2. 2 26 November 2015
To be connected or not to be at all ?
Building & Home Automation
Industrial Automation
Fire & Security
Metering
3. 3 26 November 2015
M2M
IoT
Video
Cloud
Mobile
Apps
BlueTooth
Wireless – wireless – wireless
10 100 1k 10k 100k 1M 10M
(Bytes / Day)
4G3G2.5G 2.75GSigFox UNB
WmBUS
WiFi
6LoWPAN
LTE-M / NB-IOT
LoRaWAN
4. 4 26 November 2015
M2M
IoT
Video
Cloud
Mobile
Apps
BlueTooth
Wireless but no gateway / smartphone
10 100 1k 10k 100k 1M 10M
(Bytes / Day)
4G3G2.5G 2.75GSigFox UNB
WmBUS
WiFi
6LoWPAN
LTE-M / NB-IOT
LoRaWAN
5. 5 26 November 2015
M2M
IoT
Video
Cloud
Mobile
Apps
Wireless Wide Area Networks – WAN
10 100 1k 10k 100k 1M 10M
(Bytes / Day)
4G3G2.5G 2.75GSigFox UNB
LTE-M / NB-IOT
LoRaWAN
Legacy
cellular
Low-Power
Wide Area Ntw
LPWAN
6. 6 26 November 2015
• PRO:
• Operated by MNOs MVNOs since 20 years
• Massive infrastructure & continued investments
• Licensed spectrum
• Ubiquitous service worldwide
• Secure communication (SIM card)
• Regulatory body = 3GPP – GSMA
• Extensive service offering
• Aiming at serving smartphones voice + data
• Aiming at increasing bandwidth 2G 3G 4G
to fight price erosion
• Legacy M2M communication channel
• CON:
• not suitable for low-cost battery-operated devices
Legacy cellular 2G 3G 4G
7. 7 26 November 2015
LPWAN for battery-operated devices
Connected
Not worth connecting
Container geolocation tag
Connected HVAC systems
Connected call buttons
Catering geolocation
Bicycle antitheft and geolocation
Industrial logistics
Consumer accessories
…
+ >200 new ideas
…75% of the M2M market by 2020!
8. 8 26 November 2015
• 2012: SIGFOX invented LPWAN with the
deployment of their UNB (Ultra-Narrow-Band)
network in FRANCE
• 2012: SEMTECH acquired CYCLEO a French
start-up inventor of the LoRa technology
• 2014: Inception of the LoRa Alliance.
• 2014: HUAWEI acquired NEUL
• 2015: 3GPP and GSMA have started working on
a NB-IoT standard aiming at providing improved
service in licensed spectrum in the frame of a 4G
upgrade
• Will LoRaWAN & SIGFOX be retained ?
• Goal is to deliver a standard by end of 2015 which
is a VERY ambitious objective
LPWAN for battery-operated devices
9. 9 26 November 2015
So they say… true or false ?
SIGFOX is a
proprietary
network while
LoRa Alliance is
an open standard
LoRa Alliance is a
SEMTECH proprietary
technology while
SIGFOX is compliant
with most transceivers
11. 11 26 November 2015
• A longer listening time per bit helps bring the
noise level down
• Additive White Gaussian Noise
Uncorrelated successive samples
Gaussian distribution of amplitude with variance s²
Variance of N-averaged samples = s² / N
• Bitrate /2 = bit duration x2
• Energy per bit x2 (+6dB)
• Noise energy x sqrt(2) (+3dB)
Improvement of Signal-to-Noise Ratio (SNR) by 3dB
• From 2G to SIGFOX
• 200kbps 100bps
• Bit duration extended by factor x2000
• Range improvement x sqrt(2000) = x45 in open
space at iso Tx power
Wider cells, less capex for operator
Same for LoRa
Why are LPWANs “long-range” ?
