Orbital Lab v1.3

Silent. Enable sound when you want it.

Quiet attachment, handover and service-loss cues. Ambience follows the vehicle and season at natural speed. Audio pauses with playback or when this tab is hidden. Illustrative sound, not an acoustic model.
T+00:00Running
Educational model · 5G TN / NTN

5G Connectivity Across Countries

Tower–satellite handover simulation on illustrative regional corridors.

Country profile

Start with a taskNominal demonstrations reset radio settings. Offshore starts are fresh observations.
Fig 1Route-plane projection · fixed spacecraft identities · travel →
Route map
Applications over this link

Network attachment—
Requested rate—
Observed performance · DL / ULCollecting packets
What is assumed?Research & methods

Rate, delay and admission assumptions

Calls are two-way: 0.6 Mbps per direction for a standard call and 1.2 Mbps per direction for a 720p reference. The stream is one-way media plus a small return stream. These CBR traffic budgets approximate application recommendations; they do not emulate Zoom, a video codec, adaptive bitrate or a player buffer. Call packet deadlines use an assumed 150 ms network target, excluding codec and playout delay. “Tight headroom” means less than a 20% capacity margin above the requested rate in either direction, an illustrative diagnostic. Rate feasibility does not guarantee perceived quality; inspect packet loss, delay and handover interruptions.

The satellite access model is a shared 5 MHz S-band example, not a commercial Ku/Ka broadband terminal. Eight users cap nominal downlink at 1.56 Mbps each before link and feeder losses. More visible satellites supply alternatives, not aggregated bandwidth. Receiver, antenna, power and allocation assumptions govern uplink independently.

View guide and stage colors

Each ID follows its own continuous orbit. Large icons mark serving/candidate satellites; off-plane or blocked objects are small and faint. Size changes are gradual and never relocate a spacecraft. Trails show the previous 24 simulated seconds. Ship side views show pitch with display smoothing; full roll and pitch still affect the antenna on simulated time. Hull and terrain scales are schematic. Accelerated playback compresses travel time, not physical distance.

Discovery & measurementHandoverConnected & completeFailure & recovery
Solid: serving · dashed: candidate/signaling or interrupted · gray: released · Space pause · → step · speed adapts to device
—
Application continuity Waiting for attachment
Journey time meeting the application target—
Packet loss · resolved—
Interruption · p95—
One-way delay · p95—

How these estimates are measured

Rates and one-way deadlines are assumed application profiles, not standardized 3GPP QoS classes. Illustrative CBR packets in both directions, separate FIFO queues of 512 packets, and application deadlines. Video-call loads are symmetric; the video-stream return load carries only control traffic. The 150 ms call network deadline is assumed and excludes codec and playout delay. Rates are fixed; adaptation and playback buffering are not simulated. Capacity comes from the link-budget proxy; base one-way delay is RTT/2. No TCP, HARQ, retransmission, codec, or playback buffer is simulated. Delay includes queueing and transmission; displayed delay percentiles use 1 ms histogram bins. Resolved handover success excludes cancellations and pending procedures. Delay variation is the mean absolute difference between consecutive delivered-packet delays, not RFC 3550 jitter. Independent uplink packets use the profile’s UL demand; downlink and uplink are both required for application satisfaction.

Packets start after the first usable attachment in each journey. Acquisition is shown separately. “Journey time meeting the application target” uses the entire observation window, including initial acquisition; a fresh offshore start measures only its own segment. The separate after-attachment figure excludes that initial acquisition period. Loss divides expired/overflow packets by resolved packets; queued/in-flight packets are reported as pending. Interruption p95 uses completed zero-capacity episodes; an episode still open at the observation end is reported separately. Automatic route returns close the current measurement and start a new one; the artificial return and subsequent reacquisition are never joined into an outage.

Uplink, downlink and antenna Waiting for attachment
Downlink capacity—
Uplink capacity—
UL SNR margin—
Satellite pointing error—

Compare handovers and antennas Paired experiments

Compare handover strategies or antenna models over the same selected route segment and application. All other settings, including capacity policy, stay fixed. Each repetition pairs the same seed and orbital offset across policies. The live simulation continues independently.

