Link Budget: EIRP, Path Loss, Sensitivity, Margin

What Is a Link Budget and Why It Matters

Definition and purpose in RF system design

A link budget is an accounting exercise: you sum every gain and subtract every loss between a transmitter's power amplifier and a receiver's demodulator, and compare what's left against the minimum signal the receiver needs to work. It is arithmetic in decibels, nothing more exotic than that — but it is the single calculation that tells you, before any hardware exists, whether a radio link will close at the range you want.

Every term in the budget is additive in dB: transmit power, antenna gains, cable losses, path loss, fade margins. That additive property is what makes the link budget tractable on a spreadsheet instead of requiring a full electromagnetic simulation. You're not modeling the physics in detail — you're bookkeeping the net effect of each stage.

Why link budgets predict real-world range and reliability

A product spec that says "500 m range" is meaningless without a link budget behind it, because range is not a fixed property of a radio — it's the output of a calculation involving transmit power, antenna gain, frequency, receiver sensitivity, and the propagation environment. Change any one input and the achievable range changes with it. The link budget is what lets you say "500 m in free space at 915 MHz with these antennas and a 10 dB margin" instead of just "500 m," and the qualifiers are the part that actually matters when the product ships into a warehouse full of steel shelving instead of an open field.

When to perform link budget analysis in a project

Do it at the architecture stage, before silicon and antennas are chosen, because the link budget drives component selection rather than the other way around. Redo it whenever a transceiver IC, antenna, or enclosure changes — a plastic housing with a metal shield layer can cost you 3-5 dB of radiated power that a datasheet-only analysis would never catch. A final link budget, updated with measured (not simulated) antenna gain and measured transmit power, belongs in the design history file before you commit to production.

EIRP: Effective Isotropic Radiated Power

EIRP formula: transmit power + antenna gain

EIRP is the power an isotropic radiator would need to emit to produce the same signal strength, in the direction of peak antenna gain, as your real transmitter:

EIRP (dBm) = Pconducted (dBm) + Gantenna (dBi) - Lcable (dB)

A transmitter putting out 20 dBm (100 mW) into a cable with 1 dB of loss, feeding a 3 dBi antenna, produces an EIRP of 22 dBm. That 22 dBm, not the 20 dBm the PA datasheet advertises, is the number that goes into the path-loss calculation.

Relationship between conducted power and radiated power

Conducted power is what a power meter reads at the antenna connector. Radiated power is what actually leaves the system, and the two diverge because of antenna gain (which can add several dB in a preferred direction) and because of losses in cables, connectors, and matching networks between the PA and the antenna feedpoint. A high-gain antenna doesn't create energy — it concentrates the same conducted power into a narrower solid angle, trading beamwidth for range along the main lobe. For a detailed look at how a two-port matching or filter network inserted in that feed path affects the power actually delivered to the antenna, the two-port network parameter formalism (S, Z, Y, ABCD) is the right tool for computing insertion loss and mismatch loss precisely rather than assuming a flat dB figure.

Regulatory limits and their impact on system design

Regulatory bodies cap EIRP, not conducted power — FCC Part 15, ETSI EN 300 220, and similar regional rules set a maximum radiated power or power spectral density for unlicensed bands. This means a higher-gain antenna can force you to reduce conducted power to stay compliant, which is a real design trade-off: a 6 dBi antenna and a 14 dBm PA may hit the same EIRP ceiling as a 0 dBi antenna and a 20 dBm PA, but with very different current draw and PA linearity requirements. Always check the specific regional limit that applies to your band before locking in antenna gain.

How to measure and verify EIRP in practice

EIRP is verified in a calibrated chamber (anechoic or OATS) by measuring field strength at a known distance and converting back to EIRP using the free-space formula, or substitution against a reference antenna. Conducted power alone, measured with a power meter at the antenna connector, is not sufficient for compliance testing, but it is useful for catching gross errors early — if conducted power plus the antenna's datasheet gain doesn't roughly match a chamber measurement, suspect cable loss, a bad connector, or an antenna that isn't resonant at your operating frequency. Confirming the match and conducted power accurately requires the kind of de-embedded measurement a properly calibrated VNA provides; see VNA calibration methods for how SOLT, TRL, SOLR, and LRM remove fixture and cable error so the number you measure reflects the device, not the test setup.