Listening time per bit
12. 12 26 November 2015
P
b
p
B
• A longer listening time per bit helps bring the noise
level down
• Two signals have the same power Pt at transmitter
output:
• Red is narrowband low bitrate
• Bandwidth = b << B
• Power Spectral Density = P >> p
• Green is wideband high bitrate
• Bandwidth = B >> b
• Power Spectral Density = p << P
• Same power Pt = P.b = p.B
• After travelling to a distant base-station, the
signals have been attenuated
• Wideband signal has sunk under thermal noise floor
• Narrowband signal still peaks above noise floor and
can be demodulated
• Same mechanism with spread-spectrum signals
although they seem to disappear below noise floor
Low bitrate signals travel longer distances
Bandwidth (Hz)
PSD
(dBm/Hz)
Power = PSD.BW
13. 13 26 November 2015
• SIGFOX or LoRawAN
LPWAN operators are
following similar network
architectures:
• Base-stations in the field
acting as 1 global base-
station
• Simplified MAC layer cloud-
operated
• API or callback interface to
retrieve or push messages to
customer’s application server
LPWAN global architecture
API
CallBack
3G, DSL, Ethernet, Satellite
LPWAN
operator
14. 14 26 November 2015
• Both SIGFOX and the LoRa
Alliance have a mandatory
certification program
• Necessary in order to validate
that the implementation of the
communication is 100%
compliant with the specification
• In case communication issues
arise, a certificate protects the
customer against the operator
• True with BlueTooth, 3G, WiFi,
ZigBee, etc
• Pre-certified modules and
reference-designs are a great
enabler!
Certification – why it is mandatory
EZR32
16. 16 26 November 2015
So they say… true or false ?
LPWAN data plans
cost 10x less than
GPRS/3G data plans
because 10x less
network infrastructure
required
True: for volumes >100k,
cost of communication is
~1€/node/y
But: be prepared for
aggressive competitive
2G/3G M2M data plans
in 2016…
17. 17 26 November 2015
Legacy cellular network layout complexity
18. 18 26 November 2015
• A moving UE (User Equipment) will have to switch to these 4
different cells in order to maintain connectivity
• The network manages this in real-time: Handover
• Complex mechanism supposing that the UE is updated in
real-time with the frequencies/codes of surrounding cells and
that the network pre-allocates bandwidth/slots in surrounding
cells in case the UE moves
Your smartphone keeps talking the network even if you do not
use it: battery drains in days/weeks
• 2G networks were designed for voice service: architecture
towards limiting latency to a few 10ms and maintaining audio
QoS at an average of 5-12kbps full-duplex
• 3G and 4G networks have added the capability to handle
transfers of large amount of packetized data at rates up to
50Mbps in order to serve mobile apps with smartphones
Still very different from an LPWAN usage!
How it impacts the User Equipment power consumption
19. 19 26 November 2015
LPWAN layout simplicity
• 5-10x less base-stations than 3G/4G
cellular networks
• All base-stations listen to the same
frequencies
Acting together as 1 global base-station
• Spatial redundancy helps securing
connectivity for objects in harsh
environments
• Filling the blanks is easier: just drop an
extra base-station, no modification of the
existing network required
• Very attractive for LPWAN operators: low
capex, low opex, low complexity
20. 20 26 November 2015
Outdoor Indoor = Outdoor – 20dB
French SIGFOX network
21. 21 26 November 2015
Bouygues Telecom
• End 2015 : 30 urban areas over 200k
people, ie 1500 towns
• H1 2016 : 80% population
• End 2016 : Full nationwide coverage
• Total planned number of BS = 4-5k
• To be compared with SIGFOX’s
current 1400 BS
Orange
• Current pilot in Grenoble
• H1 2016: nationwide deployment
• End 2016: full nationwide coverage
French LoRaWAN deployments
22. 22 26 November 2015
• Henri Crohas has announced the launch of
a LoRaWAN collaborative network in June
2016
• ARCHOS will subsidize 200,000 LoRa
picoWAN plugs and distribute them across
Europe at once
• Probably along with the sale of a nice
peripheral and killer application associated
with a service
• Each plug will create a small LoRaWAN
bubble strengthening the network inside
homes and buildings
• Very complementary to territorial MNO
deployments
ARCHOS – LoRaWAN collaborative network
23. 23 26 November 2015
So they say… true or false ?
