Lpwan Comparison: Low Energy Consumption With Nb-Iot, Lorawan and Sigfox [PDF]

Zitiervorschau

LPWAN COMPARISON Low Energy Consumption with NB-IoT, LoRaWAN and Sigfox WILHELM OELERS & HARALD NAUMANN

1

Version 2021-04-20 Copyright © 2021 Triptec HL UG, Lübeck Rubinweg 1, D-23566 Lübeck, Germany E-mail: [email protected] Phone: +49(0)451 / 30 40 718

2

1. 1. 2. 3. 4. 5. 6.

7.

8. 9. 10. 11. 12. 13.

14.

Table of Contents

Table of Contents List of figures List of tables Review 0G and outlook 5G mobile communications What does LPWAN mean? Why a report on LPWAN energy consumption? 6.1 Factors influencing the running time 6.2 Link budget 6.3 Maximum link budget of LPWAN technologies 6.3.1 Maximum link budget Sigfox 6.3.2 Maximum link budget LoRaWAN 6.3.3 Maximum link budget NB-IoT 6.4 Maximum payload in uplink and downlink of LPWAN technologies 6.4.1 Maximum payload in UL and DL at Sigfox 6.4.2 Maximum payload in UL and DL with LoRaWAN 6.4.3 Maximum payload in UL and DL for NB-IoT 6.5 Maximum number of messages in UL and DL with LPWAN 6.5.1 Maximum number of messages in the UL and DL at Sigfox 6.5.2 Maximum number of messages UL and DL for LoRaWAN 6.5.3 Maximum number of messages UL and DL for NB-IoT 6.5.4 Selection of modules for the test 6.5.5 Sigfox Module Selection 6.5.6 Selection LoRaWAN modules 6.5.7 NB-IoT Module Selection Link budget and energy consumption 7.1 Determining the maximum energy consumption with SIGFOX 7.2 Determining the maximum energy consumption with LoRaWAN 7.3 Determining the maximum energy consumption for NB-IoT 7.4 Determining the maximum energy consumption for SMS Range with LPWAN Range of a private LoRaWAN Analysis of the HS Offenburg study Analysis of the study by the University of Singapore Testing the range of LoRaWAN in Gleiberg Comparison of energy consumption Sigfox, LoRaWAN, NB-IoT 13.1 Uplink 12 bytes in comparison (calculated) 13.2 Uplink 24 bytes in comparison (calculated) 13.3 Uplink 64 (72) bytes in comparison 13.4 Uplink 512 bytes in comparison (calculated) 13.5 Expected energy consumption of NB-IoT, LoRaWAN and Sigfox Measurement of energy consumption 3

3 5 6 7 9 11 11 12 14 15 15 17 18 18 18 20 20 20 21 22 22 22 22 22 23 23 23 24 24 24 26 28 29 30 32 33 36 36 37 37 38

14.1 Measurement SIGFOX module Wisol (Seong Ji) 14.2 Measurement LoRaWAN Module Acsip S76 S 14.3 Measurement NB-IoT module Quectel BG95 , BC66 and BC68 compared to LoRaWAN Acsip and Sigfox Seong Ji 14.3.1 Measurement series 12, 24, 64, 512 bytes at 134 to 154 dB link budget 15. Conclusion

4

39 39 40 41 46

2.

List of figures

Figure 1 From 0G in 1918 to 5G in 2019 ............................................................................. 8 Figure 2 5G Massive IoT is NB-IoT ..................................................................................... 9 Figure 3 LPWAN compared with tractors, source LPWAN Cookbook .............................. 10 Figure 4 Calculation of the link budget, source Harald Naumann, LPWAN Cookbook ..... 14 Figure 5 Sigfox uplink sensitivity -142 dBm, link budget 163 dB........................................ 16 Figure 6 Interference Measurements in the European 868 MHz ISM Band with Focus on LoRa and Sigfox, ............................................................................................................... 17 Figure 7 NB-IoT Link budget calculator ............................................................................ 18 Figure 8 LoRa modem calculator, SF12 in EU, 255 bytes, Time on air 9019 ms .............. 19 Figure 9 LoRa Modem Calculator, SF12 in EU, 25 Bytes (12 Bytes net), Time on air 1483 ms...................................................................................................................................... 22 Figure 10 LoRa Modem Calculator, EU: 25 bytes with optimiser on = 1483 ms................ 25 Figure 11 LoRa Modem Calculator, EU: 51 bytes with optimiser on = 2793 ms................ 25 Figure 12 LoRa Modem Calculator, US: 11 bytes limitation ............................................. 25 Figure 13 Range calculation with Okumura-Hata model.................................................... 26 Figure 14 Municipality of Kirchheim with its 12 villages. Inhabitants per village marked in blue circles. Source LPWAN Cookbook............................................................................. 27 Figure 15 Simulation of the first LPWAN gateway in the main village of Kirchheim........... 28 Figure 16 Simulated network coverage of the municipality of Kirchheim with 12 LoRaWAN gateways .......................................................................................................................... 28 Figure 17 Presentation at the Wireless Congress 2019 in Munich, Field tests - RF coverage & signal quality measurements, slide 36............................................................ 29 Figure 18 Package loss of 15% to 95% packet for LoRaWAN with SF 12 ........................ 31 Figure 19 Test of range and max. link budget with LoRaWAN EU with SF12, -121 dBm with -17.2 dB SNR ............................................................................................................. 32 Figure 20 SMS with 33 dBm transmission power .............................................................. 33 Figure 21 Energy consumption 12 bytes with NB-IoT, LoRaWAN and Sigfox - calculated, not measured .................................................................................................................... 34 Figure 22 Data volume at 144 dB link budget with LoRaWAN EU, Sigfox and NB-IoT ..... 36 Figure 23 Energy consumption 72 bytes with NB-IoT, LoRaWAN and Sigfox - calculated, not measured..................................................................................................................... 37 Figure 24 Performance of NB-IoT, LoRaWAN and Sigfox over time ................................. 38 Figure 25 Keithley DMM6500 Multi meter - Front panel- Credit: TEKTRONIX INC ........... 40 Figure 26 Keithley DMM6500 Multi meter - 10 pA to 10 Ampere - Credit: TEKTRONIX INC 41 Figure 27 Measurement set up with DUT in the basement. .............................................. 41 Figure 28 Measurement set up with Quectel BC66, BC68 and BG95-M2 ......................... 42 Figure 29 Five LPWAN test PCBs ........................................................................................................................................... 44 Figure 30 Energy consumption in average during uplink of 12 bytes with 134, 144 and 154 dB link budget ................................................................................................................... 45 5

Figure 31 Energy consumption in average during uplink of 12 bytes with 134, 144 and 154 dB link budget ................................................................................................................... 46 Figure 32 Energy consumption in average during uplink of 64 bytes with 134, 144 and 154 dB link budget ................................................................................................................... 47 Figure 33 Energy consumption in average during uplink of 512 bytes with 134, 144 and 154 dB link budget ........................................................................................................... 48 Figure 34 Energy consumption with 12, 64 and 512 Bytes at 134, 144 and 154 dB link budget................................................................................................................................ 49

6

3.

List of tables

Table 1 RSRP related to the link budget ........................................................................................................................................... 49

7

4. Review: From 0G to the outlook for 5G mobile communications Before we start comparing the relatively new LPWAN technologies, let's take a look back at the history of mobile telephony. 0G was a 1918 German invention, the first mobile telephone call1 to the public telephone network was made in that year from a German express train.

Figure 1 From 0G in 1918 to 5G in 2019

The radio tower of Babel of 1G standards disappeared and was replaced by digital Esperanto at 2G. With 2G came data transmission in addition to voice communication. Digital Esperanto has been optimised over generations to 5G. 5G is available on the ground, via satellites in the sky and, if desired, underground. In May 2019, it was announced that 5G broadcast could replace DVB-T2 from 20272. Digital radio DAB+ could be the next victim. With 5G, the frequency range is no longer limited to long wave as it was with 0G in Germany and now extends from 400 MHz to 28 GHz. With 5G, there is no separation of voice and data, there is high speed to transmit data, short response time, low power devices and support for a range of broadcast options including augmented reality to live sports events, 8K streamed video and more. A network for LPWAN, streaming video to your smart phone and 3D gaming is becoming a reality ahead of further exciting developments. The holodeck of the Star ship Enterprise can now be projected into Hyde Park or the middle of the desert by 3D VR glasses. Wireless technologies have always evolved continuously to meet increasing demands and higher 1Cf.

Hessberger, Stephan: ÖbL, Zugfunk 1918 - 1926 - 1940, in: Öffentlicher beweglicher Landfunk, [online] http://öbl.de/A-Netz/Rest/Zugfunk/Zug1926.html [06.02.2021]. 2 Vgl. Sawall, Achim: 5G Broadcast soll ab 2027 DVB-T2 ablösen, in: Golem.de, 21.04.2019, https://www.golem.de/sonstiges/zustimmung/auswahl.html?from=https%3A%2F%2Fwww.golem.de%2Fnew s%2Ffernsehen-5g-broadcast-soll-ab-2027-dvb-t2-abloesen-1904-140937.html (abgerufen am 18.04.2021).