No comparison has been run.

Experimental method and limits

Signal ranking uses the strongest admitted candidate. Visibility ranking compares the six strongest admitted satellites, looking ahead up to 240 s in 10 s increments. Predictive selection checks geometry, canopy and current assumed budgets up to 60 s ahead at 2 s resolution; it can request an early handover if the serving path is predicted to block within 5 s. Forecasts refresh once per simulated second, assume constant vehicle speed and known terrain, and hold the current satellite shadow fade; terrestrial forecasts use the seeded route-channel profile. They are ideal route-knowledge surrogates, not operational predictions. Terrestrial event thresholds and satellite range-trigger rules otherwise remain fixed.

A run ends at the chosen time or just before the route boundary, whichever comes first. Choose To route end and start at 0 km for a full journey; these runs may take substantially longer. Use multiple repetitions to assess variation under the model assumptions. Repetition i uses seed + i and a 137i-second orbital offset, shared by every policy. Report median and range across repetitions, completed-outage p95, delivery-delay p95 and resolved packet loss; pending packets, open outages, cancelled/pending handovers and unattached trials remain explicit. No improvement is assumed and no significance claim is made. These are model comparisons, not field validation.

Connection metrics, nodes and handover log
Serving link
—
—
Radio RSRP proxy
—dBm
—
Per-vehicle downlink
—Mbps
—
Estimated RTT
—ms
—
Handovers
0 · 0 · 0
tower↔sat · sat↔sat · cell↔cell

Nodes

NodeRoleTo the vehicleRSRPState

Inspector

Signal at the vehicle L3-filtered RSRP · dBm · last 3 min

Tower → satellite when the tower is below A2 (−110 dBm) and a satellite above A5 (−124 dBm) for 640 ms. Satellite → tower when a tower is above A4 (−103 dBm) for 1.28 s, at least 20 s after the last handover. Illustrative failure when DL raw RSRP or UL budget availability fails continuously for 1 s; this is not full 3GPP radio-link monitoring.

Handover log

TimekmEventDetail

Country assumptions

The water model uses a line-of-sight horizon, free-space loss and synthetic obstructions. It is not calibrated to this sea or lake.

Natural Earth map boundaries are generalized and do not define legal borders. Route coordinates are rounded illustrative references.

Technical reference appendix in English

Scientific method & reproducibility

Question: how do terrain, geometry, and assumed radio budgets affect connectivity? Change one factor at a time, reset to the same seed, compare exported runs, and repeat across seeds. A single realization is not a coverage probability.

Verification checks equations against analytic limits and reference values. Validation still requires surveyed terrain, real antenna/EIRP data, drive-test RSRP/SINR, traffic and latency measurements. Shadowing σ describes only the assumed random component; it is not a confidence interval for model accuracy.

