NetSim NTN Q&A

5G non-terrestrial networks · LEO, MEO, GEO · Full-stack simulation

This page answers questions on how to simulate 5G non-terrestrial networks end to end. It covers the orbits, beams and bands that NetSim models, the contents of the link budget, and scheduling when the round trip is hundreds of milliseconds. It also covers how many satellites a full protocol stack can carry, and what the outputs look like. Q1 to Q6 cover scope and positioning. Q7 to Q31 cover the product, including what is not modelled today.

Request an evaluation Back to NetSim NTN

On this page

Thirty-one questions in six parts, from what the tool is to what it does not do yet.

Q1–Q6 Scope and positioning Q7–Q12 Constellations, orbits and beams Q13–Q18 Radio modelling Q19–Q24 Scheduling, feeder link, handover, scale Q25–Q29 Running studies and reading results Q30–Q31 Limits and direction

Part 1: Scope and positioning

What the simulator is, which class of tool it belongs to, and how it relates to coverage planning.

Q1. What is NetSim NTN?

NetSim NTN is a full-stack, packet-level simulator for 5G non-terrestrial networks. It simulates LEO, MEO and GEO systems that carry real 5G NR traffic between UEs, satellites, gateways, gNBs, the 5G core and remote servers.

The NTN library shares its protocol stack with the terrestrial 5G library. PHY, MAC, RLC, PDCP, RRC and the core therefore use the same code paths as terrestrial NR studies. The results are ordinary network metrics (throughput, delay and loss), from a network-wide view down to individual links, devices and packets.

Q2. Where does it sit relative to a link-level tool or an orbital and link-budget tool?

Three classes of tool answer three different questions.

Class of tool What it answers What it cannot produce
Link-level What a single radio link achieves. One terminal, one beam. No scheduler and no second user, so no handover rate, no per-beam distribution, no scheduling outcome.
Orbital and link-budget Where satellites are and whether a link closes. Scales to thousands of satellites because it computes geometry and link closure rather than protocol behaviour. Capacity is a Shannon bound, not a delivered throughput. No packets, no queues, no protocol.
Full-stack, packet-level What the network delivers to the user. A packet goes from an application through PDCP, RLC, MAC and PHY, and over the service link to the satellite. It then goes over the feeder link, into the 5G core and out to a server. Constellation scale, because a full protocol stack costs far more per node than geometry does. This is the class NetSim NTN occupies.

Related: NetSim Astra is the second class, and the two are complementary rather than competing. See Q3.

Q3. Should I use NetSim NTN or NetSim Astra?

Use NetSim Astra to decide where coverage exists, whether the link closes, and what the access and gap statistics look like over a region. Astra propagates the constellation, computes an ITU-R link budget at every grid point and time step, and renders coverage on a 3D globe.

Use NetSim NTN to find out what throughput and latency a user actually gets, and how the scheduler, beams and interference behave under load. Coverage first, then carried traffic. Many studies use both.

Q4. Which 3GPP specifications does it follow?

  • Channel model and path loss: TR 38.811.
  • Architecture, link budget and reference scenarios: TR 38.821, including the CNR expression in Section 6.1.3.1 and the Set 1 evaluation.
  • Uplink scheduling: Configured Grant Type 1 per TS 38.214 Section 6.1.2.3 and TS 38.321 Section 5.8.2.
  • Satellite antenna pattern: the circular aperture model of TR 38.811 Section 6.4.1.

The feeder link carries F1 over the Satellite Radio Interface. The service link uses NR-Uu.

Q5. Does it support regenerative payloads?

Not today. The architecture is the transparent payload, or bent pipe, of TR 38.821. The satellite filters, converts frequency and amplifies, but does not otherwise change the waveform. The whole gNB is on the ground.

A gNB CU/DU split with configurable DU placement, and a full regenerative payload terminating NR-Uu on board, are both planned. See Q30.

Q6. Who uses NetSim NTN?

  • Satellite operators sizing constellations and beam plans.
  • Researchers evaluating 5G NTN against 3GPP reference scenarios.
  • Defence and government programmes assessing satellite-backed connectivity.
  • Equipment vendors testing terminal and scheduler behaviour.
  • Universities teaching NTN with reproducible experiments.

Part 2: Constellations, orbits and beams

This part describes how NetSim creates and propagates satellites, how it places beams, and how a UE selects a beam.