SIGFOX is not
bidirectional while
LoRaWAN is
False: both are
bidirectional with the
same ETSI limitations
But: LoRa has a better
downlink sensitivity
with currently available
transceivers
24. 24 26 November 2015
• Both Up and Down link in same
unlicensed 868MHz ISM band
• European Telecommunications
Standards Institute (ETSI) regulates the
band usage
• Important rule: a base-station cannot talk
for more than 10% of the time
1 BS serving 100k – 1M objects cannot
answer them all!
• + Base-Stations operate in half-
duplex mode: blind when transmitting
Network is mostly uplink – downlink
used scarcely
LPWAN bi-directionality limitation in Europe
1%
10%
1%
1%
1%
25. 25 26 November 2015
• Service asymmetry is often pointed as a
service weakness
• Whereas an uplink-only service has a
great advantage
• Jamming a communication = jamming the
receiver
• Jamming an uplink protocol = jamming the
base-stations with a jammer in immediate
proximity
When the object is heard at 20km, very
difficult to jam all base-stations simultaneously!
• When engineering a service, the return-
channel can be served with another
technology
• Ex: a smoke detector cannot move by itself
to improve its network reach
How to turn an apparent weakness into a strength ?
26. 26 26 November 2015
SIGFOX
• Sensor-initiated bi-directionnal
communication
LoRaWAN
• Class-A
• Equivalent to SIGFOX with same
constraints (ETSI)
• Class-B
• Sensors are synchronized by network
beaconing - TDMA
• Unlikely in early ISM-band public
European deployments
• Useful in private networks for throughput
optimization
• Class-C
• Mains-powered sensors/actuators can be
in listen-mode full-time
LPWAN communication protocols
t
uplink
Tx 868.1MHz
Sleep downlink
Rx 869.5MHz
Dt, Df
1% duty-cycle
+14dBm
10% duty-cycle
+27dBm
27. 27 26 November 2015
So they say… true or false ?
SIGFOX function
HW cost is $1,
compatible with
all sub-GHz
transceivers in
the market
False: HW cost for
MCU+TRX+TCXO is > $1
Ultra-narrow band
modulation is tricky, much
more than a stack…
But: many transceivers
can be « hand-
modulated » to do the job
28. 28 26 November 2015
• Uplink:
• DBPSK modulation @ 100 bps
• Frequency: 868.13MHz +/- 100kHz
• Link budget = +14dBm (Tx) – (-145dBm) (ntw sensitivity) =
159dB >> GPRS
• 12-byte payload per message
• Message duration = 3x2s
- Repeated 3 times on 3 random frequencies within band
• Energy spent per message Etx = 3 x 2s x 50mA = 300mAs =
83µAh
• Downlink:
• FSK modulation @ 500 bps
• Frequency: Uplink + 1.4MHz (869.5MHz +27dBm sub-band)
• Link budget = +27dBm (Tx) – (-126dBm) (node sensitivity) =
154dB >> GPRS
• 8-byte payload per message
• Message duration = 300ms with average latency of 10s
• Energy spent per message Erx = 10.3s x 15mA = 154mAs =
43µAh
• As sensitive as GPS wrt frequency stability!
LPWAN radio characteristics – SIGFOX UNB
100Hz
29. 29 26 November 2015
• Frequency stability…
• Equivalent to phase noise
• Equivalent to RF carrier drift
• 100 bps 10ms bit duration
awfully long!