Path Loss: Signal Attenuation Over Distance

Free-space path loss model and Friis transmission equation

The Friis transmission equation gives free-space path loss (FSPL) between two isotropic antennas:

FSPL (dB) = 20*log10(d) + 20*log10(f) + 20*log10(4*pi/c)

Or in the common engineering form with d in meters and f in MHz:

FSPL (dB) = 32.44 + 20*log10(d_km) + 20*log10(f_MHz)

At 915 MHz and 1 km: FSPL = 32.44 + 20log10(1) + 20log10(915) ≈ 32.44 + 0 + 59.2 ≈ 91.6 dB. At 2.4 GHz and the same 1 km, FSPL ≈ 100.4 dB — nearly 9 dB worse purely from the frequency term.

Frequency dependence: why higher bands lose more energy

The 20*log10(f) term exists because the equation is normalized to fixed-gain antennas; a physically fixed-aperture antenna's gain rises with frequency, which cancels much of this loss in practice. But for a given antenna gain spec in dBi, higher frequency genuinely costs more path loss, which is why sub-GHz ISM bands (433, 868, 915 MHz) remain popular for long-range, low-power links despite narrower available bandwidth.

Distance and geometry: near-field vs. far-field propagation

FSPL is only valid in the far field, beyond the Fraunhofer distance 2D²/λ for an antenna of aperture D. Inside that range — common for RFID, NFC, and some proximity sensors — the field doesn't fall off as a clean inverse-square law, and a link budget built on Friis will be wrong. For near-field links, use inductive or capacitive coupling models instead, and recognize that the governing physics is the full Ampere-Maxwell relation rather than the radiating far-field simplification; see Ampere-Maxwell law for the underlying field equations when you need to go beyond the far-field approximation.

Real-world effects: multipath, shadowing, and environmental factors

Free space is the best case. Real deployments add multipath fading (destructive interference from reflections, causing deep fades of 10-20 dB at specific points), shadowing from walls and terrain (modeled statistically with log-normal shadowing, typically 3-12 dB standard deviation depending on environment), and body or hand effects for wearables. None of these are captured by Friis — they're the reason link margin exists as a separate budget line rather than being folded into path loss.

Receiver Sensitivity: The Weakest Signal That Matters

Sensitivity vs. noise figure: how noise limits detection

Receiver sensitivity is the minimum input signal power at which the receiver still demodulates data at an acceptable error rate. It's fundamentally set by the noise floor: thermal noise power is kTB (-174 dBm/Hz at room temperature), the receiver's noise figure (NF) adds on top of that, and the required SNR for your modulation sets the margin above the noise floor the signal must clear:

Sensitivity (dBm) = -174 + 10*log10(BW_Hz) + NF (dB) + SNR_required (dB)

A receiver with NF = 6 dB, 200 kHz bandwidth, and an SNR requirement of 8 dB: -174 + 53 + 6 + 8 = -107 dBm.

SNR requirement and modulation scheme dependency

SNR_required depends on modulation and coding: simple OOK/ASK may need 10-15 dB SNR for a usable bit error rate, while robust spread-spectrum schemes (LoRa's chirp spread spectrum, DSSS) can demodulate below 0 dB SNR at the cost of data rate. This is why a LoRa radio at SF12 achieves sensitivity near -137 dBm in a narrow bandwidth while a Bluetooth Low Energy receiver, optimized for throughput over range, sits closer to -95 to -100 dBm.

How to extract sensitivity from datasheets

Datasheets usually state sensitivity directly for a specified data rate, bandwidth, and bit-error-rate or packet-error-rate target — read the fine print, because a headline sensitivity number is often quoted at the most forgiving (lowest data rate, highest coding gain) setting. If your application needs a higher data rate, look for the sensitivity-vs-data-rate table rather than using the headline figure.

Trade-offs between sensitivity and bandwidth

The 10*log10(BW) term means doubling bandwidth costs 3 dB of sensitivity. This is the core tension in modulation choice: wider bandwidth buys higher data rate but directly degrades the noise floor, while narrowband or spread-spectrum techniques buy sensitivity at the cost of throughput or spectral occupancy.

Link Margin: The Safety Factor

Margin equation: EIRP − Path Loss − Receiver Sensitivity

Link Margin (dB) = EIRP - Path Loss - Sensitivity

Using the earlier numbers: EIRP = 22 dBm, FSPL at 1 km/915 MHz ≈ 91.6 dB, sensitivity = -107 dBm:

Margin = 22 - 91.6 - (-107) = 37.4 dB.