Model

Model and assumptions — with sources
Model componentWhat the simulator doesBasis
Land tower path loss3GPP rural-macro (RMa) LOS/NLOS path loss, LOS probability and break-point distance, using physical mast and receiver heights above their local ground. A seeded, spatially correlated LOS/NLOS state uses the RMa LOS probability; terrain blockage forces NLOS. The hybrid heuristic uses total loss = max(RMa state loss, RMa LOS + dominant-edge diffraction), with explicit foliage and enclosure losses. Beyond 5 km NLOS / 10 km LOS, a continuous 40 dB/decade extension is flagged as extrapolation.[1]
Maritime coastal linksFree-space loss plus illustrative dominant-edge diffraction and 2.5 dB correlated spatial shadowing. Antenna elevations above sea level determine a 4/3-effective-Earth line-of-sight envelope. Beyond-horizon paths are excluded by this scoped baseline. No coherent sea reflection, ducting or calibrated maritime channel prediction. Coastal towers are hypothetical.Geometry / FSPL baseline; illustrative channel
Tower antennaVertical pattern A(θ)=−min[12((θ−θ_tilt)/θ_3dB)², SLA] with 3° electrical downtilt, 10° beamwidth, 20 dB side-lobe floor.[1]
Terrain diffractionKnife-edge diffraction loss J(v) over the elevation profile, using the dominant sampled edge and Earth bulge for a 4/3 effective Earth radius.[6][8]
ForestFoliage and canopy attenuation use illustrative country profiles. These coefficients are not locally calibrated. Snow mode is offered for Türkiye, Sweden and Canada; wet and dry states are synthetic weather assumptions, not a climate calendar.[7]
Vehicle & displayThe car front follows increasing route distance (right on screen), with pitch from its two displayed wheel contacts. Elevation is linearly interpolated from 50 m samples. Speed adjustments use the selected vehicle acceleration and braking limits. Vehicle sizes, terrain height and satellite distances use separate schematic scales. The default vehicle access minimum elevation is 20°, an assumed operating limit rather than a verified commercial S-band requirement; terrain may impose a higher mask. The gateway retains its separate 10° assumption. Playback is device-adaptive: it retains all physical integration steps and may run below the selected wall-clock speed on a busy device; no simulated time is skipped. Vehicles cruise to the full route boundary and automatically begin a new journey at the selected cruising speed, with no terminal braking or stopped phase. Orbital time and satellite identities continue; this return is a journey reset, not physical travel between the endpoints. Scenario edits apply automatically and restart the journey. A tower-free view does not imply radio silence from distant masts.Geometry / kinematics
Satellite orbitsWalker-delta shells propagated as circular orbits and rotated with the Earth: 53° (550 km, 72×22), 70° (570 km, 36×20) and 97.6° (560 km, 6×58), using illustrative shell populations inspired by SpaceX filings, not live ephemerides. A spherical Earth includes receiver altitude; perturbations are omitted. Elevation, azimuth and slant range are computed topocentrically. Spacecraft have persistent IDs and continuous propagated positions. Serving and candidate bodies smoothly enlarge at their own position; background, blocked and off-plane bodies are small and faint. Icon scale and opacity are visual only. The 24 s trails propagate the same spacecraft at earlier times in the current observer frame; their geometry is not a handover animation. These are ideal Keplerian circular orbits, not live tracked ephemerides.[9][10]
Satellite linkFSPL = 20 log₁₀(4πdf/c) at 2 GHz. Assumed EIRP is 36.2 dBm per reference resource element; receive gain, train loss, scan loss, atmosphere and canopy are explicit terms; The displayed access RSRP is an isolated component-link budget. Transparent-payload reference waveforms, analog transponder gain and cascaded noise are not synthesized; route failure can therefore coexist with an unchanged access-budget trace. NTN architecture is inspired by the 3GPP NTN studies.[2][3][11]
