NB-IoT (LTE Cat-NB1 / narrow-band IoT) performance evaluation of variability in multiple LTE vendors, UE devices and MNOs

Robinson, Daniel (2020-04)

Thesis (MEng)--Stellenbosch University, 2020.

Thesis

ENGLISH ABSTRACT: Cellular 2G/GPRS is a sun-setting technology worldwide leaving behind a void for wireless low-power widearea-networks (LPWANs) such as LoRaWAN, SigFox and NB-IoT to fill. With NB-IoT on the roadmap towards 5G New Radio (NR), it is a promising contender due to its bidirectionality, power-saving mechanisms and ease of integration with existing equipment, yet there still exists a general uncertainty with regard to adoption. Research shows that most literature on NB-IoT is based on precise mathematical models, analysis or simulations, except for a few empirical performance evaluations which find variability in devices connected to a single network. The study theorizes that networks are responsible for the variation found in metrics and estimations, due to the high underlying complexity of Long-Term Evolution (LTE) architecture on which NB-IoT is based. Thus, the study proposes an empirical investigation using mobile-network operators (MNOs) in South Africa by comparing multiple top LTE vendors including Ericsson and ZTE on MTN’s network, and on Vodacom’s network Huawei and Nokia. Furthermore, similar user equipment (UE) devices such as Ublox and Quectel are used as a control to observe network changes via RF attenuation. A set of telemetry tests are developed to capture various metrics and estimations into datasets for comparison, which include differently sized UDP packet datagrams, cellular operator selection (COPS), extended discontinuous reception (eDRX) and periodic tracking-area-updates (PTAU). Data is measured using an external energy capture device or reported by the UE device for post-processing and analysis in plots, mean distribution tables and boxplots. Metrics such as latency, power efficiency, signal strength, enhanced coverage level (ECL) classes, throughput and data overhead are included, as well as estimates for telemetry interval periodicity and battery longevity. K-means clustering is applied to the datasets to reduce the skewness induced by the increased number of low-latency values during captures to normalize the number of unique features for comparison. Most clearly visible in the tests is how MTN leads Vodacom in NB-IoT performance due to Nokia’s subpar results. Power efficiency and latency metrics show that when connected to Vodacom-Nokia, results can factor up 20 and 10 times worse, respectively. Otherwise, ZTE, Ericsson and Huawei show satisfactory latency under the 10 second 3GPP standard. Although LTE vendors meet the 164 dBm MCL requirement, Vodacom-Nokia has 10 dB less receive sensitivity, with the rest at -130 dBm. Transmit power increases at 10 dBm per RSRP decade until its maximum at 23 dBm, except for Nokia which remains at full power. ECL classes overlap with respect to RSRP, yet partially correlate, which suggests an unknown network factor or hysteresis of a few seconds in the test captures. Nevertheless, Nokia is mostly in ECL class 1, while others are a mix of ECL class 0 and 1. This has an impact on the number of dynamic repetitions of messages between UE devices and cell-tower eNodeBs. Throughput is under 10 kbps, which is half or less than UE device claims by manufacturers. A quarter of datagrams in the telemetry test set show protocol overhead extending over 512 bytes in uplink and 200 bytes in downlink, except for Nokia extending up to 10,000 bytes. Telemetry interval and battery longevity estimates on a 9.36 Wh AA battery suggest that ZTE, Ericsson and Huawei can transmit 16-512 bytes between every 5 to 30 minutes to last at least a year, or hourly to last up to 10 years, however, a device that transmits hourly on the Vodacom-Nokia network will only last 2 months. The study provides recommendations based on these results. Finally, South Africa is ready for mobile network operators to deploy national NB-IoT coverage using ZTE, Ericsson and Huawei, but not using Nokia. With a satisfactory inter-cell tower distance, UE devices avoid having to use dynamic repetitions in higher ECL classes, thus keeping the variability that affects many of the metrics and estimates in the study to a minimum.