Why positive margin is essential for reliable links

A margin of exactly 0 dB means the link closes only under the idealized free-space assumption with no fading, no interference, and no manufacturing tolerance — a link that works once on an open bench and fails the moment someone walks between the nodes. Positive margin is headroom against everything Friis doesn't model.

Typical margin targets for different applications

Static, line-of-sight, low-criticality links (a garage door sensor) can tolerate 10-15 dB margin. Mobile or NLOS indoor links typically target 20-30 dB to survive shadowing and multipath fades. Safety-critical or regulatory-mandated links (alarm panels, medical telemetry) often specify 30-40 dB or a statistical outage probability requirement instead of a flat number.

Fading margin vs. implementation margin

Split your margin budget into fading margin (statistical allowance for multipath and shadowing, drawn from a propagation model or site survey) and implementation margin (allowance for PA aging, temperature drift, manufacturing variance in antenna matching, and the gap between datasheet-typical and worst-case sensitivity). Treating all margin as one undifferentiated number hides which risk you're actually covering.

Building a Complete Link Budget Table

Step-by-step construction: transmitter to receiver

Build the table as a running sum from PA output to demodulator input: conducted TX power, minus TX cable/connector loss, plus TX antenna gain (giving EIRP), minus path loss, minus any additional environmental loss allowance, plus RX antenna gain, minus RX cable loss, giving received power at the receiver input. Compare that to sensitivity for margin.

Accounting for cable loss, connector loss, and component insertion loss

Each connector typically costs 0.1-0.3 dB, each coax run depends on length and dielectric (datasheet cable attenuation is usually given in dB/m at a reference frequency — scale it to your actual frequency, since loss rises roughly with sqrt(f) for skin-effect-dominated cable). Filters, baluns, and switches in the signal path each contribute insertion loss from their own datasheets — don't estimate these, look them up, and verify the match with S-parameter data referenced to a properly calibrated measurement.

Using dB and dBm notation correctly

dBm is an absolute power level referenced to 1 mW; dB is a ratio. You add and subtract dB values freely, but you can only add a dB (ratio) term to a dBm (absolute) term — never add two dBm values together. A common error is summing antenna gains in dBi as if adding power; gain modifies the signal multiplicatively (additively in log units), which is correct, but mixing up which column is absolute versus relative is the most frequent arithmetic mistake in a first-pass link budget.

Sensitivity to frequency and distance assumptions

Because path loss scales with 20log10(f) and 20log10(d), small changes in assumed operating frequency or deployment range swing the margin significantly — moving from 868 MHz to 2.4 GHz for the same range costs roughly 9 dB, equivalent to needing 8x the transmit power to recover. Always state which frequency and distance the budget assumes, and recompute rather than reuse an old table when either changes.

Practical Design Decisions from Link Budget Results

When to increase transmit power vs. antenna gain

Raising conducted power is simple but linear in both regulatory headroom and battery drain (doubling power costs 3 dB and roughly doubles PA current draw). Raising antenna gain is "free" in terms of power consumption but trades beamwidth — a higher-gain antenna narrows coverage angle and may demand an enclosure redesign or external antenna placement. If the application is directional or fixed-orientation, gain is usually the cheaper lever; for omnidirectional mobile use, power or receiver sensitivity improvements are often more practical.

Optimizing receiver noise figure for range extension

Every dB shaved off receiver noise figure is a dB of extra margin, equivalent to doubling transmit power roughly every 3 dB — but achieving low NF usually means a dedicated LNA ahead of the mixer, adding cost, board area, and current draw. Compare the cost of 1 dB of NF improvement against 1 dB of extra PA output power; for battery-powered nodes, the receive-side improvement often wins since it's a one-time power cost versus a per-transmission one.

Choosing modulation schemes based on available margin

If the margin calculation comes up short, switching to a more robust modulation (lower-order, more coding gain, spread spectrum) recovers sensitivity at the cost of data rate — often the cheapest fix available, since it requires no hardware change on an SDR-capable transceiver, only a configuration change.

Iterative refinement for cost-constrained designs

Link budget design is iterative: compute margin, identify the weakest term, evaluate the cost of improving it (power, antenna, NF, modulation), and re-run the table. Converge when margin meets target at acceptable BOM cost and power budget — then lock the assumptions and verify them against bench and chamber measurements before committing to production.