Satellite fading & gasesSeeded Gaussian shadowing with elevation-dependent σ ≈ 0.7–1.8 dB and an assumed 20 s temporal grid (not a measured correlation time) and gaseous attenuation ≈0.04 dB/sin(el) at S-band; scan loss uses each shell's own altitude.[2][22]
Feeder linkInstantaneous 20 GHz return and 30 GHz forward feeder budgets: P_rx(dBm)=EIRP(dBW)+30+G_rx−FSPL(f)−2 dB fixed loss−rain fade. Noise uses 100 MHz and the receiver NF. Rain fade is γ=kR^α times min(slant range, rain-cell cap, slab depth/sin(elevation)). It now determines feeder SNR, capacity and availability against a assumed threshold. P.838-3 circular-polarization coefficients at 20 and 30 GHz are used. Both directions must meet the assumed SNR threshold; forward-feeder capacity limits the vehicle downlink proxy. The transparent option uses independent-hop bottleneck budgets rather than an analog transponder gain/noise model. The uniform rain slab has an assumed depth and a finite cell-path cap; it omits P.618 reduction factors and exceedance statistics, so the result is an instantaneous estimate, not annual availability.[23][24]
Noise and capacityNoise per reference subcarrier is −174 + 10 log₁₀(Δf Hz) + NF dBm (NF=7 dB; 15 kHz, or 30 kHz for n78). Tower SINR sums other co-channel towers in linear power with load as assumed activity. NTN uses noise-limited SNR, without inter-beam interference. Mbps = bandwidth × capped η log₂(1+SINR) × assumed layers × available resource fraction. This is a capacity proxy, not measured application goodput.Link budget / Shannon bound
Applications and admissionVoice: 0.096 Mbps each direction; one-to-one call references: 0.6 or 1.2 Mbps each direction; stream: 2.5 Mbps down and 0.064 Mbps up. Fixed CBR proxies; call network deadlines 150 ms, stream 300 ms. Instantaneous feasibility checks both rates and baseline one-way delay plus packet serialization. Packet observations retain actual simulated queueing, deadline loss and interruptions. A diagnostic flags less than 20% rate headroom. Best effort preserves a viable positive-capacity bearer below application demand; full-rate admission rejects it. No codec, adaptive congestion control or video buffering is modeled.Zoom rate guidance; assumed traffic, deadlines and admission policy
Measurement filteringLayer-3 log-domain filter uses a=1−2^(−Δt/0.2) to preserve the k=4 response for an assumed 200 ms reference period as playback or integration steps change; events use the filtered RSRP, radio-link failure the unfiltered level.[4]
Mobility eventsIllustrative A2/A3/A4/A5 thresholds and time-to-trigger settings; a DL-threshold/UL-budget timeout approximates failure and recovery. Actual T310 starts from N310 lower-layer out-of-sync indications, which this simulator does not model. The B20 option uses the same simplified mobility logic, not a separate LTE protocol stack. Gateway backhaul failure does not itself cause a raw-RSRP radio-link failure. Handovers recheck radio, route and the selected capacity policy at reservation and before path switch; policy thresholds are assumptions. Best effort permits a rate-limited connected application; full-rate admission requires both requested rates.[4][5]
Satellite handoverCustom conditional handover: target slant range must be at least 40 km shorter for 1.024 s, or the serving path is lost. Candidate ranking is selectable: strongest admitted signal, visibility among the six strongest admitted satellites, or ideal route-based predictive blockage avoidance. Baseline range triggers and thresholds otherwise remain fixed. Predictive mode can request an early handover when its serving path is forecast to block within 5 s. This is not standardized D1, which uses two geographic reference locations and two distance thresholds (TS 38.331 §5.5.4.15).[12][13]
TunnelsPortal leakage with fast decay along the tunnel; leaky-feeder cells in equipped tunnels with cable loss and coupling loss as specified per IEC 61196-4; handover to the tunnel cell at the portal. Satellites are lost in tunnels.[14][15][16][17][19]
Inside the train≈30 dB car-body loss with metal-coated windows, ≈9 dB with RF-transparent windows, none with a roof antenna plus repeater or Wi-Fi.[15]

Simplifications: 2-D route profiles rather than a 3-D DEM; generic, uncalibrated transmitter powers and loads; which tunnels are equipped, and the tunnel lengths, are illustrative. The simulator is educational and has not been validated against drive-test measurements. RMa covers the selected 0.7–3.5 GHz frequencies; its NLOS/LOS distance limits are 5/10 km, with any extension shown as extrapolation. Open-water propagation uses a scoped LOS envelope: free-space loss plus illustrative dominant-edge loss, 2.5 dB spatial shadowing and a 4/3-effective-Earth horizon computed from antenna elevations above sea level. Paths beyond that horizon are excluded by this baseline; actual beyond-horizon propagation is not predicted. Coherent two-ray sea reflection, ducting, troposcatter and variable refractivity are omitted. Coastal sites and island profiles are synthetic. Ship port endpoints are approximate great-circle references, not navigable ferry tracks; forest seasons, throughput efficiencies and layer counts, fixed terrestrial RTT, signal thresholds for radio-link failure, and regenerative ISL routing are heuristic assumptions. RSRP thresholds are not a full standards-compliant radio-link-monitoring implementation. The DTC shell changes orbital geometry only; it uses the same illustrative S-band link budget, not a validated commercial direct-to-cell radio model.

Research basis & sources

  1. Application-rate reference: Zoom desktop system requirements · bandwidth for one-to-one calls. Used as fixed traffic-budget references (0.6 and 1.2 Mbps each direction), not a Zoom protocol simulation. ITU-T G.114 discusses conversational delay; the simulator’s 150 ms packet deadline is an assumed network target, not a complete mouth-to-ear budget.
  2. 3GPP TR 38.901 V17.0.0, ETSI PDF · Tables 7.4.1-1 and 7.4.2-1, Study on channel model for frequencies from 0.5 to 100 GHz (Rel-17). 3GPP portal
  3. 3GPP TR 38.811, Study on New Radio (NR) to support non-terrestrial networks (Rel-15). 3GPP portal
  4. 3GPP TR 38.821, Solutions for NR to support non-terrestrial networks (NTN) (Rel-16). 3GPP portal
  5. 3GPP TS 38.331 V17.9.0, ETSI PDF · §§5.5.3.2, 5.5.4.15 and 5.3.10, NR; Radio Resource Control (RRC) protocol specification — measurement events, TimeToTrigger, conditional handover, T310. 3GPP portal
  6. 3GPP TS 38.300, NR; Overall description; Stage-2 — NTN and mobility procedures. 3GPP portal
  7. Geographic references: rounded illustrative coordinates. Natural Earth public domain boundaries, 1:110m, v5.1.2. Neither route tracks nor radio sites are surveyed. Water model assumptions are shared across countries.
  8. ITU-R Recommendation P.526, Propagation by diffraction. ITU
  9. ITU-R Recommendation P.833, Attenuation in vegetation. ITU
  10. ITU-R Recommendation P.453, The radio refractive index: its formula and refractivity data (effective Earth radius). ITU
  11. J. G. Walker, “Satellite constellations,” Journal of the British Interplanetary Society, vol. 37, pp. 559–572, 1984.
  12. SpaceX Starlink Gen1 FCC modification (SAT-MOD-20200417-00037): shells 53°/550 km (72×22), 53.2°/540 km, 70°/570 km (36×20), 97.6°/560 km (6×58 and 4×43). Summaries: APNIC, “Navigating Starlink's FCC paper trail”; J. McDowell, constellation models.
  13. X. Lin, S. Rommer, S. Euler, E. A. Yavuz, R. S. Karlsson, “5G from Space: An Overview of 3GPP Non-Terrestrial Networks,” IEEE Communications Standards Magazine, vol. 5, no. 4, pp. 147–153, 2021. doi:10.1109/MCOMSTD.011.2100038 · arXiv
  14. E. Juan, M. Lauridsen, J. Wigard, P. Mogensen, “Handover Solutions for 5G Low-Earth Orbit Satellite Networks,” IEEE Access, vol. 10, pp. 93309–93325, 2022. doi:10.1109/ACCESS.2022.3203189
  15. E. Juan, M. Lauridsen, J. Wigard, P. Mogensen, “Performance Evaluation of the 5G NR Conditional Handover in LEO-based Non-Terrestrial Networks,” IEEE WCNC, 2022. Aalborg University
  16. A. Hrovat, G. Kandus, T. Javornik, “A Survey of Radio Propagation Modeling for Tunnels,” IEEE Communications Surveys & Tutorials, vol. 16, no. 2, pp. 658–669, 2014. dblp
  17. PTS, Bättre mobil uppkoppling för tågresenärer i hela Sverige, PTS-ER-2023:5, 2023 — tunnel coverage by leaky feeder, Trafikverket's 165 rail tunnels, ≈30 dB car-body loss (X2, Regina). pts.se (PDF)
  18. RFS, RADIAFLEX® RLKX114-50CPR radiating cable datasheet — longitudinal and coupling loss measured per IEC 61196-4 (≈7.1–8.2 dB/100 m, ≈69 dB C50 at 3.4–3.8 GHz). rfsworld.com
  19. Eupen, RMC 1-5/8″-T “A” series radiating cable datasheet — 2.76 dB/100 m and 64 dB C50 at 900 MHz, measured in tunnel per IEC 61196-4. datasheet (PDF)
  20. M. A. Samad, S.-W. Choi, C.-S. Kim, K. Choi, “Wave Propagation Modeling Techniques in Tunnel Environments: A Survey,” IEEE Access, vol. 11, pp. 2199–2225, 2023. DOAJ
  21. Natural Earth, 1:50m Cultural Vectors — Admin 0 Countries (public domain), used for the locator map. naturalearthdata.com
  22. ITU-R Recommendation P.676, Attenuation by atmospheric gases and related effects. ITU
  23. ITU-R Recommendation P.838-3, PDF · equations 1, 4–5 and Table 5, Specific attenuation model for rain for use in prediction methods. ITU
  24. ITU-R Recommendation P.839, Rain height model for prediction methods (background only; local rain heights are assumed, not read from P.839 maps). ITU