Q7. Which orbits and altitudes does NetSim support?

NetSim supports LEO, MEO and GEO, in single-satellite and multi-satellite scenarios:

  • A single-satellite study has the presets LEO 600 km, LEO 1200 km, MEO 10000 km and GEO 35786 km. A custom altitude can be any value from 1 km to 35786 km.
  • A Walker constellation takes its altitude and inclination from the wizard.
  • A live TLE constellation follows the orbits in the public catalogue.
  • A GEO satellite sits at 35786 km in a longitude slot.

Q8. What is the difference between a single-satellite and a multi-satellite study?

In a single-satellite study, one platform serves the whole UE region. This suits link-budget checks, beam coverage studies and traffic runs in which the same satellite serves every UE.

Multi-satellite places two or more satellites over the same region, which is what satellite diversity, coverage overlap, beam association, inter-satellite interference and service continuity require.

Q9. Can I simulate a real constellation such as Starlink?

Yes. NetSim loads satellites from live TLE data by constellation category, with a maximum satellite count that keeps the scenario tractable. Each satellite follows its own SGP4-propagated trajectory.

At every mobility update, NetSim takes the position from the propagated orbit. Passes, elevation changes and link delay therefore follow the real two-line elements, not a fixed analytical track.

Q10. How do I define a synthetic constellation instead?

Use the Walker Delta or Walker Star pattern. NetSim generates it from altitude, inclination, plane count, satellites per plane, RAAN spacing and phase offset. NetSim reports the resulting satellite count before beam setup.

A Cesium 3D globe renders node placement, satellite tracking and link geometry against real Earth geometry. You can therefore inspect tracks, beam footprints and coverage movement before you create the network. You can place UEs, gateways and core devices directly on the globe, and they take their coordinates from the drop point.

Q11. What is the difference between earth-fixed and earth-moving beams?

Earth-fixed beams steer to hold a ground point, so they need target locations, usually supplied as a CSV. This is the established mode, and it is what a single-satellite study uses.

Earth-moving beams point at the satellite nadir and sweep with the satellite, so they need no target file. NetSim computes the beam centre and ground footprint again at every mobility update. The multi-satellite studies use this mode.

Other beam options:

  • Any number of Earth-fixed beams, one-to-one with cells, with built-in presets of 1, 7 and 19
  • A per-beam minimum elevation: a fixed beam is inactive until its satellite clears it
  • Timed reconfiguration: change the pointing or channel of a beam at a set simulation time
  • Frequency reuse factor FRF 1, 2, 3 or 4, with the hexagonal inter-site distance from the beam diameter

Earth-fixed beams maximise the time that a UE stays under one satellite. For LEO, this is the time that the satellite stays above the horizon, typically 7 to 10 minutes.

Q12. How does a UE choose its satellite and beam?

NetSim considers only the satellites above the configured elevation mask. It ignores any satellite below the mask for that UE at that time. For each eligible satellite, NetSim computes the full link budget across its beams. The budget covers carrier frequency, slant range, elevation, path loss, shadow fading, clutter, antenna pattern gain, receiver G/T, thermal noise, SNR and SINR.

The UE attaches to the satellite and beam that give the highest SNR. NetSim evaluates this again as satellites and UEs move. If no satellite passes the coverage check, NetSim treats the UE as outside NTN coverage for that evaluation.

Part 3: Radio modelling

Bands, link budgets, antenna patterns, interference and block error rate.

Q13. Which frequency bands does NetSim support?

  • S-band in FR1: n252 and n256.
  • L-band in FR1: n253, n254 and n255. n254 is a mixed case, with an L-band uplink and an S-band downlink.
  • Ku-band: n247 and n248 in FR1, and n508 and n509 in FR2.
  • Ka-band in FR2: n510, n511 and n512.
  • C-band: provisional.

Handheld UEs connect on FR1 only. VSAT terminals connect on FR1 or FR2. For FDD carriers, the uplink centre frequency drives the uplink losses, and the downlink centre frequency drives the downlink losses.

Worth knowing: TR 38.811 tabulates shadow fading and clutter loss for S-band and Ka-band only. NetSim uses those values directly for S and Ka. L-band uses the S-band values. NetSim interpolates C-band and Ku-band on log-frequency between the S-band and Ka-band anchors, so treat them as engineering approximations.

Q14. What goes into the link budget?

The link budget uses the TR 38.821 CNR expression. It adds EIRP and receiver G/T. It then subtracts Boltzmann's constant (−228.6 dBW/K/Hz), free-space path loss, shadowing margin, scintillation loss, polarisation mismatch, atmospheric loss, additional loss, and 10·log10 of bandwidth.

Receiver G/T comes from receive antenna gain, noise figure, ambient temperature (290 K by default) and antenna temperature. You can set the scintillation, polarisation and atmospheric terms in dB. NetSim adds them to the free-space and shadowing loss before it computes the received power.

Q15. Is the uplink budget just the downlink reversed?

No. The uplink uses the UE's total EIRP rather than an EIRP density, and the satellite's G/T at the off-boresight angle rather than the UE's. NetSim computes path loss in the same way in both directions. The satellite's absolute gain is its maximum gain plus the relative gain at that angle.

Q16. Which antenna models are available?

  • The TR 38.811 circular aperture pattern, using the first-order Bessel function.
  • The ITU-R S.672 off-axis pattern for fixed-satellite service, with configurable near-in side-lobe level.
  • A Gaussian model, G(θ) = −3.01·(θ/θbw)².

The off-boresight angle is the arccosine of the normalised dot product of the satellite-to-beam-centre and satellite-to-UE vectors.

Q17. How is interference modelled?

NetSim has two models. A CIR-based model combines carrier-to-noise and carrier-to-interference into CNIR per TR 38.821 Section 6.1.3.1, with interference power back-calculated from noise, CNR and CNIR.

An exact geometric model derives interference from every overlapping co-channel beam, on the serving satellite and on every other satellite in view. The frequency reuse factor therefore determines which beams contribute. Under reuse 1, every beam interferes. Under higher reuse factors, only beams on the same channel ID interfere.

Q18. How is block error rate handled?

Today BLER is a user input between 0 and 1, which NetSim applies as a target transport-block BLER. From it, NetSim back-calculates the per-code-block error probability as p = 1 − (1 − BLER_target)^(1/N) for a transport block of N code blocks. The realised transport-block BLER then matches the configured value. HARQ is disabled, so this applies to new transmissions only.

A planned development derives BLER from per-slot SINR and the assigned MCS through link-level curves. It depends on HARQ, which must come first.

Part 4: Scheduling, feeder link, handover and scale

The parts where satellite delay and constellation size change the engineering.

Q19. How is uplink scheduling done, given the propagation delay?

Terrestrial dynamic grant assumes sub-millisecond round trips for the scheduling request, grant, buffer status report and PUSCH. In NTN, one-way delay runs from about 4 ms at LEO 600 km to 240 ms at GEO, so round trips reach 8 ms to over 500 ms. A four-step handshake would add hundreds of milliseconds before any data moves.

NetSim therefore implements Configured Grant Type 1 as a cyclic round-robin scheduler in time-division mode:

  • One dedicated uplink slot per UE per cycle, with the cycle length equal to the number of associated UEs.
  • All available PRBs to the scheduled UE, after overhead deduction.
  • An independent schedule per beam.
  • MCS selected per UE from the uplink budget: SNR to CQI to MCS.

Q20. Which downlink schedulers are available?

Today NetSim has Round Robin, with full PRB-level resource allocation on FDD carriers. Proportional Fair and Max C/I are planned, so that you can compare scheduling policies on the same constellation.

Q21. Is the feeder link modelled as a real radio link?

Not today. NetSim follows the simplification that TR 38.821 permits. It treats feeder-link SNR impairments as negligible and models only the propagation latency. For this reason, NetSim computes the link budget on the service link.

Gateway selection and the feeder-link switch work today. For each satellite, NetSim chooses the visible gateway with the highest elevation above its mask (10° by default). It checks the choice again at every satellite position update, so the feeder link switches as the satellite moves. One gateway can serve many satellites at once, and the Gateway Association Log records every change.

These feeder-link items are not available yet:

  • A feeder-link budget with its own antenna, EIRP, G/T and loss terms
  • Q and V band feeder carriers alongside Ka
  • Gateway selection that uses feeder-link quality, not only elevation

These items matter at Ka, Q and V band, where the feeder link carries a real budget.

Q22. Is handover modelled?