8

technical requirements. We didn't stop at 192-metre antennas in 1918, but have continued to refine radio technologies from 0G to 5G. Since the deployment of zero and first generation mobile networks, the telecommunications industry has faced many new challenges in terms of technology, efficient use of spectrum and, most importantly, end-user security. Future wireless technologies will enable ultra-fast, feature-rich and highly secure mobile networks, remote sensing devices will be able to operate for ten years from one battery and also offer delay times of 1 ms. If you transfer this mentally to transport, then an equivalent capability would see the same vehicle used for the speed and handling of a Formula 1 car on the Nurburgring, a heavy tractor ploughing a field and delivering the visceral acceleration of a top fuel dragster. A dragster does not have much deceleration (a one-time event with a parachute) but will smoke the Formula 1 racing car in acceleration. At the Nurburgring, the race car then wins because the dragster runs out of fuel after a few hundred metres and can't take any of the 170 corners. We can't pull a plough with the dragster and the racing car despite their power because both lack the necessary endurance and torque. Even the fastest tractors are not going to win a race with the dragster nor the Formula 1 race car. The logical conclusion is that 5G cannot work with the approaches and ideas from 0G to 3G. With 0G to 1G, only one task (voice) was ever required from the network. 2G saw the addition of data transmission to voice. 4G is closer to the ubiquity of the 5G idea, but still not optimal. NB-IoT for example operates as a network within a network. There, 200 KHz has been cut out of the LTE band and replaced by NB-IoT. The 200 KHz bandwidth for 12 NB-IoT channels is possible several times per 5 MHz bandwidth in an LTE band. However, the server for managing the subscribers and the network traffic is different from the one for the smart phones. With 5G on the other hand, all known applications, all necessary servers and the possibility of extension to use cases not yet known have been migrated to one overall system. Today's LTE Advanced Networks will transform into 5G networks bit by bit in the future. To achieve a higher data rate, 5G technology will use millimetre waves and unlicensed spectrum for data transmission. New complex modulation technology has been developed to handle the massive data rate. To make the Internet of Things real, the possible number of subscribers per base station will be increased from 100,000 to 1,000,000. The network architecture will expand functionality and analytics capabilities for Figure 2 5G Massive IoT is NB-IoT industries, autonomous driving, health care and security applications. We are moving from 1 GB/s with LTE-A to ultra-fast mobile internet 9

connections of up to 10 GB/s. Latency is reduced to milliseconds. With 5G, the frequency spectrum can be dynamically allocated to other services at any time. Frequency ranges that are allocated to the public network operator can be assigned to the police and rescue services or the fire brigade in the event of a disaster. For 0G and 1G mobile radio, large areas were covered with tall base station antennas. 150 base stations were sufficient for the German 1G cellular network called B-network to offer 95 % of the population the possibility of mobile telephony but the radio network could only actually serve 11,000 subscribers3. 5G uses small cells and beam forming to increase efficiency. 5G-U (5G unlicensed) brings the 5G base station into our living room so that everyone becomes an operator of an ultra-small base station. Side stepping 5G mobile communications with NB-IoT and LTE-M NB-IoT and LTE-M are the LPWAN technologies of cellular network operators and have now been officially integrated into 5G. In the section above, we compared the different tasks 5G is capable of, with the transport analogy of a Formula 1 race car, dragster and tractor. The tractor is slow and has high torque to pull a plough. In this white paper, we look at the slow radio technologies with high range. In this case, not all tractors are equal. There are very big differences:

Figure 3 LPWAN compared with tractors, source LPWAN Cookbook

The common feature of LPWAN technologies is low energy consumption. Below we compare the best known technologies: NB-IoT, LoRaWAN and SIGFOX. We did not compared with LTE-M for now, because LTE-M measurements was not ordered with this study. Nevertheless, a comparison with LTE-M can follow next months.

3

Vgl. Wikipedia-Autoren: B-Netz, in: Wikipedia, 11.12.2004, https://de.wikipedia.org/wiki/B-Netz (abgerufen am 18.04.2021). 10

5.

What does LPWAN mean?

LPWAN is the abbreviation for Low Power Wide Area Network. Wide Area Networks include GSM, UMTS and LTE, among others. LPWANs include NB-IoT, LoRaWAN, SIGFOX, Weightless and many more. The adoption of the phrase “low power” is unfortunate. What counts for practical applications is the energy consumption, and the energy is obtained by multiplying the power by the time. A better term would be LEWAN (Low Energy Wide Area Network). In the LPWA world, many terms are used, these are often simply stated and the practical facts are frequently not analysed. In this report we would like to illuminate this dark area of terms and point out a few differences between the LPWAN technologies. Low Power - Power - Energy As already mentioned, power is the wrong term. The battery stores energy, not power. The capacity is written on the battery in mAh, but this is not an amount of energy. The energy is obtained by multiplying voltage x current x time. If you multiply the 2000 mAh of an AA cell by the average terminal voltage, you get energy in mWh. 2000 mAh x 1.2 volts gives 2400 mWh = 2.4 Wh. In the electrical industry we tend to use Wh or Ws and joules to describe energy consumption. In the food industry, calories are used. We could express the energy consumption of an LPWAN module in calories. 1 joule is 0.239006 calories. 1 joule is equal to 1 Ws. Watt-hours (Wh), or rather kilowatt-hours (kWh) is what the "electricity meter" counts. It does not count the electricity but the energy consumed and is therefore an energy meter. A day has 24 hours x 60 minutes x 60 seconds = 86,400 seconds. Our AA cell therefore stores 2400 x 86,400 mWs = 207,360,000 mWs. Below we take a closer look at the power consumption of the Low Power Wide Area (LPWA) modules. LPWA modules compared at 154 dB link budget Sending an SMS with 33 dBm transmission power, including logging into the GSM network in band 8 (880 MHz - 960 MHz), requires approximately 4000 mWs. All LPWA modules based on NB-IoT, LoRaWAN or SIGFOX require in some conditions less energy than SMS. If you compare the technologies on the modules with LPWA chip sets in series production from 2018, the wheat is separated from the chaff. To make it comparable, we first look at the LPWA modules With 134, 144 and 154 dB link budgets (LB). The maximum link budget is the delta of radiated power and sensitivity plus antenna gain of the antennas at the transmitter and receiver minus the attenuation from transmitter to receiver. The link budget is asymmetrical for SIGFOX and LoRaWAN in uplink and downlink. This asymmetry means that a message in one direction or the other cannot be acknowledged. The 154 dB link budget is named “Coverage Enhancement Level 1 (or CE Level 1, CE1). The Coverage Enhancement Level level indicates the quality of the link. If we were to use the maximum link budget of 164 dB for comparison, SIGFOX and LoRaWAN would fall out in this comparison. SIGFOX is limited to approximately 156 dB in 11

the downlink and can also only transmit 12 bytes per telegram in the uplink. Therefore, we limit the comparison to 12 bytes per message. The 164 dB link budget for NB-IoT is 20 dB better than the maximum 144 dB link budget for GSM. 3 dB gain means that with GSM 900 at 33 dB transmit power, the average current when transmitting is reduced from 200 mA with GSM 1800 to 100 mA. Since the energy is calculated from current x voltage x time, the energy is reduced from approximately 4000 mWs to 2000 mWs with GSM 1800. But to increase the link budget from 144 dB to 153 dB (144 dB + 3 dB + 3 dB + 3 dB), we have to double the 4000 mWs three times. At 33 dBm + theoretical 9 dB = 42 dBm, we would have 4000 mWs x 8 = 32,000 mWs energy consumption.

6.

Why a report on LPWAN energy consumption?

In many places, one reads again and again about the mystical running time of 10 years from the original battery. Some authors of LPWAN articles even mention 15 years. Why not 20 or 30 years? Most of the statements on run times of 10 to 15 years do not refer to any mathematical or physical basis for calculation. Therefore, some boundary conditions have to be established. The run time of a battery operated LPWAN device is determined by many factors.

6.1 Factors influencing operating life • • • • • • • • •

Required energy consumption per transmission Repetition of messages as default (Sigfox) Repetition of messages in case of packet loss due to collision Repetition of messages in the absence of a base station in the vicinity Number of transmissions per day Message data payload constraints Capacity of the battery used Self-discharge of the used battery Self-discharge of the battery due to currents in sleep mode

In extreme cases, there is no network coverage in the buildings in the desired region. The consequence is 100% packet loss. If indoor coverage is required, then such LPWAN technology is unsuitable from the outset. The coverage of a public LPWAN should be determined with on-site tests. If you are aiming for a private LPWAN, you should not be swayed by the extreme coverage of some LPWAN evangelists. Realistic ranges with indoor coverage in urban areas are between 1 and 2 km with a link budget of 144 dB in the 900 MHz frequency range. In this paper we have a focus on the energy consumption per transmission and the repetition of messages. Each repetition due to packet loss of a message multiplies the original consumption for a packet. Furthermore, we look at the energy required for packets of 12, 24, 64 and 512 bytes. The explanations of the parameters and values in the following table can be found in the text below. 12

Sigfox EU

Sigfox US

LoRa EU

LoRa US

Max. link budget UL Max. link budget DL Worst Case Link Budget Max. Payload UL Max. Payload DL Max. Messages UL per day Max. Messages DL per day Bps @ 144 dB Link budget UL Bps @ 154 dB Link budget UL

approximately 163 dB approximately 158 dB approximately 158 dB

TBD

approximately 141 to 146 dB approximately 151 to 156 dB approximately 141 dB

TBD

NB-IoT NB2 Global 164 dB

TBD

164 dB

TBD

164 dB

12 Byte

12 Byte

243 Byte

11 Byte

12 Byte 512 Byte

8 Byte

8 Byte

243 Byte

11 Byte

144

144

Duty Cycle 1 %

No limit

317 Byte 2536 Bits 317 Byte 2536 Bits No limit

8

8

10 % distributed over all nodes

no limit

no limit

Not considered

100 bps

TBD

297 bps

TBD

27000 bps

Not considered

100 bps

TBD

0 bps, no connection

TBD

Not considered

Radio modules

Wisol WSSFM10R1 AT

not examined

Acsip S76S

not exami ned

approxi mately 6000 bps Quectel BC66, BC68, BG95M3

TBD TBD

Here in comparison 134 to 154 dB 134 to 154 dB 134 to 154 dB

12 Byte 512 Byte Not considered

5 radio modules, see left

Table 1 NB-IoT, LoRaWAN, Sigfox in comparison

6.2 Link budget The link budget is the sum of all losses and gains in the transmission of a radio wave. A radio wave is amplified or attenuated by the antenna. Data can be lost during the transmission of a signal between transmitter and receiver. The study of losses and gains is therefore important to calculate the reliability and efficiency of a radio link. For a signal to be received, the delta of the output power minus the sum of all losses must be greater than the sensitivity of the receiver. For bidirectional LPWAN radio links, the losses between transmitter and receiver apply in both directions. For LPWAN technologies in the unlicensed band, the link budget is usually unbalanced. This means that although the sender can transmit to and reach the gateway, the gateway's acknowledgement does not reach the sender. However, the reverse case can occur. Therefore, the data sheets of the radio modules and the gateways must be carefully checked. This becomes difficult if manufacturers or network operators do not publish the data. If the transmit power and sensitivity are known, only the path losses need to be subtracted from the delta to determine whether reception is possible with a satisfactory system design. 13

Some of the losses are under the control of the designer of the hardware: The antenna and the ground plane of the PCB affect the power radiated by the antenna. More than 0 dBd or the equivalent 2.15 dBi is not to be expected from a monopole antenna. In most cases it is less. Part of the radiated energy disappears as path loss in the housing of the unit. Usually there is a matching network between the radio module and the antenna. Energy is lost in the feed line to the antenna and in the matching network. With a return loss of -6 dB or a VSWR of 3, you lose 50 % of the power (3 dB). The rest that remains is not always radiated by the antenna, some may remain as thermal energy in the antenna or in the enclosure. To detect these losses, the antenna radiation pattern must be measured in at least three axes.

Figure 4 Calculation of the link budget, source Harald Naumann, LPWAN Cookbook

The energy that then enters free space is attenuated over distance. The free-space attenuation is only half the truth, because the nature of the terrain also attenuates the radio wave. A flat water surface has less attenuation than farm land. When bushes, trees and buildings are added, it gets worse. Topography also affects path loss. At the end of the long path, the radio wave hits a building wall. A cement wall attenuates a 900 MHz wave by about 20 dB, another wall in the building quickly attenuates it by another 8 dB. Losses due to fading only occur in the area near the antenna and are calculated at 8 dB. Thus, a loss of 28 dB plus free-space attenuation must be expected as a minimum inside any building. In 1968, Okumura4 intensively investigated the attenuation of radio waves in cities and

4

Okumura, Y.; Ohmori, E.; Kawano, T.; Fukuda, K.: Field strength and its variability in VHF and UHF land mobile radio service. Review of the Electrical Communications Laboratory, 1968, H. 9-10, pp. 825-873.

14

developed a theoretical model. This Okumura-Hata model5 often runs in the background of wave propagation simulation software. The software adjusts the attenuation to the topography and the nature of the earth's surface. All forests are included in the calculation with a uniform tree height, if you want a more accurate approximation, you can change the height of the trees. Buildings and their attenuation are also simulated with general values. The same applies to water areas, cultivated fields and meadows. If necessary, even the refraction of the radio wave at the top of a mountain or hill can be included. The complex refractions caused by roofs or the reflection from walls cannot be calculated with simple simulation software, this requires 3D models of the buildings in the cities and expensive simulation software. Simple simulation software is usually sufficient for planning private LPWANs. The attenuation caused by refraction on roofs and reflection on house walls can be included in the path loss in the form of a static value. The sum of the losses are identical in the same terrain or city for NB-IoT, LoRaWAN and Sigfox because these three technologies use the frequency range of approximately 900 MHz.

6.3 Maximum link budget of LPWAN technologies In most LPWAN technologies in the licence-free band, the link budget is asymmetrical for uplink and downlink. In order to6 transmit a message from the node to the base station/gateway and vice versa with an acknowledgement, the link budget must be symmetrical or one must use the worse of the two values in the consideration. If this is not observed, a message cannot be acknowledged with an asymmetrical link budget. In addition to the maximum permitted transmission power at the node, antennas with an antenna gain of 2.15 dBi are permitted at the node in Europe. In the unlicensed European bands this results in a power of 14 dBm plus 2.15 dBi antenna gain for a total radiated power of 16.15 dBi. In the following considerations, the 2.15 dBi are rounded down to 2 dBi. We set losses in the housing of the unit to 0 dB. At the gateway, 27 dBm transmission power is permitted in the unlicensed band with LoRaWAN and Sigfox. Here, too, we add the 2.15 dBi, which we round down. On the gateway side, antennas with 5.15-7.15 dBi antenna gain are not unusual. However, the higher the antenna gain, the smaller the aperture angle. If the angle becomes too small due to an antenna gain that is too high, then no signals can be received below the antenna of the gateway. We therefore use an antenna gain of 5 dBi in our considerations. This antenna gain is then 3 dB higher than allowed. To compensate for this, the maximum permitted transmission power at the gateway must be reduced by 3 dB, so that in the end there is 29 dBi radiated power. For NB-IoT the usual transmit power is 23 dBm and becomes 25 dBm radiated power due to the 2 dBi antenna gain. For cellular base stations, sector antennas with up to 20 dBi antenna gain are common. In the licensed band, there is first of all no limit to the transmit 5

Cf. Wikepedia authors: Hata model, in: Wikipedia.org, [online] https://en.wikipedia.org/wiki/Hata_model [06.02.2021]. 6 Cf. Wikipedia authors: Link budget, in: Wikipedia.org, [online] https://en.wikipedia.org/wiki/Link_budget [06.02.2021].

15

power at the base station and at the radio node. GSM modules transmit at 33 dBm and for NB-IoT the transmission power has been reduced to 23 dBm to reduce energy consumption. In the following considerations, we do not address the losses in the antenna cables from the gateway to the gateway antenna and set them to 0 dB.

6.3.1 Maximum link budget: Sigfox Since Sigfox is a proprietary radio technology, the sensitivity of the gateway is not easily visible. In addition, the antennas on the gateway are also not well documented. According to a video7 from Sigfox, the sensitivity at the gateway is -142 dBm. If you add the 14 dBm transmission power with the 2 dBi antenna gain at the radio node, the 5 dBi antenna gain at the gateway and the -147 dBm sensitivity at the Sigfox gateway, you get a link budget of 163 dB for the uplink. The link budget for the downlink can be easily determined: 27 dBm transmission power at the gateway added to 2 dBi antenna gain at the gateway +2 further dBi antenna gain at the node and -127 dBm sensitivity (according to the data sheet of the radio module) result in a maximum downlink link budget of 158 dB. For the downlink, the budget is therefore 5 dB worse than in the uplink.

Figure 5 Sigfox uplink sensitivity -142 dBm, link budget 163 dB

6.3.2 Maximum link budget: LoRaWAN In some papers on calculating the link budget of LoRaWAN, -137 dBm sensitivity at the gateway is used. The -137 dBm can only be achieved under the laboratory conditions of a measurement chamber. According to the manufacturer Semtech, the LoRa IC in Europe can still receive at 20 dB below the noise level at 125 KHz bandwidth with a spreading

7

Cf. Sigfox: YouTube, in: Radio Access Network, 08.02.2017, [online] https://www.youtube.com/watch?v=gGvM6KEDIdE [06.02.2021].

16

factor 12. According to a study8 by the University of Denmark, the noise in cities is between -85 dBm and -105 dBm.

Figure 6 Interference Measurements in the European 868 MHz ISM Band with Focus on LoRa and Sigfox,

This means that the real sensitivity of the gateway can be found at -120 to -125 dBm. If you look at the TTN user forum and take a closer look at the test series in the following chapter and analyse various test series on the Internet, you will find that -120 dBm to -125 dBm is a value that appears again and again. In our considerations we do not use the worst value of -85 dBm, but use -100 dBm noise level to calculate the sensitivity. Anyone planning their own LPWAN with LoRaWAN should determine the local specific noise level in industrial facilities in advance. 14 dBm transmit power at the node added with 2 dBi antenna gain at the radio node plus another 5 dBi antenna gain at the gateway and -120 dBm sensitivity result in a link budget of 141 dB for uplink. If we are optimistic and calculate with 5 dB less noise, we arrive at 146 dB link budget for the uplink. For the downlink we get 27 dBm transmit power at the gateway with +2 dBi antenna gain 8

Lauridsen, Mads: Interference measurements in the European 868 MHz ISM band with focus on LoRa and SigFox, in: Aalborg University's Research Portal, 19.03.2017, [online] http://vbn.aau.dk/en/publications/interference-measurements-in-the-european-868-mhz-ism-band-with-f [23.01.2021].

17

at the gateway +2 dBi antenna gain at the node plus -120 dBm sensitivity at the node results in a link budget of 151 dB. We calculate optimistically and use 146 dB link budget. However, this 146 dB link budget in the is far from the ideal values in a measurement chamber with 158 dB. Extremely idealised reports calculate using antennas at the gateway with unrealistically high antenna gain in order to show the link budget higher than 158 dB.

6.3.3 Maximum link budget NB-IoT The maximum link budget for NB-IoT for uplink and downlink symmetry is 164 dB. However, the maximum budget is not desirable because the energy consumption at 164 dB link budget is about ten times higher than at 154 dB. NB-IoT networks are often planned for a link budget of 154 dB or even 144 dB. We will therefore limit ourselves to 144 and 154 dB link budgets in our measurements. The 164 dB figure is difficult to achieve under real-world conditions. If the link budget is asymmetric, the link will easily break near 164 dB. With the online tool you can calculate the link budget at NB-IoT9 yourself. https://5g-tools.com/nb-iot-link-budget-calculator/

Figure 7 NB-IoT Link budget calculator

If you switch the bandwidth from 15 KHz per channel for NB-IoT to 3.75 KHz, you will see that this changes the sensitivity by 6 dB. The sensitivity of a receiver of any radio technology is affected by the bandwidth. The wider the frequency bandwidth, the more thermal noise is received. Therefore, below a 154 dB link budget, the sensitivity of NB-IoT is improved by changing the bandwidth. The lower the bandwidth, the lower the transmission rate. This is not only true for NB-IoT, but for all radio technologies. In order for Sigfox to achieve the extremely high sensitivity of -142 dBm, a bandwidth of only 100 Hz is used.

9

Vgl. Vinogradov, Oleg: NB-IoT Link budget calculator | 5G-Tools.com, in: 5G Tools for RF Wireless, 13.09.2020, [online] https://5g-tools.com/nb-iot-link-budget-calculator/ [06.02.2021].

18

6.4 Maximum uplink and downlink payload of LPWAN technologies 6.4.1 Maximum UL and DL payload with Sigfox The maximum payload for Sigfox is limited to 12 bytes in the uplink. In downlink, Sigfox is limited to 8 Bytes.

6.4.2 Maximum UL and DL payload with LoRaWAN The maximum payload in uplink and downlink is not the same for LoRaWAN in Europe and the USA. In Europe, the maximum payload is only limited by the physical size of 255 bytes of the LoRa IC. The header in the LoRaWAN protocol has 13 bytes. This limits the theoretical payload for a telegram to 242 bytes. However, if 255 bytes are sent with the spreading factor 12, the channel is occupied for almost 14 seconds. Another telegram cannot then be received on this channel for 14 seconds. Since LoRaWAN uses the unsynchronised Aloha method, long channel occupancy leads to an extremely high packet loss. The LoRaWAN protocol uses only 51 bytes of maximum payload for SF11 and SF12 in Europe.