• DBPSK modulation = information born by RF
carrier phase increments from bit to bit
• Constellation must not rotate from bit to bit
• 45° rotation equivalent to 3dB loss of sensitivity
at the base-station !
• 45° in 10ms = 0.007ppm/s at 868MHz !
• It takes a little bit more than a TCXO to
guarantee this in a careful design…
This is why modules and experts are so useful
Where does lay the complexity in a SIGFOX node ?
bit n
Demodulator
decision boundary
updated after bit n
0 1
bit n+1
0.007ppm
only drift!
30. 30 26 November 2015
• TD-Next modules and SiLabs
EZR reference-design
• TD12xx for Europe/ETSI
• TD15xx for USA/FCC
• Historical partner of SIGFOX
• >400 customers currently
developing a solution around this
HW
• 99.9% of the SIGFOX nodes
deployed in the field
AVMS recommended SIGFOX modules/chipsets
Support/Volume
Timetomarket
EZR32
31. 31 26 November 2015
So they say… true or false ?
LoRa function
HW cost is more
expensive than
SIGFOX, only
compatible with
Semtech TRX
True: Semtech is to date the
only provider of LoRa
chipsets
But: no TCXO required
with LoRa
+ other IC vendors are
currently acquiring the
LoRa license
32. 32 26 November 2015
• Uplink:
• LoRa CSS (Chirp Spread Spectrum)
• 0.3-5 kbits per second (Adaptive Data Rate)
• Link budget = +14dBm (Tx) – (-142dBm) (ntw sensitivity)
= 156dB >> GPRS
• 0-250 bytes/message payload
• Message duration = 40ms – 1.2s
• Energy spent per message Etx = 1.2s x 32mA = 11µAh at
full sensitivity
• Energy spent per message Etx = 40ms x 32mA =
0.36µAh at min sensitivity
• Downlink:
• LoRa CSS (Chirp Spread Spectrum)
• 0.3-5 kbits per second (Adaptive Data Rate)
• Link budget = +27dBm (Tx) – (-137dBm) (node
sensitivity) = 164dB >> GPRS
• 8dB higher link budget than uplink allowing reducing SF
by 3 for downlink (9dB) 8x shorter message, 8x
increased capacity, 8x lower power consumption @
sensor
• 0-250 bytes/message payload
• Message duration = 20-160ms with average latency of 2s
• Energy spent per message Erx = 160ms x 11mA =
0.5µAh at uplink-equivalent link budget
LPWAN radio characteristics – LoRa technology
125kHz
33. 33 26 November 2015
• Microchip module
• RN2483 for Europe/ETSI
• AVMS reference-design
• MCU cortex M3
• SX1272 868MHz
• Or SX1276 bi-band
• >150 customers currently
developing a solution around
this HW
AVMS recommended LoRaWAN modules/chipsets
Support/Volume
Timetomarket
34. 34 26 November 2015
So they say… true or false ?
SIGFOX networks
have a greater
capacity than
LoRaWA Networks
True: Ultra-narrowband
has a greater spectrum
efficiency than LoRa
But: LoRa has a very
powerful adaptive
datarate capability
enabling scalable capacity
with infrastructure density
35. 35 26 November 2015
• 200kHz allocated BW in current
implementation
• Simultaneous frequency capacity =
200kHz / 100Hz x 10% = 200
simultaneous messages
• 10% is max limit for collisions
• Message = 208bits (2.08s) repeated 3
times on 3 random frequencies
• Base-station capacity = 200 x 24h x
3600s/h / 6.24s = 2.7M mess/day @
max link budget
LPWAN capacity – SIGFOX UNB
868.13MHz
Time
36. 36 26 November 2015
• 8 frequency channels of 125kHz in the 867-
868.5MHz sub-band
• 6 Spreading Factors (SF) orthogonal between
them yielding bitrates from 300 to 5k bps
• Base-station capacity = 8 x 24 x 3600 x 20% =
138k mess/day @ max link budget (SF12) & 20-
byte payload
• Base-station capacity = 8 x 32 x 24 x 3600 x 20%
= 3.3M mess/day @ min link budget (SF7) (=
max – 15dB) & 20-byte payload
• Capacity optimization with ADR (Adaptive Data
Rate) – network controlled
• Nodes close to BS are told to communicate at higher bit
rate shorter messages in time
Capacity can be increased with infrastructure
densification and depends on application
LPWAN capacity – LoRaWAN
868.1 868.3 868.5MHz
Time SF
SIGFOX
37. 37 26 November 2015
So they say… true or false ?
SIGFOX base-
stations cannot be
jammed by
LoRaWAN nodes
False: if too many
868.1MHz LoRaWAN nodes
are close to a SIGFOX BS,
they will mask distant
SIGFOX nodes from this BS
But: if the node is heard
by other SIGFOX BS,
messages will go through
38. 38 26 November 2015
SIGFOX immunity to LoRaWAN on shared frequency
SIGFOX BS
SIGFOX BS
L
L
S
S
Masking if
L-31dB > S-7dB
L > S+24dB
at base-station
Solved by
redundancy
39. 39 26 November 2015
So they say… true or false ?
LoRaWAN
networks cannot
be jammed by
SIGFOX nodes
False: if too many SIGFOX
nodes are close to a
LoRaWAN BS, they will mask
distant 868.1MHz LoRaWAN
nodes from this BS
But: if the node is heard
by other LoRaWAN BS,
messages will go through
40. 40 26 November 2015
LoRaWAN immunity to SIGFOX on shared frequency
LoRaWAN BS
LoRaWAN BS
L
L
S
S
Solved by
redundancy
Masking if
S > L+20dB (SF12)
S > L+5dB (SF7)
at base-station
41. 41 26 November 2015
So they say… true or false ?
SIGFOX does
not support
mobility while
LoRaWAN does
True: SIGFOX has a current
limitation wrt mobility under
certain conditions
Although more robust
against mobility by
nature, LoRaWAN has
limitations too
42. 42 26 November 2015
• Due to the very low bitrate, messages are very
long in time:
• SIGFOX: 26 bytes = 208 bits = 2.08s
• LoRa: 40-60 bytes = 320-480 bits = 1-1.5s
• A lot can happen in 2.08s when in urban
environment:
• Intersection crossing
• A car drives by
• Etc
• … whereas the base-station assumes the
channel (ie RF wave propagation path) does not
change for the whole duration of the message:
Linear Time-Invariant (LTI) channel
Message can be lost
An accelerometer can help assess when to Tx
Time and spatial redundancy help
LoRa error correction mechanism helps a lot
Mobility with LPWAN
Courtesy of www.aruco.com
2.08s
43. 43 26 November 2015
So they say… true or false ?
100m geolocation
without GPS is
possible in a LoRa
network while
impossible with
SIGFOX
True: if a node is
seen by at least 3
LoRaWAN base-
stations, it can be
located with an
accuracy of 20-200m
44. 44 26 November 2015
Geolocation inside a LoRaWAN network
• Measuring distances can be done 2 ways:
• RSSI (Received Signal Strength) = imprecise
• Time-of-Flight of the radio wave = precise
• LoRa, given its spectral/time properties, makes it
possible to measure a time of flight with a precision
of 200ns-1µs under good receive conditions
• 200ns . c (3e8 m/s) = 60m
• Nodes and network are NOT time synchronized
Not possible to measure distance to 1 BS only
• If BS are time-synchronized, they can time-stamp the
same message they receive (almost) simultaneously
These tiny time-of-arrival differences can be used to
compute the relative distance of a node between 2
BS: locus = hyperbola
3 BS required to compute a location in DTOA
(Differential Time of Arrival)
Dense network needed!!
BS1
BS2
BS3
45. 45 26 November 2015
So they say… true or false ?
LoRaWAN is for
private networks
only while SIGFOX
covers countries
False: both will
cover countries
But: LoRaWAN is
also great for private
networks whereas
this is not SIGFOX’s
business model
46. 46 26 November 2015
SIGFOX network operators
• France: SIGFOX
• Spain: CELLNEX
• NL: AEREA / TELE2
• UK: ARQIVA
• Portugal: NARROW NET
• Belgium: ENGIE M2M
• Luxemburg: RMS / POST
• Italy: NETTROTTER
• Denmark: IoT DENMARK
• Ireland: VIZOR
• Poland: EED
• USA: SIGFOX
LoRaWAN MNOs
• France: BOUYGUES TEL – ORANGE
• NL: KPN
• Belgium: PROXIMUS
• Switzerland: SWISSCOM
• South Africa: FASTNET
• UAE (United Arab Emirates): du
• Russia: LACE
• USA: SENET
• India: TATA
• Private networks everywhere:
ACTILITY
Who are the operators ? Nov 2015
+ 50 pilots for both technologies
around the world
47. 47 26 November 2015
LoraWAN Private network infrastructure by ACTILITY
• Customer owns, installs and administrates his private network across his buildings and
campuses
• Connects sensors, actuators, machines inside Intranet
• Compatible with public networks when available
• Also useful to strengthen / complement a public network in harsh industrial environments
Webhosted IT
Webhosted admin
Infrastructure
guillaume.remond@actility.com
49. 49 26 November 2015
• 3GPP has launched a NB-IoT initiative in
order to normalize LPWAN in licensed
spectrum worldwide
• Backed by 20+ operators inside GSMA,
including VODAFONE, TELEFONICA,
AT&T, VERIZON, T-MOBILE, DEUTSCHE
TELECOM, TELECOM ITALIA…
• Backed by base-station manufacturers,
including HUAWEI, NOKIA, ERICSSON,
ZTE…
• Backed by cellular chipset vendors such
as INTEL and QUALCOMM
… GSMA & 3GPP strike back!
50. 50 26 November 2015
• Objective: provide improved
LPWAN services in licensed
cellular spectrum
• Premium bi-directional service
• Higher bit rates
• Lower power
• Standardized across all operators
and countries
• No extra LPWAN-dedicated
infrastructure: native in 5G base-
stations
• Deployed over full countries
overnight
NB-IoT – LPWAN integrated in a 5G service
51. 51 26 November 2015
• Licensed cellular spectrum
• Revamping of 2G channels
• 180kHz sub-bands
• LTE guard-bands
• LTE sub-carriers
• Modulations
• Uplink: 300bps – 300kbps
• FDMA – GMSK (NEUL/HUAWEI)
• SC-FDMA (Single Carrier FDMA)
• Downlink: 200bps – 100kbps
• OFDMA
• Full balanced bidirectional service
• +23dBm sensor Tx power
NB-IoT – Bits and pieces of on-going proposals
52. 52 26 November 2015
NB-IoT – Roadmap to 2020
2016 2017 2018 2019
NB-IoT
standard
initial
release
LTE/NB-IoT
infrastructure
commercially
available
NB-IoT
service
commercial
launch
2015
Early pilot
Full worldwide
NB-IoT service
complementing
unlicensed
LPWAN
54. 54 26 November 2015
Sensors and
Actuators
• Product
selection
• Connectivity
• Power
management
Gateways
• Connectivity
Management
• LAN WAN
• Network
interface
• Cloud
messaging
• Security
Comms
infrastructure
• Network
selection
• In field
experience
• MNO
relationships
• IT investments
Cloud
Infrastructure
• Messaging
• Large volume
• New IoT
network
servers
• Enterprise
grade
• Analytics and
reporting
Applications and
Enterprise IT
• Integration with
and collection of
appropriate data
from cloud
service
• Customer
application
development
Edge to Enterprise Challenges
55. 55 26 November 2015
Scope of the Visible Things Platform