Uplink and antenna method

UL received power = terminal conducted power + terminal gain + receiver gain − propagation, obstruction, enclosure, pointing and scan losses. Thermal noise = −174 + 10 log₁₀(B Hz) + NF. Availability requires an assumed −3 dB UL SNR threshold; rate uses the capped Shannon proxy. Default terminal power is 23 dBm over a 1 MHz noise-equivalent allocation, satellite receive gain 30 dBi / NF 3 dB, and terrestrial receive gains 15/15/24 dBi for n28/B20/n78 with NF 5 dB. These are adjustable or explicitly assumed values, not measured commercial terminal budgets. The allocation is an analytic bandwidth, not a full NR resource-grid implementation.

Satellite UL uses 1.99 GHz; reciprocal foliage/enclosure/fading and satellite scan loss are reused from the S-band access model. Its resource fraction is min(1, 5 MHz / (UL allocation × users)); return feeder capacity is a separate bottleneck. Terrestrial UL reconstructs the reciprocal path from DL reference-element power, using the correct tower EIRP or cable feed reference and no receive antenna gain at the cable port. Terrestrial UL uses the selected carrier frequency as a close-frequency path surrogate, without actual FDD frequency pairs or a separate UL channel realization.

The tracked terminal uses a local body-frame target vector, delayed geometric pointing, a mount offset, a projected-aperture scan term and A(θ)=min(12(θ/θ₃dB)², 30 dB). This quadratic form is inspired by TR 38.901 Table 7.3-1; the terminal parameters and scan model are illustrative. Scan and angular-rate limits are kinematic cutoffs, not a calibrated servo loop. Fixed broad and ideal-reference modes allow sensitivity comparisons. Tower links use an illustrative vertical omni pattern; cable coupling remains in the cable budget. Road/rail pitch follows physical grade; route heading is unchanged. Ship roll/pitch are deterministic sinusoids on simulation time, not hydrodynamics or measured sea state.

Voice generates 96 kbps each way; one-to-one call references generate 0.6 or 1.2 Mbps each way; video streaming generates 2.5 Mbps DL and 64 kbps UL control; telemetry generates 32 kbps each way. These are fixed CBR proxies. Call rates follow application guidance; packet deadlines, packet sizes and the 20% headroom diagnostic are assumptions. Best-effort attachment is separate from rate feasibility; full-rate admission remains selectable. Packet queues run independently in both directions, with shared handover availability; both rate requirements must be met for application satisfaction. Initial attachment, censored observations and artificial route returns retain separate accounting. A UL-budget failure can interrupt service and, after the existing illustrative timeout, trigger recovery; this is not a standards-complete 3GPP radio-link-monitoring implementation.

Inspiration and version

Inspired by the LEO TN–NTN Simulator at hsafi.com Updated