Noise floor calculator
Every receiver sits on a floor of thermal noise — kTB, which is −174 dBm in each hertz of bandwidth at 290 K. Widen the channel and the floor rises by 10 dB per decade of bandwidth: 10 kHz sits at −134 dBm, 40 MHz at −98 dBm, 500 MHz at −87 dBm. Add the receiver’s own noise figure and the SNR the modulation needs, and you have its sensitivity.
2–3 dB for a good VHF/UHF receiver, 4–8 dB for integrated Wi-Fi and cellular front ends.
−20 dB for LoRa SF12, 5–10 dB for narrow FM or QPSK, 25 dB for 64-QAM.
290 K is the standard reference; cooled LNAs go lower.
The vertical scale is signal level in dBm. The dashed line is the thermal floor kTB for the bandwidth and temperature you set. The amber block above it is the receiver’s own noise figure, the green block above that is the SNR the modulation needs, and the solid green line on top is the sensitivity that results. The labels on the left are familiar floors for comparison.
kTB covers thermal noise only. Below roughly 100 MHz, atmospheric and man-made noise usually set a higher floor than the receiver does (see ITU-R P.372).
How it works
- 01
Start with physics
Any resistor at temperature T delivers a noise power of k·T·B into a matched load, whatever it is made of. At the standard reference temperature of 290 K that is −174 dBm in every hertz. No amplifier, filter or antenna design removes it — it is the floor the whole link budget stands on.
- 02
The wider the channel, the higher the floor
Noise power is proportional to bandwidth, so the floor rises by 10·log10(B). At 290 K a 10 kHz SSB channel sits at −134 dBm, a 40 MHz Wi-Fi channel at −98 dBm, and a 500 MHz wideband capture at −87 dBm. Same physics, 47 dB apart. Doubling the bandwidth costs 3 dB; ten times the bandwidth costs 10 dB. That gap is the whole reason narrowband modes reach so much farther than wideband ones at the same transmit power.
- 03
Add the receiver’s own noise
Noise figure is how many dB the receiver adds to that floor. A good low-noise amplifier at VHF or UHF manages 1–3 dB; integrated Wi-Fi and cellular front ends are typically 4–8 dB. The first stage dominates, which is why the LNA belongs at the antenna rather than at the far end of the feedline.
- 04
Add what the modulation needs
Every scheme needs some signal-to-noise ratio to decode: roughly −20 dB for LoRa at SF12, 5–10 dB for narrow FM or QPSK, 25 dB for 64-QAM and above 30 dB for 256-QAM. Floor plus noise figure plus that required SNR, Smin = N + NF + SNRreq, is the weakest signal the receiver can actually use.
Formulas
- k — Boltzmann constant, 1.380 649×10−23 J/K
- T — noise temperature, K (290 K is the standard reference)
- B — bandwidth, Hz
- Smin — receiver sensitivity, dBm
- NF — receiver noise figure, dB
- SNRreq — signal-to-noise ratio the modulation needs, dB
Worked example
- N = −174 + 10·log10(20 000 000) = −174 + 73 = −101 dBm
- + NF 6 dB → −95 dBm effective floor
- + SNR 25 dB (64-QAM) → sensitivity ≈ −70 dBm
- → that is why “two bars” at −75 dBm still will not carry 64-QAM
FAQ
- Why is the noise floor −174 dBm per hertz?
- It is k·T for the standard reference temperature: 1.380 649×10−23 J/K times 290 K is 4.00×10−21 W per hertz, which is −173.98 dBm/Hz. 290 K is roughly 17 °C, close to a typical outdoor environment, and it is the temperature noise figures are defined against. A cooled receiver front end genuinely lowers this number.
- Can a receiver hear a signal below the noise floor?
- Yes, when it spreads the signal over time or bandwidth and integrates. GPS arrives at about −128 dBm, roughly 17 dB below the thermal floor in its 2 MHz band, and correlation recovers it. LoRa at spreading factor 12 demodulates at −20 dB SNR. Both trade data rate for that processing gain.
- What is a good noise figure?
- Below 1 dB is excellent and needs a dedicated low-noise amplifier; 2–3 dB is a good VHF/UHF receiver; 4–8 dB is normal for integrated Wi-Fi and cellular chipsets. Each dB of noise figure costs exactly one dB of sensitivity, and the first stage sets nearly all of it, so feedline loss ahead of the LNA adds to the noise figure dB for dB.
- Why do narrowband modes reach farther?
- Because bandwidth is the term you control in N = −174 + 10·log10(B). Going from a 500 MHz capture (−87 dBm) down to a 40 MHz channel (−98 dBm) and on to a 10 kHz SSB filter (−134 dBm) drops the floor by 47 dB — the same benefit as multiplying transmitter power by 50 000, and it costs nothing. Narrow modes buy range with the one term nobody can argue with, and pay in throughput.
- How can a real-time spectrum analyzer capture hundreds of MHz?
- Because the noise floor it displays follows the FFT resolution bandwidth (RBW), not the capture span. An instrument digitizing 800 MHz at once still processes it as thousands of narrow bins: at a 10 kHz RBW each bin sits near −134 dBm plus the instrument’s noise figure, not at the −87 dBm a single 500 MHz-wide detector would see. That is why DANL (displayed average noise level) is quoted in dBm/Hz or at a stated RBW — narrow the RBW and the displayed floor drops 10 dB per decade, paid for in sweep or processing time. Capturing 800 MHz does not mean detecting each signal against an 800 MHz-wide floor.
Sensitivity is one end of the link budget. Simulate whether the signal that actually arrives clears it — over real terrain, free in your browser.
Simulate received levels →Related guides
- Plan radio coverage (RF / ITM) — the “Min signal (dBm)” setting is exactly this sensitivity — anything weaker is not drawn
Sources & further reading
- ITU-R Recommendation P.372 — Radio noise ↗ — external noise, which dominates below about 100 MHz
- Keysight — Fundamentals of RF and Microwave Noise Figure Measurements (AN 57-1, PDF) ↗ — noise figure, noise temperature and cascade behaviour
- Johnson–Nyquist noise — Wikipedia ↗ — the thermal noise the kTB formula describes
Related tools
Link budget calculator
Received power and fade margin from TX power, gains, losses and path loss.
LoRa airtime calculator
Time on air, duty-cycle limits and data rate from spreading factor, bandwidth and payload.
dBm ↔ watt converter
Convert between dBm, milliwatts and watts, with common reference levels.
Wi-Fi channel to frequency converter
Center frequencies and widths for 2.4, 5 and 6 GHz channels.