IoT NTN — NB-IoT & eMTC over Satellite

Introduction

IoT-NTN is the 3GPP work that lets the two established cellular-IoT radios — NB-IoT and eMTC (LTE-M) — run over a satellite instead of a terrestrial eNB. Both radios were designed in the LTE era for massive machine-type communication (mMTC): tiny payloads, deep-coverage penetration, and battery lives measured in years. IoT-NTN keeps that design intact and simply changes the access node from a tower to a spacecraft, so the same low-cost, frugal modules keep working far outside terrestrial coverage.

It sits in the device lifecycle at exactly the same point terrestrial NB-IoT/eMTC does — attach, occasional small-data transfer, then back to deep sleep — but across a link that can add hundreds of milliseconds of delay and tens of kHz of Doppler. The study item is TR 36.763; the normative work landed in Rel-17 on the LTE (36-series) air interface and, in most deployments, on the EPS/EPC core the two radios already used on the ground.

It matters because coverage, not throughput, is the binding constraint for the mMTC market. A container tag mid-ocean or a pipeline sensor in the desert has no tower to reach; a delay-tolerant, repetition-heavy narrowband radio rides a satellite comfortably, where a full NR modem would be overkill on power, cost, and complexity.

Why IoT-NTN is needed

Concretely, three needs drive the whole design. First, coverage everywhere: mMTC assets live where terrestrial coverage does not — oceans, deserts, farmland, remote infrastructure — so the network must reach them from space. Second, battery life measured in years: these devices are unattended and often unpowered, so the radio must sleep aggressively and wake only to send a handful of bytes. Third, cost: they ship in the millions, so reusing existing NB-IoT/eMTC chipsets, the LTE air interface, and the EPC ecosystem is far cheaper than porting everything to NR. Satellite fills the coverage gap; the low-power LTE radios keep the devices cheap and long-lived.

NB-IoT and eMTC, lifted to orbit

NB-IoT and eMTC are the two cellular IoT radios 3GPP standardised in the LTE era. NB-IoT is an ultra-narrowband, extremely low-power radio for tiny amounts of data and years of battery life; eMTC (marketed as LTE-M) is a slightly richer LTE-based machine radio with more bandwidth and mobility. IoT-NTN is the work item that lets both operate through a satellite instead of a terrestrial eNB, so the same cheap, frugal devices keep working far outside terrestrial coverage.

The physical channels underneath

Both radios keep their LTE-era channel sets unchanged over satellite — that is the whole point of reuse. NB-IoT carries its downlink on NPBCH (broadcast), NPDCCH (control) and NPDSCH (shared data), and its uplink on NPUSCH (with formats F1 for data and F2 for HARQ/ACK signalling) and NPRACH (random access). Downlink uses a single 15 kHz subcarrier spacing inside the 180 kHz carrier; the uplink can be single-tone at a 3.75 kHz or 15 kHz tone spacing, or multi-tone (3, 6 or 12 tones at 15 kHz). eMTC keeps the standard LTE 15 kHz numerology but confines a transmission to a 1.08 MHz narrowband (6 PRBs), hopping between narrowbands within a wider LTE carrier. Neither adds an NR-style flexible numerology — IoT-NTN inherits LTE timing wholesale.

Same sky, same challenges, same tools

A satellite does not care whether it is carrying NR or NB-IoT — the physics is identical. IoT-NTN faces the same long propagation delay and the same large, fast-changing Doppler shift as NR-NTN, so it reuses the same core ideas rather than inventing new ones.

The device uses its GNSS position together with broadcast satellite information to pre-compensate both timing and frequency before it ever transmits: it advances its uplink timing to cancel most of the round-trip delay (the NTN timing-advance idea, see ntn-timing-advance) and pre-corrects its carrier frequency to cancel most of the Doppler (see ntn-doppler). To do that it needs to know where the satellite is and how the common delay behaves, so IoT-NTN adds new system information carrying satellite ephemeris and a common-TA term — carried in LTE SIBs, the direct analogue of NR's SIB19 (see ntn-sib19). And because the round trip is long, it applies scheduling offsets and timer scaling — the same family of ideas as NR's K_offset and scaled timers (see ntn-timers).

What "pre-compensation" costs an IoT device

The catch is that pre-compensation makes GNSS effectively mandatory in an IoT-NTN device — a cost, in power and BOM, that terrestrial NB-IoT never paid. The UE applies the full timing advance itself (its own UE-specific component plus the broadcast common TA that covers the feeder link and satellite-to-reference-point delay), and it pre-corrects the full Doppler on the service link. The broadcast ephemeris comes with a validity duration: the device may reuse a GNSS fix and the satellite position for as long as they remain accurate, then must refresh before its timing or frequency error drifts outside the receiver's window. Rel-17 leaned on frequent GNSS use; later releases work to relax that so the receiver can stay asleep longer, because on these devices every GNSS acquisition is a measurable dent in battery life.