Figure 8 LoRa modem calculator, SF12 in EU, 255 bytes, Time on air 9019 ms

But if you are planning a global IoT product, you have to take into account the boundary conditions for LoRaWAN in the USA. In the USA, the channel may only be occupied for 19

400 ms in uplink or downlink. If you then use the LoRa calculator, you find that you can use a maximum spreading factor of 10 and that the payload is limited to 11 bytes. Changing the spreading factor from 12 to 10 causes the sensitivity to drop by 6 dB. This in turn can be compensated for by increasing the transmit power by 6 dB. The typical maximum power of a LoRa module is 20 dBm, which can be achieved without an amplifier. If a transistor is used for amplification, then these LoRaWAN modules automatically become more expensive. Therefore, if you want to develop a global product on LoRaWAN, you have to limit yourself to an 11 byte payload.

6.4.3 Maximum UL and DL payload for NB-IoT For NB-IoT, the maximum payload in uplink and downlink is asymmetrical. Furthermore, the payload is usually specified in bits in the public literature. In addition, the payload is unequal for NB-IoT NB1 and NB-IoT NB2. NB2 brings an improvement due to the larger maximum payload and larger amounts of data no longer have to be broken down into many small packets. 3GPP barrier 3GPP Release Voice SMS Bandwidth Bandwidth per channel

NB-IoT NB1 Release 13 No Mainly not 200 KHz 15 or 3.75 KHz

NB-IoT NB2 Release 14 Push to talk Mainly not 200 KHz 15 or 3.75 KHz

Number of channels Data Max. link budget Latency Mobility Max uplink data rate Max uplink TBS Max DL data rate

12 Half duplex 164 dB 0.5 - 10 seconds Up to 100 km/h, Nomadic mobility (reconnection) 65 kbps 1000 bits, 125 bytes 27 kbps

12 Half duplex 164 dB 0.5 - 10 seconds Up to 100 km/h, Better mobility (reconnection) 159 kbps 2536 bits, 317 bytes 127 kbps

Max downlink TBS

680 bits, 85 bytes

2536 bits, 317 bytes

Power level Positioning

20, 23 dBm Cell ID

14, 20, 23 dBm OTDOA, E-CID

Table 2 Comparison NB-IoT NB1 and NB2

6.5 Maximum number of messages in UL and DL with LPWAN 6.5.1 Maximum number of messages in UL and DL with Sigfox The maximum number of messages in the uplink is limited to 140 messages per day with Sigfox. This limit is based on the European duty cycle of 1 %. Sigfox repeats a telegram twice with each transmission. Each telegram occupies the channel for approximately 2 20

seconds. In total, this results in approximately 6 seconds per message. 1 % of 1 hour comprising 3600 seconds results in 36 seconds. 36 seconds/6 seconds results in 6 telegrams per hour. 6 × 24 results in 144 per day. A Sigfox device is therefore limited to 144 messages per day by the European regulations with the 1 % duty cycle. Sigfox itself has then set itself a limit of 140 messages per day. If you look more closely at the 6 messages per hour, you will see that results in transmission every 10 minutes. For tracking moving objects, an interval of 10 minutes is quite probably unsuitable. One may transmit the six messages per hour at shorter intervals, but then one must stop transmitting for a correspondingly long time to stay within the duty cycle limit. Sigfox limits the downlink to four messages per day. In addition, only 8 bytes are possible in the downlink. This extreme limitation is necessary in the unlicensed frequency band because a Sigfox gateway is limited to 10 % duty cycle per hour. These 360 seconds must be sufficient to supply all participants of a gateway with receipts. In the downlink, there is no protection of the telegrams by repetition. It is sent with 600 bit/s. 8 bytes +12 byte header result in 20 bytes gross or 160 bits. 160 bits/ 600 bits per second results in a channel occupancy time of 267 milliseconds. 360 seconds/267 ms results in a maximum of 1350 downlinks per gateway and hour or 1350 × 24 hours equals 32,400 downlinks per day. As a result, the Sigfox protocol is unsuitable for serving a larger number of participants in the downlink.

6.5.2 Maximum number of messages in UL and DL with LoRaWAN At first glance, LoRaWAN does not limit the number of messages in the uplink or downlink. At second glance, you can see 1 % duty cycle in the uplink and 10 % duty cycle in the downlink. If one limits oneself to approximately 12 bytes of payload and uses spreading factor 12, then a telegram occupies the channel for 1.483 seconds. If you add packet losses due to collision or interference on the radio channel similar to Sigfox, then you occupy the channel for 4.45 seconds and can therefore only transmit 8 messages per hour.

21

Figure 9 LoRa Modem Calculator, SF12 in EU, 25 Bytes (12 Bytes net), Time on air 1483 ms

In the downlink, one is limited by the 10 % duty cycle to 360 seconds of transmission time at the gateway. With a spreading factor of 7, the channel is only occupied for a few milliseconds. With a spreading factor of 12 with a 12-byte payload, 1.483 seconds. On average, the channel is therefore occupied for 0.74 seconds for an acknowledgement. If you divide the 360 seconds by 0.74 seconds per telegram, you get a maximum of 486 acknowledges per hour. With 1000 or even 5000 participants per gateway supplied, a real receipt operation (downlink or acknowledgement) is therefore impossible.

6.5.3 Maximum number of messages in UL and DL with NB-IoT With NB-IoT there is no duty cycle and therefore no time limit. One LTE block with 200 KHz bandwidth gives us twelve channels for NB-IoT. The transmission speed is a maximum of 21,000 bit/s and for an extreme link budget of 164 dB only 300 bit/s. In our comparison, we limit the link budget to 154 dB and thus get a lower speed of about 6,000 bit/s. But 6,000 bit/s is much faster than 100 bit/s with Sigfox or 297 bit/s with LoRaWAN. The capacity of a block in NB-IoT is not infinite, but it is many times greater than in the unlicensed band. NB-IoT has no limitation in the data volume per time because it is running in the licensed band. NB-IoT is mostly limited by the network operator.

22

6.5.4 Selection of modules for the test 6.5.5 Sigfox Module Selection When selecting the Sigfox module in September 2020 for our tests, it was noticeable that on the Sigfox website there was no single radio module with a radio certification in all regions. There were radio modules that supported the frequencies in the different regions, but these modules lacked certification. An IoT device based on a single Sigfox module with pre-certification in the designated Sigfox territories is therefore not possible. This leads to possible chip level testing in multiple regions where Sigfox is present and consequently the cost of certification becomes extremely high. In addition, if there is a mistake in the design of the module, no change can be made and certification cannot be granted. We therefore decided to compare products from Wisol (today Seong Ji) because they offer radio modules that are pre-certified in the necessary regions and these modules have a low price.

6.5.6 LoRaWAN module Selection After consultation with Semtech before the project start on 1 September 2020, there was still no radio module available based on the new LoRaWAN chip set . Radio modules with certification in all frequency ranges for LoRaWAN could not be found either. We therefore decided to use radio modules from Acsip, because they offer radio modules that are pincompatible for different regions.

6.5.7 NB-IoT Module Selection For the NB-IoT radio module, we have chosen radio modules from Quectel because the manufacturer can supply the widest range of radio modules and utilise different chipset manufacturers. Furthermore, we receive direct support from Quectel.

7.

Link budget and energy consumption

7.1 Determining the maximum energy consumption for SIGFOX Since the Sigfox radio protocol is completely static, it is easy to determine the energy consumption. A Sigfox module in Europe sends each message three times on three different radio channels with approximately 2 seconds per message. This then results in 6 seconds of transmission time for 12 bytes of payload. Since a Sigfox module does not check whether there is a gateway nearby at all, one only has to transmit the 12 bytes with AT commands, send and measure the current. The Sigfox module selected by us can be operated with 2.4 V. Two alkaline AA cells with 1.2 V nominal voltage are sufficient to operate the radio module.

23

7.2 Determining the maximum energy consumption for LoRaWAN The EU LoRaWAN protocol is dissimilar to that of the USA. We therefore limited our measurements to the European LoRaWAN with 14 dBm transmit power and spreading factor 12. We did not use a connection to the LoRaWAN gateway for the measurements, but operate the LoRa module in peer to peer mode. We emulate the payload of 12 bytes and 13 at header by sending 25 bytes and setting the known parameters of the protocol. For spreading factor 11 and 12, the low data bit must be set. LoRaWAN payload, spread factor (SF) and packet size in Europe: • • •

51 bytes for the slowest data rates, SF10, SF11 and SF12 on 125 KHz 115 bytes for SF9 on 125 KHz 222 bytes for faster rates, SF7 and SF8 on 125 KHz

TTN10 in the Netherlands operates a free LoRaWAN with an extreme limit of 30 seconds uplink per day and 10 acknowledgements or downlink per day. In the USA, the maximum packet size is limited to 11 bytes of11 payload with a spreading factor of ten. In the USA, the radio channel may only be occupied for no more than 400 ms. With a 12-byte payload, this 400 ms would be exceeded.

10

Vgl. van Boven, Arjan: Limitations: data rate, packet size, 30 seconds uplink and 10 messages downlink per day Fair Access Policy [guidelines], in: The Things Network, 26.07.2020, [online] https://www.thethingsnetwork.org/forum/t/limitations-data-rate-packet-size-30-seconds-uplink-and-10messages-downlink-per-day-fair-access-policy-guidelines/1300 [06.02.2021]. 11 van Bentem, Arjan: Airtime calculator for LoRaWAN, in: Github.io, 06.09.2020, [online] https://avbentem.github.io/airtime-calculator/ttn/eu868 [23.01.2021].

