An SNR value belongs in a wireless link calculation only when its signal power and noise power use a consistent boundary: the same reference point, a defined bandwidth, and compatible measurement conditions. A spectrum analyzer reading, a receiver-reported SNR, and an SNR calculated from received power and noise figure are not automatically interchangeable.
For capacity estimates, the question is whether the SNR describes the noise in the modeled channel bandwidth. For receiver sensitivity, the question is whether the noise model and required SNR match the receiver’s operating mode and error-rate criterion. Resolve those definitions before comparing link options or interpreting a calculated margin.
Define the SNR Before Entering It
For average desired-signal power (P_s) and average noise power (P_n):
=10\log_{10}\left(\frac{P_s}{P_n}\right)
=P_{s,\mathrm{dBm}}-P_{n,\mathrm{dBm}}
Both powers must describe the same reference point and compatible filtering. In the band-limited additive white Gaussian noise, or AWGN, model, the denominator is the total noise power within the channel band—not the noise in one spectrum-analyzer bin. MIT’s communication theory notes explicitly define this relationship as (P/(N_0B)).
Record the following information with any SNR used for planning:
| Item | What must be established |
|---|---|
| Reference point | Antenna terminal, receiver input connector, receiver output, or a documented internal estimator location |
| Desired-signal power | Whether the value isolates the desired signal or includes noise and other transmissions |
| Bandwidth | Channel bandwidth, receiver equivalent noise bandwidth, or instrument integration bandwidth |
| Noise content | External noise only, external plus receiver-added noise, or noise plus interference |
| Operating state | Frequency, channel width, receiver gain state, modulation/coding mode, and relevant firmware or estimator configuration |
| Time basis | Active-burst power, continuous power, averaging duration, and whether signal and noise observations represent comparable conditions |
A value labeled “SNR” without these details may still be useful for trending one radio under unchanged settings. It is not yet a defensible input for comparing different measurement systems or receiver thresholds.
Keep Channel Bandwidth, Noise Bandwidth, and RBW Separate
Three bandwidths commonly appear in the same calculation, but they serve different purposes.
Channel bandwidth in a capacity calculation
For an ideal band-limited AWGN channel:
where:
- (C) is theoretical capacity in bits per second.
- (B_c) is the modeled channel bandwidth in hertz.
- (\gamma) is the linear signal-to-noise power ratio over that channel.
Convert an SNR expressed in decibels before using it:
This formula is a capacity bound for the stated channel model. It is not a prediction of Ethernet throughput, a supported modulation mode, or a radio’s measured performance. The AWGN definition assumes noise independent of the desired signal and a specified average signal-power constraint.
Equivalent noise bandwidth in a sensitivity calculation
Receiver noise depends on the filter’s equivalent noise bandwidth, (B_n). This is the bandwidth of an ideal rectangular filter that would pass the same white-noise power as the actual filter, after consistent gain normalization. It need not equal the nominal channel width or the filter’s 3 dB width. Keysight’s instrument documentation makes this distinction explicitly.
Using (B_n=B_c) can be a reasonable screening assumption when an ideal rectangular channel is intended. For an actual receiver, establish the applicable noise bandwidth from its filter configuration, manufacturer documentation, or a suitable measurement.
Resolution bandwidth in a spectrum measurement
Resolution bandwidth, or RBW, determines how a spectrum analyzer resolves frequency components. It does not normally represent the full communication channel.
An individual noise marker in a narrow RBW cannot be subtracted directly from full-channel signal power to obtain channel SNR. For approximately white noise with a calibrated density (N_{0,\mathrm{dBm/Hz}}):
=
N_{0,\mathrm{dBm/Hz}}
+
10\log_{10}\left(\frac{B_n}{1\ \mathrm{Hz}}\right)
If starting from an unnormalized noise-power reading, use the instrument’s effective noise bandwidth and documented corrections. Do not assume that displayed RBW always equals equivalent noise bandwidth: Keysight notes that swept analyzers generally specify a 3 dB width, whereas its 89600 VSA defines RBW through equivalent noise bandwidth.
When noise or interference varies significantly across the channel, use a properly configured integrated-power measurement rather than extrapolating one noise marker across the entire band.
Match the Reference Point and Count Receiver Noise Once
A useful thermal-noise sensitivity estimate at the receiver input is:
\approx
-174
+
10\log_{10}\left(\frac{B_n}{1\ \mathrm{Hz}}\right)
+
NF
+
\mathrm{SNR}_{\mathrm{req,dB}}
Here:
- (-174\ \mathrm{dBm/Hz}) is the approximate thermal-noise density at the standard reference temperature of 290 K.
- (NF) is the receiver system noise figure in decibels, referenced to the chosen input.
- (\mathrm{SNR}_{\mathrm{req,dB}}) is the required detector/demodulator SNR represented by this model.
- (P_{\min}) is the estimated minimum desired-signal power at that input.
Analog Devices derives this relationship from thermal noise, receiver noise figure, and the SNR required for detection at an acceptable error rate. Its source equation uses (10\log_{10}(kTB/1\ \mathrm{mW})) for thermal-noise power.
The corresponding input-referred effective noise floor is:
\approx
-174
+
10\log_{10}\left(\frac{B_n}{1\ \mathrm{Hz}}\right)
+
NF
The word “effective” matters. Receiver-added noise is represented as an equivalent input quantity; it is not necessarily an external noise power that a separate instrument would measure at the connector.
Apply the noise figure only where it belongs:
- If SNR is derived from desired-signal power and a thermal (kTB) floor, account for receiver-added noise through the appropriate system noise figure.
- If a receiver’s documented SNR estimator already includes its internal noise, do not subtract noise figure again.
- If a spectrum analyzer measures noise at the receiver input, the reading includes the analyzer’s own measurement contribution, not automatically the radio receiver’s noise contribution.
- If receiver-output powers are referred back to the input using gain, retain the receiver noise already present in that output measurement rather than adding it twice.
The standard (-174+10\log_{10}B_n+NF) expression assumes a 290 K source-noise reference. For a matched source with a materially different noise temperature (T_s), a more explicit input-referred model is:
where (T_e=(F-1)T_0), (F=10^{NF/10}), and (T_0=290\ \mathrm{K}). Receiver noise temperature and external source-noise temperature are separate quantities; the radio enclosure’s ambient temperature is not a substitute for both.
Feedline losses also require a consistent boundary. Received power at the antenna terminal must be moved through the feedline loss before comparison with a sensitivity specified at the radio connector. Alternatively, model the feedline and receiver together at the antenna reference plane. Do not use one boundary for signal power and the other for sensitivity.
Obtain a Measurement That Represents the Intended Ratio
A practical measurement sequence should preserve the channel definition and distinguish desired signal from unwanted power.
1. Fix the channel and instrument configuration
Specify center frequency, integration bandwidth, RBW, detector, averaging method, input attenuation, preamplifier state, and cable corrections. Check for overload and determine whether instrument noise materially contributes to the result.
Use power-domain averaging for average noise power. Keysight explains that averaging logarithmic readings does not produce the same result as averaging linear power; its documentation identifies errors as large as −2.51 dB for log-scale averaging of noise-like signals.
2. Measure unwanted power under representative conditions
If the desired transmitter can be muted, measure the same channel with the desired signal absent. Keep instrument settings unchanged and preserve the interference conditions as closely as possible.
A sequential signal-off observation cannot establish what the interference was during an earlier burst if channel occupancy changed. Where that difference matters, use an appropriate gated or time-correlated measurement and retain the observation timing.
3. Separate desired-signal power from total channel power
A signal-on channel-power measurement may contain desired signal, noise, and interference. If unwanted power remains unchanged between observations:
This subtraction must occur in linear power units. With input readings in dBm:
=
10\log_{10}
\left(10^{P_{\mathrm{on,dBm}}/10}
–
10^{P_{\mathrm{off,dBm}}/10}
\right)
The terms inside the parentheses are powers in milliwatts. The expression requires (P_{\mathrm{on}}>P_{\mathrm{off}}).
When the two readings are close, small errors or changes in unwanted power can dominate the difference. A displayed positive result alone is not evidence of a reliable desired-signal estimate. Check measurement uncertainty before using that result near a receiver threshold.
4. Label the denominator accurately
If the unwanted-power measurement includes interference, the result is:
where (P_i) is in-band interference power and all terms use compatible bandwidths and reference points.
Do not label this ratio thermal-noise SNR. Substituting SINR into the AWGN capacity equation is a planning approximation only when treating the interference as independent additive Gaussian noise is defensible. A narrowband interferer, intermittent collision, or blocker that changes receiver behavior is not fully described by one average noise-equivalent number.