Yes, as a selection procedure. The UE switches to another satellite only when that satellite's SNR exceeds the serving SNR by the handover margin (3 dB by default). This stops ping-pong between satellites of similar strength. Beam changes on the same satellite are immediate. The Serving Transition Log records every change. NetSim tracks the serving satellite, beam and SNR per slot, and evaluates the association again as described in Q12. The switch itself is instantaneous, with no control-plane exchange and no measurement reporting.

Handover as a measured procedure is planned. It includes:

  • Measurement reporting, decision, execution and completion, with configurable margin, time to trigger and interruption time
  • Conditional handover triggered on predicted satellite visibility, which the short LEO visibility window demands
  • KPIs for attempts, successes, failures, ping-pong, interruption time and packets lost or buffered per handover

Q23. Is Doppler modelled?

Today NetSim assumes ideal Doppler compensation. It assumes that every UE has GNSS and pre-compensates Doppler and delay from the satellite ephemeris and its own position, with perfect timing advance.

These Doppler items are planned:

  • Explicit Doppler shift and rate for each link, from satellite and UE velocity along the slant path, at every mobility update
  • Common network-side pre-compensation, with the residual frequency offset carried into the link abstraction so that its effect on demodulation is visible
  • Ephemeris-assisted acquisition with a configurable K-offset per Release 17

Q24. How many satellites can it simulate?

Today a multi-satellite scenario runs the full stack across all its satellites. The practical ceiling depends on beam count, UE count and run length, not on a fixed number. The NTN page shows two examples: a 300-satellite LEO constellation over Tokyo for five hours, and a 225-satellite constellation over four cities. Tell us the study, and we will size it.

The ceiling is not specific to one product. In a system-level wireless simulation, the channel computation between transmitter and receiver pairs takes most of the cost. That cost grows quadratically with the number of nodes that share a channel. Tools that report results at thousands of satellites usually abstract the radio layer down to geometry, routing and latency. Tools that keep a full protocol stack usually show results for one or a few satellites.

A planned development raises that ceiling and keeps the stack. It uses two-tier fidelity in a single simulation:

Tier What NetSim models Scale
Full-stack Serving satellite and interfering neighbours. Complete PHY, MAC, RLC, PDCP, RRC and 5G core, per-beam PRB scheduling, link adaptation, packet traffic, end-to-end latency and throughput. Tens of satellites, hundreds of UEs
Constellation All remaining satellites, propagated by SGP4. Position, velocity, visibility, elevation, slant range, propagation delay, beam footprint and aggregate interference. No per-packet processing. Hundreds of satellites

The idea: NetSim promotes a satellite into the full-stack tier when it comes into service and demotes it when it passes out of view. The tier of each satellite therefore follows the orbital geometry. A UE experiences a full constellation, while the simulator pays for a full stack only on the links that carry its traffic.

Part 5: Running studies and reading results

What comes out of a run, and which studies ship ready to modify.

Q25. What comes out of a run?

Standard NetSim metrics, in tabular and graphical form, plus NTN-specific logs:

  • Radio Measurement log, one row per sub-frame: time, gNB and UE names, slant height, elevation angle, carrier and band, EIRP, path loss, shadow fading, additional loss, clutter loss, total loss, beamforming gain, angular antenna gain, UE receive gain, received power, SNR, SINR, thermal noise, interference power, CQI index and MCS index.
  • Beam Association log: serving satellite and beam over time, including no-coverage intervals.
  • Resource Allocation log: per-slot PRB, MCS and transport block size.
  • Serving Transition log and Gateway Association log: every change of serving satellite and of gateway.
  • PRB Utilization and Code Block logs.
  • Packet trace: individual packets across the gateway, satellite and UE hops, alongside the event trace.

Q26. Can I drive satellite or UE movement from my own data?

Yes, through file-based mobility. A CSV in the format <time in seconds>,<device id>,Lat(deg),Lon(deg),Alt(m) sets positions for satellites or UEs over time, which is the mechanism for custom trajectories and externally computed tracks.

Aeronautical and maritime UE mobility models are planned.

Q27. Can I reproduce the 3GPP TR 38.821 Set 1 reference scenario?

Yes. A featured example reproduces the TR 38.821 Set 1 configuration. It has LEO at 600 km, S-band, 19 beams in a hexagonal layout, 10 UEs per beam (190 in total), a transparent relay, full-buffer downlink traffic and Round Robin scheduling.

