NTN Link Budget & Coverage
Why the satellite link runs at very low SNR — free-space path loss over the slant range, satellite EIRP and G/T, the TR 38.821 reference budgets, and a worked handheld example.
A terrestrial cell is a few kilometres away; an NTN satellite is hundreds to nearly thirty-six thousand kilometres away, with a strictly limited power budget on both ends. The consequence is unavoidable: NTN operates at very low signal-to-noise ratio — often near or even below 0 dB — and every design choice, from waveform to HARQ, falls out of that one fact. This page builds the link budget from the ground up: free-space path loss, the downlink carrier-to-noise chain, why the handset uplink is the direction that actually limits coverage, and how minimum elevation angle sets the worst case. The reference scenarios come from TR 38.821, which defines link budgets for handheld and VSAT terminals across GEO and LEO in S-band and Ka-band.
Introduction
A link budget is the accountant's view of a radio link: a running sum, in decibels, of every gain and loss between the transmitter's power amplifier and the receiver's demodulator. Start with the transmitted EIRP, subtract the free-space spreading loss and everything else the path takes away, add back the receiver's figure of merit, and what remains is the carrier power relative to the noise floor — the signal-to-noise ratio the decoder actually has to work with. If that number sits above the threshold the chosen modulation and coding need, the link closes; if not, it does not, no matter how clever the higher layers are.
In a terrestrial network the budget usually closes with comfortable margin, so it is a background calculation. In NTN it is the foreground constraint. The distances are enormous, the satellite platform is power- and aperture-limited, and one end of the link is a pocket-sized handset. Those three facts squeeze the margin down to a few dB — sometimes to nothing — and the whole air-interface design (robust MCS, repetition, long TTIs, low-PAPR uplink, HARQ changes) is a direct response to that squeeze.
You meet the link budget the moment coverage is being dimensioned: it decides which band a service can use, which orbit and terminal class are viable, what data rate is achievable at the edge of a beam, and which robustness features are mandatory rather than optional. Read this page as the "why" underneath most of the rest of the NTN section — it is the number every other NTN mechanism is trying to protect.
On this page
Why the link budget rules NTN
In plain words: shining a torch across a room lights up a wall; shining the same torch at a hilltop a kilometre away barely registers. The light did not weaken — it just spread out over a vastly larger area by the time it arrived. A satellite is that distant hilltop, and its transmit power is a torch that cannot be made much brighter. NTN is the art of reading a message by torchlight from very far away, so every scrap of energy has to be conserved and every source of loss accounted for.
Received power falls with the square of distance, and NTN distances are astronomical compared with terrestrial ones. A satellite also has a finite platform: limited DC power, limited antenna aperture, so its EIRP (effective isotropic radiated power) cannot simply be cranked up to compensate. Put a large path loss against a capped transmit power and a modest receiver, and the arithmetic lands you at an SNR of a few dB at best — sometimes negative.
A link budget is the running tally, in dB, from transmit EIRP down through path loss and other losses to the received carrier power, compared against the noise floor to give C/N0 and finally SNR.
It tells you whether a link closes at all, and with how much margin. In NTN the margin is thin, so the budget dictates the achievable data rate, the coverage edge, and which robustness features are mandatory.
TR 38.821 fixes reference terminals (handheld, VSAT), orbits (LEO, GEO) and bands (S, Ka), then tabulates EIRP, G/T and losses so every proposal is measured against the same worst case.
The governing fact: NTN is a low-SNR regime by construction. You cannot out-power the path loss from a satellite, so the system is engineered to work at low SNR rather than to avoid it.
Free-space path loss and the elevation penalty
The dominant term is free-space path loss (FSPL), the spreading loss of a wave over distance. Physically it is the Friis relationship — the fraction of radiated power an isotropic receiver captures shrinks with the square of both distance and frequency — recast into decibels. In the convenient form with frequency in MHz and distance in km:
The constant 32.45 simply absorbs the unit conversions and the 4π geometry when frequency is in MHz and distance in km; use metres and Hz and the constant changes but the physics does not. Two things jump out. First, path loss grows with frequency: doubling the frequency adds 6 dB, and a factor of ten adds 20 dB — which is exactly why moving from S-band (2 GHz) to Ka-band (20 GHz) costs a flat +20 dB for the same geometry. Second, distance is not the orbital altitude but the slant range: when the satellite is low on the horizon, the line-of-sight path is far longer than when it is straight overhead, so lower elevation means more path loss and a longer atmospheric path.
The slant range itself follows from spherical geometry — with Earth radius RE, orbital altitude h and elevation angle ε, the range is d = √((RE+h)² − (RE·cosε)²) − RE·sinε — which is why the range grows steeply as ε drops toward the horizon and the satellite's signal skims through progressively more atmosphere.
| Scenario | Slant range | FSPL @ S-band (2 GHz) | FSPL @ Ka-band (20 GHz) |
|---|---|---|---|
| LEO 600 km, overhead (90° elevation) | 600 km | ≈ 154.0 dB | ≈ 174.0 dB |
| LEO 600 km, 30° elevation | ≈ 1075 km | ≈ 159.1 dB | ≈ 179.1 dB |
| GEO 35,786 km | 35,786 km | ≈ 189.6 dB | ≈ 209.6 dB |
Notice that dropping the same LEO satellite from overhead to 30° elevation adds about 5 dB of path loss purely from the longer slant range — and GEO sits roughly 35 dB worse than an overhead LEO in the same band. The minimum elevation angle a system supports (often 10°–30°) therefore defines the coverage edge: it is the longest slant range, the highest FSPL, and the most atmospheric attenuation the link must still close. Every budget is dimensioned to that worst-case geometry, because a UE at the edge of the beam is the one most likely to drop.
The downlink C/N0 chain, worked
Path loss is only the biggest term. The full downlink carrier-to-noise-density budget chains EIRP against all losses and the receiver figure of merit:
SNR (dB) = C/N0 − 10·log10(BWHz)
Here k is Boltzmann's constant, −228.6 dBW/Hz/K; L_other gathers the losses that are not free-space spreading — atmospheric absorption, scintillation, shadowing and polarisation mismatch; and G/T is the receiver figure of merit (antenna gain over system noise temperature), which for a handheld is small and negative. Once you have C/N0, subtracting the log of the occupied bandwidth gives the SNR the demodulator actually sees. The reason the budget is expressed against noise density N0 first, and only later against bandwidth, is precisely so you can see the rate-versus-margin trade explicitly: C/N0 is fixed by geometry and hardware, and you choose how much of it to spend on bandwidth.
A representative handheld downlink at S-band, 30° elevation, makes the low-SNR reality concrete. The EIRP and G/T below are representative values from the TR 38.821 reference scenarios, not fixed spec constants — the method is exact, the vendor numbers illustrative:
| Term | Value | Note |
|---|---|---|
| Satellite EIRP | +34.0 dBW | representative, TR 38.821 reference scenario |
| − FSPL | −159.1 dB | S-band, 30° elevation (slant ≈ 1075 km) |
| − Lother | −3.0 dB | atmospheric + scintillation + polarisation |
| + G/T | −31.6 dB/K | handheld, representative (TR 38.821) |
| − k | +228.6 dBHz | Boltzmann, −(−228.6) |
| = C/N0 | 68.9 dBHz | carrier-to-noise density |
| − 10·log10(5 MHz) | −67.0 dB | occupied bandwidth |
| = SNR | ≈ 1.9 dB | a few dB — thin margin |
Under two dB of SNR, and that is the downlink with a satellite doing the transmitting. Widen the bandwidth for more throughput and the SNR falls further, dB for dB, because the same C/N0 is spread over more noise bandwidth. That trade — rate against margin — is the everyday tension in NTN scheduling. And note there is no separate margin line reserved here: at 1.9 dB the operating point already sits close to the threshold of a robust QPSK code rate, so any fade eats directly into decodability. That is why the link-margin term, which is comfortable and often unstated in terrestrial budgets, becomes something NTN actively fights to create.
Why the handset uplink is the hard direction
The downlink was tight; the uplink is tighter, and it is the direction that actually limits coverage for a handheld. The reason is the terminal. A handheld has roughly a 0 dBi antenna and about 23 dBm (0.2 W) maximum transmit power — a rounding error next to a satellite's EIRP and large receive aperture. On the downlink the big antenna and power sit at the satellite; on the uplink the weak end is doing the transmitting, so the same FSPL bites into a far smaller starting EIRP.
The uplink budget starts from a handheld's ~0 dBi, ~23 dBm EIRP, so after NTN path loss the received uplink carrier is weaker and the link closes with less margin than the downlink.
You cannot enlarge a phone's antenna or battery to satellite scale. The asymmetry between a powerful satellite and a pocket-sized terminal makes the uplink the coverage-limiting direction.
Systems fight back with low bands and heavy robustness: S/L-band keeps FSPL and antenna penalty manageable (see NTN spectrum), and Rel-18 adds uplink coverage and capacity enhancements (see NTN evolution).
Frequency choice is deliberate here. Every octave up costs 6 dB of FSPL and a stiffer antenna-efficiency penalty on a tiny terminal, so direct-to-handset NTN uses low bands (S/L) precisely to keep the uplink survivable, leaving Ka-band for VSAT terminals with real dishes. This is the budget reason behind the band split covered in NTN spectrum.
TN ↔ NTN: in a terrestrial cell the base station and the handset are only kilometres apart, FSPL is 120–140 dB, and the downlink is usually the limiting direction (the base station serves many UEs from a shared power budget). In NTN the numbers invert: FSPL is 150–210 dB, margin is measured in single dB, and the uplink is limiting because the weak end — the phone — is now transmitting across a satellite path. The same link-budget equation applies to both; only the term magnitudes change, and that change is enough to flip which direction fails first.
Designing for low SNR
Once you accept that the operating point is a few dB or less, a familiar set of choices follows — the same tools terrestrial coverage extension uses, pushed harder:
- Robust MCS. Low-order modulation and low code rate, trading throughput for the ability to decode at low SNR.
- Repetition / coverage enhancement. Sending the same content multiple times to accumulate energy at the receiver.
- Longer TTIs. Spreading a transport block over more time to collect more energy per bit.
- DFT-s-OFDM on the uplink. The low-PAPR waveform lets the handheld power amplifier run closer to saturation, squeezing more useful EIRP out of that 23 dBm budget.
- HARQ-feedback-less operation with RLC ARQ. Over a long round trip, waiting for HARQ feedback stalls the pipeline, so NTN often disables HARQ feedback and leans on RLC ARQ for reliability instead — see HARQ over a long round-trip.
One thread: robust MCS, repetition, long TTIs and a low-PAPR uplink are all the same move — spend time and simplicity to buy energy per bit, because the budget will not give you SNR for free.
⚠ Common pitfalls / gotchas
- Budgeting at overhead, deploying at the horizon. Sizing coverage against the 90° slant range hides the ~5 dB (LEO) or more that the minimum-elevation geometry actually costs — the beam edge is where UEs drop.
- Forgetting the bandwidth term. A healthy C/N0 can still yield a negative SNR once spread over too wide a channel; throughput ambitions must be reconciled against 10·log10(BW).
- Treating Lother as a constant. Scintillation and rain fade (especially in Ka-band) are time-varying and can swing several dB, so a budget with no fade margin closes on paper and fails in the field.
- Assuming the downlink limits coverage. Terrestrial instinct says downlink; in direct-to-handset NTN the uplink is the binding constraint, so link-budget effort belongs there.
- Reading EIRP/G/T table values as spec constants. They are reference-scenario illustrations in TR 38.821; real vendor payloads and terminals differ, so re-run the chain with actual hardware figures.
Summary
The link budget is the one number the entire NTN air interface is built to protect. Chain it in order and the story tells itself: start from EIRP, subtract FSPL (which grows 20 dB per decade of frequency and rises as elevation drops toward the horizon), subtract the other losses, add the receiver's G/T, remove Boltzmann's constant to get C/N0, then divide out the bandwidth to get the SNR the decoder sees. For a representative handheld at S-band and 30° elevation that SNR lands under 2 dB — the low-SNR signature of NTN.
Two structural facts follow. First, the uplink is the coverage-limiting direction, because a ~0 dBi, ~23 dBm handset is transmitting across the same brutal path that a powerful satellite handles easily on the downlink — which is why direct-to-handset service lives in low S/L-band and Ka-band is reserved for VSAT dishes. Second, the minimum elevation angle sets the worst case, because it is the longest slant range and the highest path loss the link must still close.
Everything else in the NTN section — robust MCS, repetition, long TTIs, low-PAPR DFT-s-OFDM, and the HARQ feedback changes — is a way of spending time and simplicity to buy back the energy-per-bit the budget refuses to hand out for free. When an NTN link fails to close, walk the budget term by term: the first line that is worse than you assumed is your root cause.
Q. Why does NTN operate at such low SNR?
A. Because path loss scales with distance and NTN distances are 600 km to 35,786 km — so FSPL runs 154 dB to 190 dB at S-band — while the satellite's EIRP and the terminal's G/T are both bounded. Chain EIRP − FSPL − L_other + G/T − k and divide by the bandwidth and you land at an SNR of only a few dB, sometimes below 0 dB. You cannot out-power the path, so the system is built to work at low SNR rather than to escape it.
Q. Why is the handheld uplink the limiting direction, and how much does Ka-band cost versus S-band?
A. On the downlink the powerful, large-aperture satellite transmits; on the uplink a handheld with ~0 dBi and ~23 dBm transmits, so the same path loss eats into a far smaller EIRP and the uplink closes with less margin. That is why direct-to-handset NTN uses low S/L-band. Ka-band (20 GHz) costs a flat +20 dB of FSPL versus S-band (2 GHz) for the same distance — a factor of ten in frequency is 20·log10(10) = 20 dB — which is why Ka is reserved for VSAT terminals with real antenna gain.
Q. Why does the minimum elevation angle set the worst-case budget?
A. Elevation controls slant range: overhead a 600 km LEO is 600 km away, but at 30° it is about 1075 km away, adding roughly 5 dB of FSPL plus a longer atmospheric path. The lowest supported elevation is therefore the longest range, highest path loss and most attenuation the link must still close, so budgets are dimensioned to that edge-of-coverage geometry.
Related NTN topics
The budget decides the band, the geometry and the reliability scheme — follow each thread from here.