24

Figure 10 LoRa Modem Calculator, EU: 25 bytes with optimiser on = 1483 ms

Figure 11 LoRa Modem Calculator, EU: 51 bytes with optimiser on = 2793 ms

Figure 12 LoRa Modem Calculator, US: 11 bytes limitation

7.3 Determining the maximum energy consumption for NBIoT If you want to determine the energy consumption of NB-IoT statically in a laboratory, then you need a rather expensive NB-IoT measurement station. So that our measurement series can be replicated by any developer, we chose an inexpensive approach. We reduced the received signal at the antenna with attenuators until the desired path loss of 134, 144 and 154 dB was achieved. For this test setup to work, the test object with the attenuators must be placed in a shielded box. Without this shielding, the test boards would receive a signal via other paths and thus falsify the measurement. The selected NB-IoT modules all have AT commands to read out RSSI and RSRP.

7.4 Determining the maximum energy consumption for SMS All the LPWAN technologies mentioned require less energy than an SMS. Therefore, we added an SMS with 12 bytes and 140 bytes payload at GSM 900 with 33 dBm transmission power to the measurement series.

25

8.

Range with LPWAN

Ranges of LPWAN technologies are determined by the maximum link budget, the transmission frequency, the height of the base station antenna and the height of the antenna at the sensor node. We have determined the maximum link budget in the previous sections and used it in the table below. Furthermore, we have reduced the absolute value by 20 dB attenuation for the first wall and an additional 8 dB for fading. With these parameters for the net budget we have determined the largest possible ranges for an antenna mast with 30 m and 60 m height with an online tool. Everyone is welcome to "play" with a lower path loss for the first wall or with other heights for the antenna to see how this affects the results, however, there is no evidence of the repeatedly quoted range of 50 km or more. Gross link budget (dB) Max. LB LoRaWAN EU and NB-IoT CL0 NB-IoT CL1 Max. LB Sigfox EU UL Max. LB Sigfox EU UL NB-IoT CL2

144 dB

Net link budget = gross - (20 dB first wall + 8 dB Fading) (dB) 116 dB

Hata propagation calculation, mast 30 m, mobile 1 m, distance (m) 490 m

Hata propagation calculation, mast 60 m, mobile 1 m, distance (m) 600 m

154 dB 158 dB

126 dB 130 dB

900 m 1190 m

1200 m 1610 m

163 dB

135 dB

1650 m

2300 m

164 dB

136 dB

1700 m

2400 m

Table 3 Range Indoor NB-IoT, LoRaWAN, Sigfox calculated with Candy Tools 12

Figure 13 Range calculation with Okumura-Hata model

The example in figure 10 shows a 30 m high mast and assumed attenuation in a city (Area option 4). The values for NB-IoT CE1 (154 dB link budget) with 20 dB loss in the first wall 12

Cf. Gütter, Dietbert: CANDY - Tools - Outdoor wave propagation models, in: http://www.guetter-web.de/, [online] http://www.guetter-web.de/mini-tools/candy-prop-outdoor.htm [03.02.2021].

. 26

and 8 dB for fading were used. So in total a path loss of 28 dB. The table of results clearly shows that 10 dB more in link budget leads to a doubling of the range. From 600 m at a 144 dB link budget the range is doubled to 1200 m when link budget increases to 154 dB. Increasing the budget to 164 dB, achieves a range of 2400 m. Since LoRaWAN’s link budget is limited to between 141-146 dB, the range is 50 % worse compared to NB-IoT with its 154 dB link budget. Since the area of a circle is calculated from where r in this case is the range, this means that four times more antenna sites are needed in a city to provide indoor coverage with LoRaWAN than is the case with NB-IoT.

9.

Range of a private LoRaWAN

Figure 14 Municipality of Kirchheim with its 12 villages. Inhabitants per village marked in blue circles. Source LPWAN Cookbook

To further illustrate the task of creating a private LPWAN13 and to provide a practical example, the main village of Kirchheim was studied along with the 11 other villages in the municipality. The blue outlined ovals represent villages with the number of inhabitants recorded in 2016. Kirchheim, with 1,853 inhabitants, is spread over two valleys. Some of the hamlets have less than 80 inhabitants and four are located at the end of a small side valley. The main valley is marked with a solid red line. The side valleys are marked with dashed red lines. Some of these have further small branches. The first LPWAN gateway was installed on the slope above Kirchheim on the roof of a two13

Vgl. Naumann, Harald: LPWAN Cookbook: How to develop LPWAN technology - real products, real networks, 1. Aufl., 2021, [online] https://www.linkedin.com/pulse/lpwan-cookbook-table-content-haraldnaumann/.

27

storey residential building. We used a mobile installation for testing purposes with a portable LPWAN gateway. This arrangement ensures that we can test the same antenna system at different locations in the community. From the chimney of the house we have a view into the valley of the river Aula and the village. Since Kirchheim branches out into a side valley, two gateways are necessary.

Figure 15 Simulation of the first LPWAN gateway in the main village of Kirchheim

Figure 16 Simulated network coverage of the municipality of Kirchheim with 12 LoRaWAN gateways

In the graphic above, the purple markers show the 12 locations for the LPWAN gateways. To increase contrast, the map in the background was selected with grey tones without contour lines. Areas marked in green indicate regions for indoor coverage. Areas marked in red indicate regions of outdoor coverage. The opacity of the green and red colour layers was set to 40 %. When two green areas or more overlap, the result is an area with a dark green hue. As soon as a green area or several green areas or several red areas and a green area overlap, brown tones are created. If three red layers overlap with one green layer, then the eye may no longer recognise the minimal brown due to the mixing of the colours. 28

These twelve villages are distributed over approximately 50 km². If we divide the 50 km² by twelve, this results in an average of approximately 4.2 km² per LoRaWAN gateway. 4.2 km² / Pi and taking the square root gives 1.15 km radius per gateway. Where did the 50 km or 150 km range go?

10. Analysis of the University Offenburg study The slide below is from a presentation by Offenburg University of Applied Sciences14 at the Wireless Congress 2019 in Munich. These four illustrations clearly show that NB-IoT is clearly superior to LoRaWAN. The top left graph, like many others on the internet, shows that LoRaWAN can only receive down to about -120 dBm. The upper right curve shows Signal to Noise Ration (SNR) versus distance. You can see about -18 dB for LoRaWAN at 9 km and about 30 dB SNR for NB-IoT. Sigfox shows approximately 20 dB SNR. The most interesting graph, however, is the third one. At distance of 9 km and maximum path loss, 80% packet loss was documented. Even at 6 km, the packet loss is still 40 %. Similarly high packet loss is found in other reports on LoRaWAN. For this reason, we assume two repetitions of the telegrams corresponding to Sigfox for LoRaWAN in our test series.

Figure 17 Presentation at the Wireless Congress 2019 in Munich, Field tests - RF coverage & signal quality measurements, slide 36

14

Vgl. Sikora, Axel: Field tests - RF coverage & signal quality measurements, Folie 36, in: Wireless Congress 2019, 23.10.2019, [online] https://electronica.de/de/rahmenprogramm/electronicaconferences/wireless-congress/ [30.10.2019].

29

11. Analysis of the study by the University of Singapore LoRaWAN compensates for path loss by changing the Spread Factor (SF). In the study by the University of Singapore15, it can be seen that at a distance of 1 km, the packet loss is between approximately 15 % and 90 % depending on how SF changes between SF 7-12. LoRaWAN can readjust the spread factor at a short distance from the gateway starting from SF7. At a distance of 2 km, an adjustment of the spreading factor is also still recognisable. However, the packet loss with SF12 is already approximately 25 %. More than SF12 is not possible with LoRaWAN. At a distance of 3 km, the packet loss at maximum SF12 is already 55 %. SF 7-9 exhibit an extremely high packet loss at 3 km. If you now double the distance from 3 km to 6 km, you end up with a packet loss of about 95 %. So as soon as a LoRaWAN module has to be operated on SF12, no further adjustment (change of spreading factor) is possible and greater distance automatically ends in packet loss. However, each repetition of a packet also means a multiplication of the energy consumption. Every unnecessary repetition also means that a packet can collide with another packet from another node. Constant unnecessary repetition of packets also leads to other nodes in the licence-free band being disturbed or blocked by the packets of a LoRaWAN network. Increasing the Spread Factor slows down the data rate AND uses more energy from the battery to send a given message as a result. So it’s not just a range limit resulting in more base stations, if you push LoRaWAN performance you end up with remote devices that use up their batteries much faster. On top of that you have repeated messages due to packet loss which makes the problem even worse.

15

Vgl. LIANDO, JANSEN C./Amalinda Gamage/Agustinus W. Tengourtius/Mo Li: Known and Unknown Facts of LoRa: Experiences from a Large-scale Measurement Study, in: Github, 01.02.2019, [online] https://jansencl.github.io/publication/2019-02-01_TOSN-2019 [06.02.2021].

30

Figure 18 Package loss of 15% to 95% packet for LoRaWAN with SF 12

31

12. Testing the range of LoRaWAN in Gleiberg This test of the range of LoRaWAN took place in Gleiberg in Hesse. Gleiberg is a low mountain range surrounded by mountains. The two graphs below show that a signal between a LoRaWAN gateway and a LoRaWAN node can travel up a mountain. The signal is refracted and reflected at the top of a mountain or a building on the mountain. The green or turquoise coloured areas in the terrain section show the growth of trees or bushes. The software for simulating the radio link shows us that in the two terrain sections a line of sight is impossible. With LoRaWAN, a connection is still possible. This characteristic becomes a disadvantage when a LoRaWAN network is set up in Gleiberg and reaches other regions due to the overreach of the nodes caused by refraction by the mountains. If you take a closer look at the range test, the SNR is several times in a negative range and shows that LoRaWAN can receive signals in noise. The test also proves that the limit for reception is reached at approximately -121 dBm. The most extreme value is -121 dBm with a SNR of -17.2 dB. This value of -121 dBm is 16 dB away from the ideal value of -137 dBm in the laboratory. The packet loss was not analysed and evaluated in this range test. This test underlines once again that the maximum link budget of LoRaWAN is approximately 141-145 dB. The 165 dB quoted in many places is not possible in the real world with real noise.

Figure 19 Test of range and max. link budget with LoRaWAN EU with SF12, -121 dBm with -17.2 dB SNR

32

Figure 17 Test of range and max. link budget with LoRaWAN EU SF12, -120 dBm with -8.2 dB SNR

Comparison of energy consumption Sigfox, LoRaWAN, NB-IoT At first glance, all LPWAN technologies require less energy than an SMS. An SMS can transmit 140 characters with 1 byte each. With Sigfox, only 12 bytes fit into one message. SMS can be used with and without acknowledgement. Sigfox can only send 4 receipts per day. If more than 12 bytes are needed, an SMS sometimes requires less energy than with Sigfox because Sigfox will require more than one (effectively n*3) transmissions. In addition, with SMS the energy is readjusted depending on the link budget and with Sigfox the transmission power is static. A deeper comparison with SMS and measurement series with 12, 24, 64 and 512 byte message payloads is not part of the report.

Figure 20 SMS with 33 dBm transmission power

33

12.1 Uplink 12 bytes in comparison (calculated)

Figure 21 Energy consumption 12 bytes with NB-IoT, LoRaWAN and Sigfox - calculated, not measured

The energy consumption for 12 bytes was calculated for this graph and will be verified with measurement series in the next section. For the energy consumption with the Sigfox module, the 2 seconds for one telegram with the power consumption at 14 dBm according to the data sheet is multiplied by the nominal voltage of 2.4 V for two alkaline AA cells. This amount of power is then multiplied by three because the Sigfox protocol transmits each packet three times on three different channels to minimise packet loss. If one uses a lithium cell with 3.6 V instead of two AA cells, the energy consumption increases by 50 %. The energy required is therefore 1000 mWs for the best case, and 1000 mWs then become 1500 mWs for the lithium battery. The energy consumption calculation for the Sigfox module is simple. For LoRaWAN, best-case calculations were also made with 2.4 volts. The measurements were then carried out with 3.3 V because the selected LoRa module does not support 2.4 V. The current and the required time can be calculated with the LoRa calculator. 13 byte header and 12 byte payload result in 25 byte gross. For the 144 dB link budget, one must select SF12, 125 KHz bandwidth and "Low-Data". If you are unsure, use the online tool mentioned at TTN. This calculates correctly and also masters the LoRaWAN settings for 34

other regions (such as the USA, for example). In contrast to Sigfox and NB-IoT, LoRaWAN does not inherently support repetition, it leaves the retry to the next higher protocol layer. NB-IoT, on the other hand, repeats in the core of the protocol even at high field strength. To make the three LPWAN protocols comparable, LoRaWAN was calculated with one and two repetitions. We "invented" the spreading factor 13 to show that the energy consumption doubles or halves with each jump in the spreading factor. We operated the NB-IoT modules at 3.6 volts, although one of the three modules can also be operated at 2.4 volts. Considered over 134 dB to 154 dB link budget, Sigfox loses in the comparison because this protocol does not readjust. LoRaWAN can statically readjust and adapt to the path loss. We negotiate the spreading factor and thus the sensitivity once when registering with the Sigfox gateway. If a different spreading factor with LoRaWAN would be necessary due to a change in the noise in the channel or object in the radio field, then it is not changed and the packets are lost. There is no constant connection, but only a one-time assignment of the spreading factor. Therefore, the highest possible spreading factor is also recommended for moving objects. With NB-IoT, this is completely dynamic. After the first login, the repetition rate and the channel width are dynamically adjusted in the existing connection when sending again. More or less data will be repeated. To calculate the expected energy consumption, an Excel file from the manufacturer Quectel was used for the BC95 module. The file statically calculates from 144, 154 and 164 dB link budget. The 164 dB link budget was not considered at all in this discussion. LoRaWAN is already at its limit at 144 dB budget. With Sigfox, the downlink limit is approximately 158 dB. We was able to change all important parameters in of the NB-IoT protocol and parameters like messages per day and number of bytes. The number of messages per day and the number of bytes are easy to understand. The more messages per day, the higher the energy consumption. The more bytes per payload, the higher the energy consumption. With the many other parameters and timers, it is difficult to set them correctly without experience. We were able to change the eRDX cycle, RRC Release, DRX period, paging timer, timer T3324, timer T3412 and TAU.

35

Figure 22 Data volume at 144 dB link budget with LoRaWAN EU, Sigfox and NB-IoT

The calculations were made with NB-IoT-Non-IP and the measurements later with UDP-IP. The UDP header on IPv4 causes the amount of data to increase from 12 bytes to 40 bytes. A change from 12 bytes to 72 bytes with NB-IoT-Non-IP in the calculation causes the energy consumption to increase from 350 mWs to 360 mWs. 60 bytes more payload causes 10 mWs more energy consumption. 10 mWs / 60 results in 167 uWs per byte. Some NB-IoT developers aim for MQTT-SN on NB-IoT-Non-IP because this saves about 48 bytes. One saves about 8 mWs in relation to 350 mWs. 8 mWs / 350 mWs x 100 % = 2.3 % saving. If 64 or even 512 bytes are transmitted, the saving is reduced to 1 %. However, 1 % or 2.3 % cannot be measured at all in the subsequent series of measurements in the real NB-IoT network. A good antenna design with a 3 dB system gain gives much more of an advantage than the 2.3 % gained from MQTT-SN with NB-IoT-NonIP.

36

12.2 Uplink 64 (72) bytes comparison

Figure 23 Energy consumption 72 bytes with NB-IoT, LoRaWAN and Sigfox - calculated, not measured

The calculation for 72 bytes was done before the final request for 64 bytes and was not adjusted in the calculation. Sigfox totally loses ground because the 72 bytes have to be distributed over 6 messages. In addition, one byte or half byte of the 12 bytes must be used for the message counter and the message ID. Due to the duty cycle restrictions and the small overhead, the last packet with the remainder of the 72 bytes of user data may only be sent after 60 minutes of the first message. There are only 4 receipts per day with Sigfox. Sigfox is totally unsuitable for small quantities of only 72 bytes or more than 4 receipts per day. For LoRaWAN with SF11 and SF12, only 51 bytes are available. The LoRaWAN module must therefore send 2 messages with 2 x headers. But this happens according to the principle of hope. The LoRaWAN gateway has only 360 seconds per hour to acknowledge all messages. 512 bytes with LoRaWAN takes over an hour because more than 10 packets are necessary. Acknowledgement operation is not possible due to duty cycle restrictions for a high or even a medium number of participants. The Quectel Excel file includes NB-IoT NB1 and therefore the maximum payload of NB1. 37

For NB2, the maximum physical payload is larger. However, since 64 bytes is significantly smaller than the physical payload, NB-IoT does not need to send another packet.

12.3 Uplink 512 bytes comparison (calculated) 512 bytes was not even considered in advance with NB-IoT because the 1000 bit (125 byte) physical payload results in 4.1 packets within the logical connection. Once the connection is established, a payload of 12 bytes or 125 bytes does not make much difference, from 125 bytes to 512 bytes, hardly any more energy is required. Most of the energy is needed to establish the connection. If several packets of 125 bytes are needed, the Excel file takes this into account. The result can be found in the measurement series.

12.4 Expected energy consumption of NB-IoT, LoRaWAN and Sigfox

Figure 24 Performance of NB-IoT, LoRaWAN and Sigfox over time

The performance over time in the photo montage shows what we expect in the subsequent measurements. The power curve for NB-IoT is not linear and splits into different phases of communication. After switching from PSM mode, the microcontroller in the NB-IoT module must first power up, turn on the receiver, find the base station, establish a connection and then send the payload. Since NB-IoT transmits at 27 Kbit/s with a 144 dB link budget, logging on and sending the payload is expected to happen in a very short time. In comparison, LoRaWAN at 297 bits/s and Sigfox at 100 bits/s will take much more time. But energy is power x time, therefore due to the slow modulation with the two LPWAN technologies in the licence-free band, a higher energy consumption is to be expected than 38

with NB-IoT.

13. Measurement of energy consumption No special expensive equipment was used for the measurement setup. The measuring device for the dynamic currents in NB-IoT modules should have a sufficient sampling rate to keep the measurement error small. The sampling rate should be chosen so that a short NB-IoT burst is sampled often during transmission. At 134 and 144 dB link budget, the bursts are extremely short. This range should be taken as a benchmark. We have chosen 2 ms. With Sigfox and LoRaWAN, one memory oscilloscope is sufficient. 12 bytes with Sigfox lead to 2.08 seconds and for LoRaWAN EU SF12 to 1.483 seconds channel occupation time. The current curve on the oscilloscope is a simple rectangular curve. With 2 ms sampling rate and 2008 ms or 1483 ms, the error is about 0.1 %. If the spreading factor has to be changed by one digit or a telegram has to be repeated, then the error in consumption is 100 %. The shielding box from Willtek was modified with sealing tape to increase the attenuation to approximately 81 dB. Even the 81 dB attenuation was not enough to force the NB-IoT modules to a 134 to 154 dB link budget. In the end, these measurements were carried out in the basement to gain the necessary additional attenuation. Measuring devices / software used • Tektronik 2004 Digital oscilloscope • Keithley 2100 Digital Multimeter • Keithley DMM6500 Multimeter • Willtek 64 dB Shielding box, approximately 81dB shielding attenuation was achieved by modifications to the sealing strips • Laboratory power supply unit • Lenovo Laptop • Microsoft Excel • Quectel QCOM V1.6. Terminal software Hardware used • NB-IoT Measuring board with Quectel BG95-2 BG95M2LAR02A04_01.003.003 • NB-IoT Measuring board with Quectel BC66 BC66NAR01A03 • NB-IoT Measuring board with Quectel BC68 BC68JAR01A07 • LoRaWAN Measuring board with Asip S76S • Sigfox Measuring board Seong Ji WSSFM10R1AT • NB-IoT-enabled SIM card from Deutsche Telekom • NB-IoT base station in the Lübeck area with Cell-ID 3740 in band 8 • Antenova Wideband Antenna SREL036-10P • Radiall type SMA attenuators 2 dB, 4 dB, 6 dB, 10 dB, 20 dB • 50 Ohm dummy load 39

The measurement boards with the five radio modules were specially developed for the measurement series on a board the size of a credit card so that all modules have the same physical conditions. The PCB antennas in the boards are designed to cover the required frequency bands. For this series of measurements, the internal PCB antennas are unnecessary, but they are needed for other series of measurements planned for the future. The PCBs have a minimal assembly to avoid unwanted leakage currents. For this reason, no evaluation kits from the three manufacturers were used. All modules therefore have the same requirements. The test setup can therefore be easily reproduced in, for example, Australia and the USA. The firmware of the Quectel BG95-M2 BG95M2LAR02A04_01.003.003 was valid at the start of the project. In the meantime there is firmware release BG95M2LAR02A04_01.003.004. With this later version the measurement result could be still better. The same applies to the Quectel BC66 and BC68. Only with the Sigfox and LoraWAN module has there been no change for a long time. The Sigfox protocol is now 10 years old and LoRaWAN has not received any change that optimises energy consumption since version 1.0 in January 2015.

Figure 25 Keithley DMM6500 Multi meter - Front panel- Credit: TEKTRONIX INC

40

Figure 26 Keithley DMM6500 Multi meter - 10 pA to 10 Ampere - Credit: TEKTRONIX INC

Figure 27 Measurement set up with DUT in the basement.

41

Figure 28 Measurement set up with Quectel BC66, BC68 and BG95-M2

Point 1 and 2: The digital multi meter DMM6500 was configured before the measurements so that it takes a measurement every 3 us and adds up 12 measurements in the mean value. The output was in CSV format with measurements values every 36 us. Point 3: The measured values every 36 us were summed up again by script every 40 ms. One measured value in CSV format was created from 1111 measured values. Point 4: Despite the summary in measured values every 40 ms, at 2 seconds for a cycle from PSM mode into transmission mode back into PSM mode, 50000 measured values still come together. With a high link budget, the cycle can also be 10 seconds long. This then results in 25,000 values. The energy consumption of LoRaWAN Sigfox was very easy to measure. A memory oscilloscope is enough. Current x voltage x time gives the energy consumption in mWs. With NB-IoT, the energy consumption is divided over different cycles with different times and different currents.

13.1 Measurement of SIGFOX module Wisol (Seong Ji) At 2.20 Euros each, the Seong Ji WSSFM10R1AT is probably the cheapest Sigfox module on the market. It cannot be used worldwide. There are derivatives for other regions. A worldwide product in all Sigfox regions is not possible on a pre-certified Sigfox module because, based on our research, there is no module for this as of September 2020. The measurement with the Seong Ji WSSFM10R1AT was carried out with standard commands for sending 12 bytes. The operating voltage was set to 3.3 volts because this is common for microcontrollers. Message receipt was not checked since Sigfox is a connection less protocol, one simply sends and hope that a packet arrives. This means that a Sigfox module sends without knowing whether a Sigfox gateway is reachable. 42

Therefore, the expected 1000 mWs for 12 bytes in the calculation were confirmed by measurement to be 980 mWs. To make the measurement, one simply connects a 50-ohm dummy load instead of the antenna. The figures for 24, 64, 72 and 512 bytes can be easily derived mathematically from the measurement for 12 bytes. 64 bytes with acknowledgement are impossible with Sigfox because the number of acknowledgements is limited to 4 per day in the downlink.

13.2 Measurement of LoRaWAN Module Acsip S76 S The Acsip module S76S is a low-cost LoRaWAN module. It cannot be used worldwide. The operating voltage had to be 3.3 volts because of the requirements of the module’s MCU. The Acsip module is also available with LoRaWAN US with FCC approval. Since LoRaWAN is a connection less protocol that primarily hopes for the message to arrive, you don't need a gateway to measure the energy consumption. The LoRaWAN EU protocol consists of 13 byte headers and a maximum of 51 byte payload. The Acsip S76S can be used with the LoRaWAN protocol with LoRa peer-to-peer. 13 byte header with 12 byte payload results in a 25 byte payload in peer-to-peer mode. 24 bytes of payload can still be transmitted with one telegram. At 64 bytes, the payload must be divided into two telegrams. 512 bytes are further divided into many telegrams and an acknowledgement becomes impossible with just a few radio nodes. The figures for 24, 64 and 512 bytes can be easily derived mathematically from the measurement of the 12 byte payload. The results of the theoretical calculation and the actual lab measurements are similar. The calculation is even a little more accurate because it was performed at a low operating voltage.

43

13.3 Measurement of NB-IoT module Quectel BG95 , BC66 and BC68 compared to LoRaWAN Acsip and Sigfox Seong Ji Circuit boards for measurement

Quectel BC66 Quectel BC68 NB-IoT NB-IoT PCB antenna PCB antenna

Quectel BG95-M2 NB-IoT, LTE-M, GPRS, GNSS PCB antenna

Acsip S76S LoRaWAN EU PCB antenna

Seong Ji SSFM10R1AT Sigfox EU PCB antenna

Figure 29 Five LPWAN test PCBs

All boards have been designed to be the same size for all radio modules and LPWAN protocols. The dimensions are 55 mm x 85 mm, just like a credit card. The circuit boards were designed so that they can be used for further series of measurements to compare other parameters. The four-pole connector on the board is used to connect a control cable with converter from UART to USB. The two-pole connector on the left side of the board is used to connect the laboratory power supply or battery. The two yellow plug connectors between the plug for the operating voltage and the control cable are used to connect the measuring device for measuring current consumption. The NB-IoT coverage at our site is so good that the 80 dB shielding box used was not sufficient to force the 134, 144 and 154 link budget. The radio wave comes through the external antenna (the PCB antennas are capped and not in use), is attenuated with attenuators, and more radio energy comes through the shield box to the PCB. This unwanted energy also reaches the NB-IoT module. To achieve the target of 134 to 154 dB LB, the measurements were moved to the basement. The Gerber files, parts list, circuit diagrams, measurement of the antenna (return loss / radiation pattern in 3 axes) are available free of charge on request. Anyone interested can thus reproduce the measurements with minimal effort.

44

13.3.1 Measurement series 12, 24, 64, 512 bytes at 134 to 154 dB link budget 12 byte uplink - 134 to 154 dB link budget

Figure 30 Energy consumption in average during uplink of 12 bytes with 134, 144 and 154 dB link budget

To make the LPWAN technologies comparable, 12 bytes was chosen as the lowest payload because Sigfox can only transmit a maximum of 12 bytes in the uplink. Since a Sigfox module always repeats a packet twice and NB-IoT repeats once even at high field strength with 134 dB link budget, LoRaWAN with repetition once and twice was also applied. The fact that 90 % packet loss is also achieved with LoRaWAN at extremes was not taken into account here. Without repetition, Sigfox and NB-IoT cannot achieve the desired result of as little packet loss as possible. The benchmark is 3530 mWs for an SMS at 144 dB link budget. NB-IoT and LoRaWAN are better at 144 dB. The 3530 mWS includes waking up, logging into the GSM network and transmitting 12 bytes by SMS. At 144 dB link budget, NB-IoT is better than LoRaWAN and Sigfox. At 154 dB, Sigfox is better than NB-IoT. LoRaWAN EU has a packet loss of 100 % at 154 dB. At 134 dB, Sigfox even loses to the age-old SMS. LoRaWAN EU is better than NB-IoT at 134 dB link budget because the statistical control of LoRaWAN fits better in our steady-state consideration due to the simplicity of the protocol.

45

24 byte uplink - 134 to 154 dB link budget

Figure 31 Energy consumption in average during uplink of 12 bytes with 134, 144 and 154 dB link budget

Sigfox is easy to explain. Sigfox must send two messages to achieve the 24 bytes. For every 12 bytes of payload, there are 12 bytes of headers. If you increase the payload from 12 bytes to 24, the amount of energy required also increases. Even at 24 byte payload with 154 dB LB, Sigfox is better than NB-IoT or LoRaWAN EU. LoRa cannot do 154 dB link budget and ends up with 100% packet loss. At 144 dB and 134 dB LB, NB-IoT and LoRaWAN EU outperform Sigfox. The 10-year-old, very simple proprietary protocol of Sigfox is beaten by standardised NB-IoT and LoRaWAN. Even the ancient SMS beats the old Sigfox at 134 dB link budget.

46

64 byte uplink - 134 to 154 dB link budget

Figure 32 Energy consumption in average during uplink of 64 bytes with 134, 144 and 154 dB link budget

At 64 bytes, Sigfox loses to SMS at 144 dB LB. At 154 dB LB, Sigfox is similar to NB-IoT. However, NB-IoT constantly readjusts. At 134 dB link budget, even SMS is better than LoRaWAN with one repetition. If one does not plan a repetition, LoRaWAN immediately looks better. Sigfox recognised the risk of packet loss 10 years ago and integrated this into the protocol. However, a radio cell does not only consist of the extreme link budget 154 dB, but reaches far below to a 134 dB link budget. At 64 bytes and 134 dB link budget, NBIoT pulls ahead of LoRaWAN and Sigfox.

47

512 byte uplink - 134 to 154 dB link budget

Figure 33 Energy consumption in average during uplink of 512 bytes with 134, 144 and 154 dB link budget

512 bytes is only possible with Sigfox and LoRaWAN with extremely high energy expenditure. Due to the 1 % duty cycle in the uplink, 512 bytes must be distributed over a long period of time. Acknowledgement operation is not possible because the gateway is limited to 10 % duty cycle for all nodes. With 1,000 nodes and thus 360 ms time for the downlink per node, an acknowledgement with 144 dB link budget already requires more than 360 ms. Acknowledgements are impossible at 512 bytes with LoRaWAN and Sigfox. With 64 bytes, however, this is also impossible with 1,000 bytes. If you look closely at NB-IoT, you find that 154 link budget with 12 bytes to 512 bytes requires 5,393 to 5,606 mWs. Whether the payload is 12 bytes or 512 bytes it does not make much difference. This continues with a 134 or 144 dB link budget.

48

24, 64 and 512 bytes in tables

Figure 34 Energy consumption with 12, 64 and 512 Bytes at 134, 144 and 154 dB link budget

The tables above show the measurement series with two of the three selected Quectel NB-IoT modules . In the table, the energy consumption was estimated at 134,144 and 154 dB link budget with 24,64 and 512 bytes. The link budget is obtained assuming that -104, -114 and -124 dBm RSRP correspond to the desired link budget. The limits for the RSRP were taken from the 3GPP documents and the Excel file to calculate the Quectel energy consumption. RSRP -104 dBm -114 dBm -124 dBm -134 dBm

Link budget 134 dB 144 dB 154 dB 164 dB

Table 4 RSRP related to the link budget

The location for all measurements was always the same. The required attenuation was achieved by attenuators behind the antenna and in front of the test PCB. The PCB with the NB-IoT module was brought to the basement in a shielding box with 80 dB shielding attenuation because the signal levels in the lab were still too high despite the shielding box. The NB-IoT signal passed through the shielding box to the test PCB in the lab, and the board acted as an antenna. In contrast to Sigfox and LoRaWAN, the NB-IoT protocol behaves very dynamically. This dynamic behaviour leads to readjustment even during a connection or data transmission. If the noise increases, the minimum receiving level deteriorates. The NB-IoT protocol immediately readjusts so that the connection is maintained. Each of the ten measurements consists of multiple measurements, and the 49

minimum and maximum signal level per measurement are shown in the table. The Excel file for the calculation only details a 144, 154 and 164 dB link budget. Therefore we estimated the 134 dB link budget calculation and confirmed that estimate in our series of measurements.