Satellite EIRP density 34 dBW/MHz, 0.5 m aperture, 47.22 km beam radius, 30 MHz per beam, UE noise figure 7 dB, 30 second run. It produces CDFs of SINR, per-UE throughput and per-beam throughput.

Q28. What parameter studies ship as worked examples?

  • LEO altitude against SNR and path loss.
  • SNR against transmit power, in rural and urban environments.
  • SNR and path loss against elevation angle.

You can run each sweep again with your own values, in addition to the TR 38.821 Set 1 scenario in Q27.

Q29. Can I combine NTN with cyber attacks or machine learning?

Most attacks in the cyber library can be ported to NTN scenarios with minor code modifications. The Python and MATLAB interfaces let you put an external model in the loop. Ask us for a worked NTN example, and do not assume a particular integration point.

On rain attenuation specifically, see Q30. Today you enter it as an atmospheric-loss value. ITU-R P.618 and P.676 models are planned.

Part 6: Limits and direction

This part lists what is out of scope today and what is in development.

Q30. What is not modelled today?

  • Inter-satellite links.
  • Regenerative payloads. The whole gNB sits on the ground, with no CU/DU split.
  • HARQ. It is disabled, and RLC runs in UM mode only.
  • Native rain attenuation. Rain and gaseous absorption enter today as a user-entered atmospheric-loss value in dB, applied in the link budget.
  • Terrestrial to NTN coexistence and handover.
  • Broadcast and multicast transmissions.
  • SIB19, with SIB and MIB behaviour following terrestrial 5G.
  • TDD carriers, since NTN carriers are FDD today.
  • Antenna polarisation.

Four items are idealised, not absent. Know them before you interpret a result:

  • The RACH procedure is ideal.
  • Channel state information is perfect at both transmitter and receiver.
  • The UE, satellite and gNB are perfectly synchronised in frequency and time.
  • Doppler compensation happens at the UE, as described in Q23.

Several of these are planned developments, not permanent limits:

  • The gNB CU/DU split with configurable DU placement, on the ground beside the gateway or on board with F1 over the Satellite Radio Interface. The DU placement is the MAC scheduler placement, so it sets the scheduling round-trip time and the achievable link adaptation.
  • A full regenerative payload that terminates NR-Uu on board
  • Inter-satellite links with routing across the constellation
  • HARQ sized for the NTN round-trip time, with RLC AM alongside UM
  • TDD with NTN-sized guard periods
  • Native ITU-R P.618 rain attenuation with P.676 gaseous absorption

Q31. Can I study multi-orbit systems, for example steering traffic between GEO and LEO?

Not in a single scenario today. A planned development puts a GEO layer and one or more LEO shells in one run, each with its own band plan, beam layout and gateway set.

The intent is to choose the serving orbit per bearer from QoS class and traffic priority, not from SNR alone. Latency-sensitive traffic then goes to LEO, and capacity-oriented or delay-tolerant traffic goes to GEO. NetSim will measure the delay, packet loss and throughput transient on each switch, as for a handover.

Evaluation

Run your own constellation

Request a demonstration or an evaluation license. A typical evaluation starts from the TR 38.821 Set 1 scenario, because it gives a known reference point. It then moves to your own constellation, band plan and traffic model.

Request a demonstration NetSim NTN product page

Sources, scope and trademarks

Scope. Answers describing current behaviour reflect NetSim v15.1 as of October 2026. Specification references are to 3GPP TR 38.811, TR 38.821, TS 38.214 and TS 38.321, and to ITU-R Recommendations S.672, P.618 and P.676. Capabilities change between releases. Confirm the current position with Tetcos.

Planned items. Answers marked Planned describe intended development. They are indicative, carry no delivery commitment, and are subject to change. Nothing on this page should be read as a contractual undertaking. If a planned capability matters to a decision you are making, ask us for the current status rather than relying on this page.

Trademarks. Starlink is a trademark of Space Exploration Technologies Corp. MATLAB is a registered trademark of The MathWorks, Inc. Cesium is a trademark of Cesium GS, Inc. All other product and company names are trademarks of their respective owners, used here solely to identify the products discussed. Tetcos is not affiliated with, endorsed by or sponsored by any of them, and no such relationship is implied.