Connect the Capacity and Sensitivity Calculations
The two calculations answer different questions:
| Calculation | Inputs that must be established | Meaning of the result |
|---|---|---|
| Shannon capacity | Modeled channel bandwidth and matching linear SNR | Ideal AWGN information-rate bound |
| Thermal-noise receiver sensitivity | Noise bandwidth, source-noise assumption, system noise figure, and required SNR | Estimated minimum received power for the stated detection criterion |
Use the Shannon Channel Capacity calculator for the bandwidth-and-SNR planning calculation and the Thermal-Noise Receiver Sensitivity calculator for the noise-budget calculation.
Their live input forms and executable implementations were not retrievable for this review. The calculations below therefore provide independent reference values, not a claim that either tool’s current defaults, unit conversions, or outputs have been checked.
A receiver sensitivity estimate needs a required SNR associated with an actual operating condition. Obtain that threshold from documentation or testing that identifies modulation, coding, channel width, error criterion, and relevant receiver configuration. A Shannon-derived minimum SNR is not a substitute for that equipment-specific threshold.
Worked Example: One Boundary for Both Calculations
Consider an illustrative single-channel model with these assumed inputs:
| Input | Assumed value |
|---|---|
| Desired-signal power at receiver connector | −75 dBm |
| Modeled channel bandwidth, (B_c) | 20 MHz |
| Equivalent noise bandwidth, (B_n) | 20 MHz |
| Source-noise temperature | 290 K |
| Receiver system noise figure | 5 dB |
| Required SNR for the hypothetical mode | 15 dB |
| Interference | None in this example |
The 15 dB threshold is an example assumption, not a manufacturer specification. The equal bandwidths represent an idealized rectangular channel.
Using the rounded (-174\ \mathrm{dBm/Hz}) thermal-noise reference:
=
-174+10\log_{10}(20{,}000{,}000)
=
-100.99\ \mathrm{dBm}
Including receiver noise figure:
=
-100.99+5
=
-95.99\ \mathrm{dBm}
The effective receiver SNR is:
=
-75-(-95.99)
=
20.99\ \mathrm{dB}
For the capacity calculation:
=
20{,}000{,}000\log_2(1+125.59)
\approx139.7\ \mathrm{Mbit/s}
This is the AWGN capacity of the modeled channel with receiver noise represented in its effective noise term. It is not an expected payload data rate.
For the sensitivity calculation:
=
-95.99+15
=
-80.99\ \mathrm{dBm}
The received-power margin above that assumed threshold is:
=
-75-(-80.99)
=
5.99\ \mathrm{dB}
The example therefore has approximately 6.0 dB of margin under its stated noise and threshold assumptions. That does not establish an outage probability or link availability: no fading distribution, time-varying interference, equipment tolerance, or site measurement has been included.
Widening the channel changes the SNR
Suppose the modeled bandwidth and noise bandwidth both increase from 20 MHz to 40 MHz while total desired-signal power remains −75 dBm and noise figure remains 5 dB.
Noise power rises by:
The effective noise floor becomes −92.98 dBm, and SNR falls to 17.98 dB. Recalculating the AWGN capacity gives approximately 239.8 Mbit/s—not twice the original 139.7 Mbit/s.
This comparison holds total received signal power constant. Holding SNR constant while doubling bandwidth describes a different condition: with unchanged white-noise density, total received signal power must also double. Any real radio comparison must establish its transmit-power behavior and mode-specific threshold rather than carrying the original SNR into a wider channel unchanged.
Decide Whether a Reported SNR Is Usable
Before accepting an SNR for a link calculation, make the following decisions:
- If the noise reading is per hertz or per RBW, convert it to the applicable channel noise power before forming the ratio.
- If signal and noise use different reference points, translate them through the documented gain and loss chain.
- If receiver-added noise is already included, do not apply noise figure again.
- If interference is present, identify the result as SINR and evaluate whether an AWGN approximation is appropriate.
- If the radio reports a per-chain, per-subcarrier, or post-processing estimate, use its documented definition; do not assume it represents one full-channel pre-detection SNR.
- If comparing against sensitivity, match the threshold’s bandwidth, modulation/coding mode, and error-rate criterion.
When those definitions cannot be established, retain the reading as a device-specific observation rather than converting it into a precise capacity or sensitivity conclusion. When they can be established, the resulting calculation supports a traceable planning comparison—with its assumptions attached to the result.