14. Conclusion The prediction of the calculated amount of energy for 12, 24, 64 and 512 bytes with LoRaWAN and Sigfox was confirmed. The calculation for NB-IoT with a very complex Excel spreadsheet was also proven. If you compare the LPWAN technologies NB-IoT, LoRaWAN and Sigfox with GSM 900 and SMS using 33 dBm transmission power, all three are usually better than SMS. However, when a GSM module is placed closer to the base station, the transmission power of the module decreases when transmitting. This can reach the point where a GSM module with SMS ends up using less power than a Sigfox module. Sigfox modules use a completely static radio protocol and do not readjust the energy consumption. As soon as more than 12 bytes of payload are sent, SMS is also better in terms of energy consumption, because 140 bytes can be sent in an SMS. The Sigfox module must divide the 140 bytes into 12 packets. A packet with the Sigfox EU protocol occupies the channel for 2.06 seconds, and by repeating it twice, 6.18 seconds. Due to the duty cycle of 1 %, one is limited to 36 seconds of transmission time per hour in Europe. The 140 bytes that are possible with a single SMS will therefore take over 2 hours to send. With LoRaWAN it looks better because there it leads to 3 messages with a channel occupation time of 10.7 seconds in total. For LoRaWAN, however, it becomes impossible to acknowledge the telegrams even with only a few nodes per gateway. Acknowledgement operation is impossible for all LPWAN technologies in the 868 MHz band in Europe with a medium number of nodes, due to the 10 % duty cycle. For example the 360 seconds with 1,000 nodes only provides 360 ms transmission time for acknowledgement. However, an acknowledgement with LoRaWAN EU with SF12 or Sigfox requires more than 360 ms channel occupancy time. Analysis of the measurement series on energy consumption At 164 dB link budget, the energy consumption for NB-IoT increases tenfold compared to 154 dB according to the Excel calculation. Measurements in the real NB-IoT network are almost impossible at 164 dB, because connection breaks can be expected there. To increase the link budget by 3 dB, the power must be doubled. 3 dB +3 dB +3 dB plus 1 dB = 10 dB. 2 × 2 × 2 x 1.25 = 10. Since the transmission power is not increased further, the increase per byte takes place via the additional time. Energy = power x time = voltage times current x time. At 164 dB link budget, LoRaWAN and Sigfox no longer work in bidirectional mode. We have successfully tested NB-IoT up to a 154 dB link budget and don't intend to test further using a 164 dB link budget. The maximum link budget of Sigfox is 158 dB in bidirectional operation is close to NB-IoT with Coverage Enhancement Level 1 (CE1) at 154 dB. Sigfox will consume approximately 50

980 mWs. LoRaWAN needs infinite energy because no more message can be evaluated. At 144 dBm link budget, NB-IoT requires approximately 362 mWs and thus significantly less energy than Sigfox. The Sigfox module remains static at approximately 980 mWs seconds in energy consumption. With approximately 410 mWs for 12 bytes with LoRaWAN per transmission, this results in approximately 625 mWs with three transmissions. If one assumes the 80 % packet loss of the Offenburg University paper, five telegrams would be necessary. As soon as the packet size of 51 bytes is exceeded, the energy consumption of LoRaWAN EU increases drastically because 13 bytes of headers are added to the 51 bytes net. 64 bytes gross occupy the channel for a very long time, and thus the probability of a collision increases. The 512 bytes payload can theoretically be reached with multiple telegrams with LoRaWAN. This leads to a further increase in the probability that the 512 bytes can be transmitted without loss. The 2 to 3 calculated messages per 51 bytes may not be sufficient for 512 bytes. Regardless of the high energy consumption with LoRaWAN EU at 144 dB link budget, the protocol is unsuitable for transmitting 512 bytes. Even 64 bytes is difficult because the probability of collision of the two packets increases. At 134 dBm link budget, NB-IoT automatically reduces the transmission power of the module. Theoretically, the energy consumption should improve by a factor of ten with a 10 dB better link budget. In practice, the readjustment of NB-IoT is apparently not good enough. If you take Semtech's LoRaWAN calculator to hand, you will see that when you change from spreading factor (abbreviation SF) 12 to SF 9, LoRaWAN reduces the time needed for a telegram from 1483 ms to about 206 ms. It is not a tenth, but the reduction in energy consumption is better with LoRaWAN at 134 dBm than with NB-IoT. LoRaWAN can readjust up to SF7. After that, there is no more improvement in energy consumption by changing the spreading factor. Due to the not regulated Sigfox protocol the approximately 980 mWs is much worse than LoRaWAN and NB-IoT. If you increase the payload from 12 bytes to 24 bytes, Sigfox immediately loses to NB-IoT. Sigfox can only transmit 12 bytes. A header would have to included in the 12 bytes in order to merge the split telegrams. Since Sigfox only offers 4 messages with 8 bytes in the downlink, a request to repeat messages is only possible to a limited extent. 120 bytes would be spread over 11 messages and take more than 120 minutes. The already poor energy consumption of Sigfox would increase by about eleven times. The situation is similar with LoRaWAN US with 11 byte payload. There, the problem can be mitigated by foregoing approximately 3 dB sensitivity and limiting SF 10 with 11 bytes to SF9 with a maximum of 53 bytes. The 53 bytes in the USA are then similar to the 51 bytes in Europe and lead to split telegrams and possible requests for repetition. With 2794 ms for 51 bytes for the first split telegram and the following telegram plus 1810 ms, the radio channel is occupied for 2794 ms + 2794 ms + 1810 ms = 7398 ms. If the 2 necessary repetitions are added, the channel is occupied with 6 telegrams for approximately 22 seconds. With 1 % duty cycle, 22 sec. x 99 = 2,178 sec. (36 minutes) my further message may now be sent. In addition, a LoRaWAN gateway in Europe is limited to 360 seconds downlink for all participants to request messages. The necessary countermeasure is the reduction to SF 11 or better SF 10. However, SF10 means approximately 6 dB loss in the 51

link budget with LoRaWAN. LPWAN for moving objects Since the maximum link budget for Sigfox is static and is not readjusted, Sigfox can be used well for moving objects. The disadvantage is the consistently high energy consumption due to the lack of regulation of the modulation or TX power level. In principle, LoRaWAN only readjusts the link budget when logging on to the gateway. The procedure is called Adaptive Data Rate (ADR). This leads to a high packet loss with moving objects or to unnecessarily high energy consumption with a fixed high spreading factor. The packet loss is 100 % if, for example, a spreading factor of 10 was negotiated by ADR and a factor of 11 or even 12 is required. As soon as the negotiated spread factor with ADR is too low, 100 % packet loss is inevitable. NB-IoT can be used for moving objects. Due to the lack of handover, an NB-IoT module will keep the connection to the base station as long as possible, even though there may be another base station nearby that could reduce energy consumption. An NB-IoT module will only switch to a new station after the connection is broken. Such a switch means that it searches radio modules for the base station for a short time. Since the transmission frequency is the same, this does not take very long. If the so-called tracking area zone has remained the same, then no complete registration is necessary. A so-called reattach takes place. This takes less time and energy than an attach. The energy consumption in any LPWAN network must not only be considered at one extreme. One must take into account the changing link budget and thus the possibility of dynamic energy consumption. For NB-IoT and LoRaWAN, the network operator can optimise the energy consumption with the density of the base stations. The denser the population of base stations, the smaller the expected average consumption of all participants in the network. NB-IoT is superior to LoRaWAN EU in a network with 144 or 154 dB link budget for coverage in buildings via high masts outside the building. For local LPWAN in a building, LoRaWAN is superior to NB-IoT because it can fine-tune in steps of approximately 3 dB via the spreading factor 7 - 12. This is partly implemented in German LoRaWAN networks. The lack of hand over with a moving NB-IoT device leads to unnecessarily high energy consumption due to base station carryover. The brief interruption of a connection with NBIoT and the reestablishment takes only a few seconds. In contrast, Sigfox is limited by the duty cycle to a cycle of 10 minutes. LoRaWAN EU SF12, with 36 seconds maximum transmission time due to duty cycle and approximately 4.5 seconds (including 2 repetitions) channel occupancy, can only transmit approximately every 8 minutes. LoRaWAN is a good alternative if you need to build your own private LPWAN. If one respects the real outermost link budget or even does without spreading factor 11 and 12 and thus reduces the approximately 141 dB link budget to 135 dB at SF10, then an LPWAN with LoRaWAN with primary uplink similar to Sigfox is feasible. An absolute statement as to which of the three LPWAN protocols is better in terms of 52

energy consumption cannot be made. If we look at the range for the link budget (LB) of 144 to 154 dB, NB-IoT is better than the protocols of LoRaWAN and Sigfox from 12 bytes to 512 bytes. 154 dB link budget is only possible with LoRaWAN EU in the laboratory. Below 134 dB LB, LoRaWAN becomes better than NB-IoT if the number of participants is kept small and the number of messages per hour is limited. Packet loss can be caused by collision with participants in one's own LPWAN network and by other participants in other radio networks (e.g. other LoRaWAN, Sigfox, W-Mbus). The studies mentioned by the University of Applied Sciences in Offenburg, the University in Singapore and the University in Aalborg in Denmark point to the expected loss. The packet loss will be significantly lower with LoRaWAN EU when sensors and gateway are operated in the same building, because other participants outside will be attenuated by 15 to 20 dB at the outer wall. But LPWAN means Long Range Wide Area Network, and a radio network in the building is inverse to "Long Range".

53