Where IoT-NTN differs from NR-NTN

The shared physics hides real design differences. IoT-NTN is aimed at a different job, sits on a different core, and uses a very different radio underneath. The table makes the contrast concrete.

A few NB-IoT specifics are worth spelling out because they explain the whole design. NB-IoT lives in a single 180 kHz narrowband. Its uplink can be single-tone (as little as a 3.75 kHz or 15 kHz tone) or multi-tone, trading rate for reach and device simplicity. It leans on heavy repetition for coverage enhancement — sending the same content many times so it decodes deep inside a building or across a satellite link. It is half-duplex (never transmitting and receiving at once), runs at much lower data rates than NR, and — the number interviewers love — has only 2 HARQ processes, against NR's up to 32. eMTC is a little richer (roughly 1.08 MHz, higher rates, better mobility) but shares the same low-power, coverage-first philosophy.

Coverage enhancement and power saving, quantified

Two mechanisms carry most of the weight. Coverage enhancement uses repetition: eMTC defines CE Mode A (light repetition) and CE Mode B (deep coverage, up to hundreds or thousands of repetitions), while NB-IoT organises access into coverage-enhancement levels, each with its own NPRACH resource, so a device in a worse link budget automatically picks a more-repeated preamble. Repetition trades throughput for reach — exactly the right trade over a lossy satellite link. Power saving uses two features: PSM (Power Saving Mode), where the UE stays registered but becomes unreachable and shuts down almost completely between transfers, and eDRX (extended DRX), which stretches the paging cycle so the device wakes to listen only rarely. Together they are what turn "years of battery" from a slogan into a spec.

Why a separate track at all

You might ask why 3GPP didn't just let NR-NTN cover everything. The answer is that IoT devices optimise for the opposite things. A soil sensor or a container tag needs years of battery from a small cell and needs to be heard from places with terrible link budgets — it does not need throughput. The NB-IoT/eMTC design — narrowband, half-duplex, repetition-heavy, delay-tolerant, with deep sleep modes — fits that job far better than a full NR modem, which is bigger, hungrier and built for capacity the IoT device will never use.

This is also why GEO is such a natural fit: IoT traffic tolerates the ~half-second GEO round trip without complaint, and a GEO satellite gives a fixed, always-there coverage footprint that a sleepy sensor can wake into on a schedule. LEO is supported too, but the delay-tolerant nature of NB-IoT is what makes GEO IoT so comfortable.

How IoT-NTN keeps evolving

The separate track keeps getting its own enhancements. Rel-18 extends IoT-NTN alongside NR-NTN — improving mobility and service continuity, handling discontinuous coverage (so a device in a sparse constellation can predict when a satellite is overhead and sleep through the gaps), and relaxing the GNSS burden so the receiver spends less energy re-acquiring position. Rel-19 and beyond continue in the same direction: better power efficiency, store-and-forward operation that suits an intermittently-visible satellite, and steps toward regenerative (on-board) processing. The theme never changes — squeeze more coverage and more battery life out of a deliberately minimal radio. The broader release arc is traced in ntn-evolution.

The IoT-NTN picture

Put together, the end-to-end path is simple: a frugal IoT device pre-compensates using GNSS, talks through a GEO or LEO satellite to a ground gateway, and lands on an EPC rather than a 5GC.

Summary

IoT-NTN is the low-power cousin of NR-NTN: the LTE-based NB-IoT and eMTC radios (36-series, TR 36.763) taken over satellite, usually on EPS/EPC. It shares the satellite physics — long delay, large Doppler — and therefore reuses NR-NTN's toolkit wholesale: GNSS-based UE pre-compensation of timing and frequency, ephemeris and common-TA broadcast in LTE SIBs (the analogue of NR SIB19), and scheduling offsets with scaled timers to absorb the round trip.

Where it diverges is by design: a deliberately minimal radio — narrowband, half-duplex, repetition-heavy for coverage enhancement (eMTC CE Mode A/B, NB-IoT CE levels), only 2 HARQ processes in NB-IoT, and deep sleep via PSM and eDRX — optimised for years of battery and reachability in terrible link budgets rather than for speed. That is exactly why it earns a separate track, and why GEO suits it so well: delay-tolerant traffic rides the half-second round trip happily under a fixed coverage footprint. Rel-18 and beyond keep sharpening the same edges — mobility, discontinuous coverage, and a lighter GNSS burden.