AFRIKAANSE OPSOMMING: Sellulêre 2G/GPRS is ’n einde-van-leeftyd tegnologie wat wêreldwyd ’n leemte agterlaat, wat deur draadlose lae-krag-wye-netwerke (LPWAN’s) soos LoRaWAN, SigFox en NB-IoT gevul sal word. NB-IoT se prominensie op die padkaart na 5G New Radio (NR), maak dit ’n belowende aanspraakmaker vanweë die tweerigtingkommunikasie, kragbesparingsmeganismes en die gemak van integrasie met bestaande toerusting, maar daar bestaan steeds ’n algemene onsekerheid oor die aanvaarbaarheid daarvan. Navorsing toon dat die meeste literatuur oor NB-IoT gebaseer is op presiese wiskundige modelle, analise of simulasies, behalwe vir ’n paar empiriese prestasiebeoordelings wat wisselvalligheid vind in toestelle wat aan ’n enkele netwerk gekoppel is. Hierdie studie stel voor dat netwerke verantwoordelik is vir die variasie in statistieke en beramings as gevolg van die hoë onderliggende kompleksiteit van die Long-Term Evolution (LTE) argitektuur waarop NB-IoT gebaseer is. Die studie stel dus ’n empiriese ondersoek in Suid-Afrika voor, wat gebruik maak van mobiele netwerkoperateurs (MNO’s) en deur verskeie top-LTE-verkopers, waaronder Ericsson en ZTE, op MTN se netwerk en op Vodacom se netwerk Huawei en Nokia te vergelyk. Verder word soortgelyke toestelle vir gebruiker-toerusting (UE) soos Ublox en Quectel gebruik om ’n netwerkverandering via RF-demping te waarneem. ’n Stel telemetrie-toetse word ontwikkel om verskillende statistieke en beramings op te stel in datastelle vir vergelyking, wat verskillende grootte UDP-pakkedatagramme, seleksie van sellulêre operateurs (COPS), uitgebreide diskontinue ontvangs (eDRX) en periodieke opdaterings vir opsporing van gebiede (PTAU) insluit. Data word gemeet met behulp van ’n eksterne energie metingstoestel of deur die UE-apparaat gerapporteer vir na-verwerking en ontleding en analises. Maatstawwe soos latensie, drywingseffektiwiteit, seinsterkte, verbeterde dekkingvlakklasse (ECL), deurset data en oorhoofse data is gebruik, sowel as skattings van telemetrie-intervalperiode en batteryleeftyd. K-gemiddelde-groepering word op die datastelle toegepas om die skeefheid wat veroorsaak word deur die verhoogde aantal lae-latenstydwaardes tydens opnames te verminder, om die aantal unieke eienskappe te vergelyk. Die toetse dui duidelik aan aan hoe MTN se NB-IoT beter vaar as Vodacom s, as gevolg van Nokia se ondergeskikte resultate. Kragdoeltreffendheids- en latenstatistieke toon dat die resultate, as dit met Vodacom-Nokia gekoppel is, onderskeidelik 20 en 10 keer erger kan wees. Andersins vertoon ZTE, Ericsson en Huawei bevredigende vertraging onder die 10 sekonde 3GPP-standaard. Alhoewel LTE-verkopers aan die MCLvereiste van 164 dBm voldoen, het Vodacom-Nokia 10 dB minder sensitiwiteit, met die ander op -130 dBm. Transmissiedrwying neem toe met 10 dBm per RSRP dekade tot die maksimum op 23 dBm, behalwe vir Nokia wat op volle krag bly. ECL-klasse oorvleuel ten opsigte van RSRP, maar korreleer tog gedeeltelik, wat dui op ’n onbekende netwerk eienskap of histerese van enkele sekondes in die toetsopnames. Nietemin, is Nokia meestal in ECL-klas 1, terwyl die ander ’n mengsel van ECL-klasse 0 en 1 is. Dit het ’n invloed op die aantal dinamiese herhalings van boodskappe tussen UE-toestelle en eNodeBs. Die deurset is minder as 10 kbps, wat die helfte of minder is as wat UE-toestelle se vervaardigers beweer. ’n Kwart van die diagramme in die telemetrie-toetsstel toon die oorhoofse protokol wat strek oor 512 bytes in oplaaikanaal en 200 bytes in aflaaikanaal, behalwe vir Nokia wat tot 10.000 grepe strek. Telemetrie-interval- en batteryleeftydberamings dui daarop dat ZTE, Ericsson en Huawei 16-512 byte tussen elke 5 tot 30 minute kan oordra met ’n 9.36 Wh AA-battery wat minstens ’n jaar sal hou, of uurlikse transmissie wat tot tien jaar sal duur. Toestel wat uurliks op die Vodacom-Nokia-netwerk uitstuur, sal slegs 2 maande duur. Die studie bied aanbevelings gebaseer op hierdie resultate. Ten slotte, is Suid-Afrika gereed vir mobiele netwerkoperateurs om die nasionale NB-IoT-dekking te gebruik met behulp van ZTE, Ericsson en Huawei, maar nie Nokia nie. Met ’n bevredigende afstand tussen die toring van die sel, vermy UE-toestelle om dinamiese herhalings in hoër ECL-klasse te gebruik, en sodoende word die veranderlikheid wat baie van die statistieke en ramings in die studie beïnvloed tot ’n minimum beperk.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/107854
This item appears in the following collections: