dh0 is the reference distance at which the reference
path loss PL0 is measured. In [9], measurements were conducted and a path loss model
was specifically developed for a hospital indoor environment. Authors found in [15]
that for a LOS hospital room case n = 1.2, which we are using in our calculations in the
rest of the paper.
The signal to interference ratio (SIR) at the affected receiver under multiple radios in-
terference can be computed as [14]
(2)
where the desired signal power is PS and Pi is the power of the i:th interferer (in dB).
The distance between the desired signal Tx and Rx nodes is L, and di is the distance
from the i:th interferer to the affected receiver. is a coefficient that limits the in-
terfering power to the bandwidth occupied by the technology being interfered with. It
is defined in [15] as follows
if
if ,
(3)
where BI is the bandwidth of the interferer signal and BDS is the bandwidth of the target
node receiver filter. For this study the BLE is assumed to use GFSK modulation with
bandwidth 1 MHz, bit rate Rb = 1 Mbit/s, BT = 0.5 and modulation index h = 0.5. For
non-coherent demodulation, the symbol error rate (SER) is calculated as [14] [15]
,
(4)
where Es is the energy per symbol and N0 is the noise power spectral density per Hz.
Here we assume a worst-case scenario where full collision of interfering packets
and the useful packet occurs, therefore SER can be assumed to be same for each trans-
mitted symbol of the BLE packet. The PER for the affected BLE link can be calculated
as
,
(5)
where K is the length of the packet of the desired signal and ɛ is the SER that can be
calculated using Eq. 4.
5 Heikki Karvonen et al.
5
4 Measurement devices
In our measurements, one of the first commercial chipsets that support BLE 5.0, the
nRF52840 [16] from Nordic Semiconductor, was used. The nRF52840 is a system on
chip (SoC) integrating a multiprotocol 2.4 GHz transceiver with an ARM Cortex-M4F
based microcontroller. The chipset was programmed with S140 SoftDevice v6.0.0,
which is a precompiled and linked binary software implementing BLE protocol devel-
oped by Nordic Semiconductor.
In the experiments we have used two nRF52840 Preview DK development kits
shown in Figure 1. The firmware for them was developed in this work based on the
ATT_MTU Throughput Example of the nRF5 software development kit (SDK)
v15.0.0. One of the boards was programmed to act as an advertiser, and the other one
– as scanner. The BLE physical layer to be used by the boards (i.e., LE 1M, LE 2M or
LE Coded) can be selected during the startup using the control buttons.
Figure 1. nRF52840 Preview DK device used for measurements.
The methodology of our experiments was as follows. After placing the BLE boards
in the specified locations, the scanner board was connected to a computer via serial over
USB interface, configured to operate using the required PHY layer option, and forced
to continuously scan a single advertisement channel. Approximately every second the
scanner reported via serial interface the number of the received advertisements from
the advertiser board, as well as the received signal strength indicator (RSSI) and the
sequence identifier for the last advertisement it has received. Once the scanner board
was activated, the advertiser was powered up and its PHY layer was configured. The
advertiser started periodically sending the advertisements, each of which contained a
unique sequence number. At the end of the experiment the PER was calculated from
the total number of the packets received by the scanner and the sequence number of the
last received packet.
In order to introduce the ZigBee interference to BLE communication, we used in our
measurements the CWC-MOD-POW platforms (version two) [17] [18] illustrated in
Figure 2. These boards are built around Texas Instruments’ CC2650 multi-standard
system-on-chip [19].The core middleware is based on CWC CC2650 IEEE Std.
6 Heikki Karvonen et al.
802.15.4 proprietary driver and firmware, and TI CCS 7.4.0.00015 IDE. ZigBee nodes
are equipped with an external antenna, Taoglas FXP70 [20]. The nodes were configured
to start spamming the ZigBee packets with maximum possible payload without using
any form of listen before talk at the same channel where BLE devices operate immedi-
ately after power up. The time between two sequential packets (due to radio re-config-
uration and uploading of the new packet) was well below 1 ms. In order to ensure con-
tinuity of the interferences multiple ZigBee interferers was used in our experiments.
Figure 2. CWC-MOD-POW platforms
In our measurement the BLE boards were set at the same height (1m) so that anten-
nas were pointing each other creating a LOS link. Three interfering ZigBee nodes were
set around the BLE receiver, all at the same distance (Case1 = 4 m and Case2 = 6 m)
to BLE receiver antenna. Different BLE link lengths were used (4 - 11 m) and number
of transmitted and received packets was recorded for 10 minutes duration at each case
(resulting in at least 10 000 BLE packets being sent).
Measurement environment was a restaurant at the University of Oulu during a time
when it was there was not customers, i.e., it was empty. This environment appeared to
provide similar path loss as the hospital room LOS model introduced in [15] with path
loss exponent n = 1.2. Therefore, this was a good environment to obtain results that can
be applied also to hospital case. Spectrum sniffers were used to find out that there was
not interference from WiFi or Bluetooth at the same band were our measurements were
conducted.
7 Heikki Karvonen et al.
7
5 Results
The developed analytical model was implement to Matlab and measurements were con-
ducted to evaluate BLE PER under ZigBee interference. Table 2 shows the parameters
used in analytical and experimental performance evaluation of BLE 4. In addition,
measurement were done for BLE 5 coded (S = 8) mode with the same parameter set-
tings to find out the gain provided by forward error correction.
Table 1. Parameters for analytical and experimental performance evaluation
Parameter
Value
Number of interfering nodes
3
Distance to interferers
Case1 = 4m; Case2 = 6m
Target BLE link length
4 - 11 m
Frequency
2.480 GHz (BLE CH#39, ZigBee CH#26)
Transmit power, BLE
0 dBm
Transmit power, ZigBee
0 dBm
Path loss exponent, n
1.2
RSSI at 1 m, BLE
-15 dBm
ZigBee Tx to BLE Rx loss
-12 dBm
Payload length, BLE
12 octets
Payload length, ZigBee
116 octets
Packet rate BLE node
One packet every 50 ms
Packet rate ZigBee node
One packet every 5 ms
Data Rate (BLE)
1 Mbps
Data Rate (ZigBee)
250 kbps
Figure 3 shows PER results for the scenario where three ZigBee nodes are interfering
LOS BLE link which length was varied. ZigBee nodes were set at 4 meters distance
from BLE receiver to create LOS interference. From Figure 3 it can be observed that
the effect of interference becomes visible when the BLE 4 link distance is longer than
5 meters. PER increases rapidly when the BLE link distance is increased and reaches
its maximum value when BLE link length is 10 meters. After that point, almost all
packets are lost since interference is too strong in comparison to BLE 4 signal strength.
As a reference result, in case without interference, the PER of BLE link remained below
15 % for a link distance of 80 meters. Further it can be observed that the measurement
results match well with the analytical results of BLE 4. It must be noted that in analyt-
ical calculations we used -12 dBm loss for ZigBee to BLE receiver. The rationale for
this negative gain is different antenna types and their orientations used in the ZigBee
and BLE nodes, which are assumed to decrease the strength of experienced interference
at BLE receiver. BLE 5 coded case measurement results of Figure 3 shows that the
error correction enables to maintain low PER until the link distance increases to longer
than 9 meters, enabling 3 meters (50%) higher communication range. After that point
the PER increases rapidly also in the coded mode, i.e., the coding cannot correct the
errors created by interference.
8 Heikki Karvonen et al.
Figure 3. BLE PER under interference of three ZigBee nodes at 4 m distance.
Figure 4. BLE PER under interference of three ZigBee nodes at 6 m distance.
9 Heikki Karvonen et al.
9
Figure 4 shows the PER for the case were ZigBee nodes were set at 6 m distance
from BLE receiver. As expected, it can be observed that BLE 4 link can be longer in
this case before the PER starts to increase due to interference. Also in this case it can
be observed that the PER starts to increase rapidly when the BLE link length is in-
creased beyond 6 meters. This results verifies further that the analytical results are
matching well with the measurement results even there is bit more variation in the re-
sults in comparison to Figure 3 case. BLE 5 coded mode measurement results shows
similar behavior to that in Figure 3, the coding gain being quite modest (2 – 3m) in
terms of increased BLE link length.
6 Conclusions
This paper proposes an analytical model and reports the results of an experimental
packet error rate evaluation of BLE under ZigBee interference. Analytical results were
derived for the BLE 4 mode while measurement were conducted not only for BLE 4
but also for BLE 5 coded mode. Measurement results verified the analytical model re-
sults. Analytical model can be used to derive results for other scenarios and also for
other type of interferers.
Our results show that the worst-case interference is very harmful for BLE commu-
nication even when using the BLE 5 coded mode. Here the worst-case interference
means that the interferers are at the same channel than useful signal and full packet
collisions occur. In terms of BLE link distance, the error correction coding gain was
found to be only 2 to 3 meters, i.e., approximately one third of the used communication
ranges. Results highlight that it is very important to pay attention to different technol-
ogies coexistence since the amount of IoT devices is increasing rapidly creating inter-
ference to each other.
Resilience towards interference is especially important in applications which require
high reliability communication. Erroneous packet receptions will also decrease the en-
ergy efficiency which is highly important in IoT applications. Results of this paper
show that the BLE communication performance will decrease drastically if there are
interfering ZigBee nodes in a close vicinity (< 6m) at the same frequency channel. In
future studies were are going to evaluate different coexistence scenarios using analyti-
cal modeling and experimental measurements.
Acknowledgments
This research has been financially supported by Academy of Finland 6Genesis Flagship
(grant 318927).
10 Heikki Karvonen et al.
References
1. P. Lamkin, “Wearable Tech Market To Be Worth $34 Billion By 2020,”
https://www.forbes.com/sites/paullamkin/2016/02/17/wearable-tech-market-to-be-worth-34-
billion-by-2020/#4cbe76e83cb5 (2016)
2. Tractica, “Healthcare Wearable Device Shipments to Reach 98 Million Units Annually by
2021,” https://www.tractica.com/newsroom/press-releases/healthcare-wearable-device-ship-
ments-to-reach-98-million-units-annually-by-2021/ (2016)
3. R. Haydon, “Bluetooth Low Energy: The Developer’s Handbook”, Pearson Education Inc., New
Jersey, USA, 2013.
4. IEEE Standard for Low-Rate Wireless Networks, "IEEE Std 802.15.4-2015 (Revision of IEEE
Std 802.15.4-2011)” (2016)
5. ZigBee Alliance, http://www.zigbee.org/.
6. IEEE Std. 802.15.6: IEEE Standard for Local and metropolitan area networks—Part 15.6: Wire-
less Body Area Networks. Standard, The Institute of Electrical and Electronics Engineers, Inc.
(2012)
7. M. Hämäläinen et al., "ETSI TC SmartBAN: Overview of the wireless body area network stand-
ard," International Symposium on Medical Information and Communication Technology
(ISMICT) (2015)
8. H. Karvonen, M. Hämäläinen, J. Iinatti and C. Pomalaza-Ráez, “Coexistence of Wireless Tech-
nologies in Medical Scenarios,” European Conference on Networks and Communications
(EUCNC), Oulu, Finland (2017)
9. Nokia white paper, “LTE evolution for IoT connectivity,” (2017)
10. H. Karvonen, C. Pomalaza-Ráez, K. Mikhaylov, M. Hämäläinen and J. Iinatti, “Interference of
Wireless Technologies on BLE Based WBANs in Hospital Scenarios,” IEEE International Sym-
posium on Personal, Indoor and Mobile Radio Communications (PIMRC), Montreal, Canada
(2017)
11. Bluetooth SIG, “Bluetooth Core Specification v 5.0” https://www.bluetooth.com/specifica-
tions/bluetooth-core-specification (2016)
12. Bluetooth SIG, “Bluetooth Core Specification 5.0 FAQ” https://www.bluetooth.com/~/me-
dia/files/specification/bluetooth-5-faq.ashx?la=en (2016)
13. R. de Francisco, “Indoor Channel Measurements and Models at 2.4 GHz in a Hospital,” IEEE
Global Telecommunications Conference (GLOBECOM), Miami, FL, USA, 2010.
14. H. Karvonen, K. Mikhaylov, M. Hämäläinen, J. Iinatti and C. Pomalaza-Ráez, "Interference of
wireless technologies on BLE based WBANs in hospital scenarios," International Symposium
on Personal, Indoor, and Mobile Radio Communications (PIMRC), Montreal, QC, 2017.
15. R. Natarajan, P. Zand and M. Nabi, “Analysis of Coexistence between IEEE 802.15.4, BLE and
IEEE 802.11 in the 2.4 GHz ISM band,” 42th Conference of the IEEE Industrial Electronics
Society (IECON 2016), Florence, Italy, 2016, pp. 6025-6032.
16. Nordic nRF52840, https://www.nordicsemi.com/eng/Products/nRF52840.
17. K.Mikhaylov, “Plug and play reconfigurable solutions for heterogeneous IoT”, Ph.D. thesis,
University of Oulu, 2018, online http://jultika.oulu.fi/Record/isbn978-952-62-1841-0
18. K. Mikhaylov and J. Petäjäjärvi, "Design and Implementation of the Plug&Play enabled Flexi-
ble Modular Wireless Sensor and Actuator Network Platform", Asian J. Control, Vol. 19, Issue
5, pp. 1-21, Sept. 2017.
19. Texas Instruments, CC2650, online http://www.ti.com/product/cc2650
20. Taoglas, FXP70 data sheet, online https://eu.mouser.com/datasheet/2/398/FXP70.07.0053A-
1219667.pdf
