Before this:Decibels & signal powerHow an SDR receiver works
Noise, SNR & sensitivity
Key takeaways Every receiver sits on top of a noise floor — a background hiss from thermal noise, the environment, and its own electronics. A signal decodes not because it’s strong in absolute terms but because it clears that floor by enough margin: its SNR (signal-to-noise ratio, measured in dB). Sensitivity is how weak a signal a receiver can still pull out of the noise. Crucially, more gain doesn’t add SNR — see gain & AGC.
Beginners naturally reach for “signal strength” as the thing that matters. It isn’t. A booming signal buried in an equally booming noise floor won’t decode, while a faint one sitting in silence will. This lesson explains the quantity that actually decides a decode — SNR — and the receiver properties that set the floor beneath it.
The noise floor
Point an antenna at empty spectrum and the receiver still shows a level. That’s the noise floor, and it has three sources.
The first is thermal noise — the random jostling of electrons in any conductor above absolute zero. It’s unavoidable and its power depends only on temperature and bandwidth, captured by the expression kTB (Boltzmann’s constant × temperature × bandwidth). At room temperature this works out to roughly −174 dBm/Hz — the noise power in a single hertz of bandwidth. That figure is the floor beneath the floor: no receiver can do better.
The second is man-made noise the antenna picks up — switching power supplies, LED lighting, chargers, motors, and general urban electrical hash. In a noisy environment this often dominates the thermal contribution entirely.
The third is the receiver’s own noise, added by its amplifiers and mixers (more on that below).
Notice the bandwidth term in kTB: the wider your receive bandwidth, the more noise power you collect. Doubling bandwidth adds about 3 dB of noise. This is why narrow, matched filtering ahead of the demodulator helps — you want the noise only from the channel you care about, not the whole band.
Signal-to-noise ratio (SNR)
SNR is the gap, in dB, between your signal and the noise floor. If a signal peaks at −80 dBm and the floor sits at −100 dBm, the SNR is 20 dB. That difference — not the −80 dBm by itself — is what the demodulator has to work with.
This is why absolute dBm is a poor predictor of success. A −40 dBm signal sounds strong, but if local noise or overload has pushed the floor up to −45 dBm, you have only 5 dB of SNR and it won’t decode. Meanwhile a −95 dBm signal over a quiet −115 dBm floor has a healthy 20 dB and locks fine. The ratio wins, every time.
How much SNR do you need? It depends on the protocol, but as a rough guide digital voice systems need on the order of 10–20 dB to lock cleanly. The real gate at the receiver is the constellation — the demodulated symbols have to land close enough to their ideal positions (low error vector magnitude, EVM) for the error correction to recover them. SNR is just the convenient meter reading that tracks it.
Sensitivity & the minimum discernible signal (MDS)
Sensitivity is a receiver’s ability to recover weak signals — how far down toward the noise floor it can still hear. It’s often quoted as the minimum discernible signal (MDS): the input power in dBm that produces an output just detectable above the noise, conventionally the point where the signal equals the noise (a few dB of SNR).
MDS follows directly from the noise-floor discussion. It’s roughly the thermal floor (−174 dBm/Hz), plus 10·log₁₀ of the bandwidth in hertz, plus the receiver’s noise figure, plus whatever small SNR you define “discernible” to mean. Narrow the bandwidth or lower the noise figure and the MDS drops (improves). A receiver specified at, say, −120 dBm MDS in a given bandwidth is more sensitive than one at −110 dBm.
The practical upshot: sensitivity is not something you fix with the volume knob. It’s set by physics and by the front-end hardware.
Noise figure — the receiver adds its own noise
No amplifier is silent. Every stage in the receive chain adds a little noise of its own, degrading the SNR that arrives at its input. Noise figure (NF) quantifies that penalty in dB — how much worse the SNR is coming out of a stage than going in. An ideal noiseless stage has NF = 0 dB; real ones add anywhere from a fraction of a dB (a good low-noise amplifier) to several dB.
The subtlety is that the first stage matters most. Once the first amplifier has set the noise contribution, later stages add relatively little because the signal is already amplified when it reaches them. That’s why a clean, low-noise-figure first amplifier — ideally right at the antenna — does more for weak-signal reception than anything downstream. The front-end lesson covers where that amplifier belongs and how to avoid overloading it.
Why more gain isn’t more SNR
Here’s the point that trips people up. Gain multiplies the signal and the noise by the same factor, so it doesn’t change the ratio between them. If a signal enters an amplifier 12 dB above the noise, it leaves 12 dB above the noise — just louder.
SNR is essentially fixed by the time the signal reaches the amplifiers you control: it’s set by the antenna, the environment, and the first stage’s noise figure. Turning gain up past the point where the signal is already clear of the receiver’s internal noise buys nothing but risk — it pushes everything toward the ADC ceiling and invites clipping and overload. The right amount of gain lifts weak signals clear of the ADC’s own noise and then stops. Beyond that, SNR is what it is.
Reading SNR in practice
GopherTrunk shows SNR live on its tuning meters alongside the level in dBFS. When you’re evaluating whether a system is decodable, watch the SNR, not just the level:
- SNR climbing as you raise gain, then flattening — normal. Once it plateaus, you’ve lifted the signal clear of internal noise; stop there (see the gain lesson).
- High level but low SNR — the noise floor is up with the signal. Suspect local interference or front-end overload, not a weak transmitter.
- SNR comfortably in the double digits with a tight constellation — you’re in good shape to decode.
- SNR hovering at the threshold — expect intermittent locks and dropped voice; you need more signal or less noise (a better antenna usually beats more gain).
Quick check: a signal reads a strong −40 dBm but won't decode. Most likely cause?
Recap
- Every receiver has a noise floor from thermal noise (kTB, ~−174 dBm/Hz), man-made noise, and its own electronics.
- Wider bandwidth collects more noise power — about 3 dB per doubling.
- SNR — the dB gap between signal and floor — is what decides a decode, not absolute dBm.
- Digital voice typically needs roughly 10–20 dB SNR (low EVM) to lock.
- Sensitivity/MDS is the weakest signal a receiver can hear, set by bandwidth and noise figure; the first amplifier matters most.
- More gain doesn’t add SNR — it lifts signal and noise together. SNR is set before the ADC.
Next up: antennas 101 — the first thing that decides how much signal, and how much noise, reaches your radio.
Frequently asked questions
What is a good SNR for decoding?
There’s no single number, but digital voice systems typically need roughly 10–20 dB of signal-to-noise ratio to lock reliably, with the exact threshold depending on the protocol and how demanding its error correction is. What actually matters at the demodulator is a clean constellation and low symbol error (low EVM); SNR is the practical proxy you can read off a meter. Below the threshold you’ll see intermittent locks and dropped voice; well above it, more SNR buys you little.
What is the noise floor?
The noise floor is the background level of random energy present at the receiver even with no signal of interest — it sets the bottom of what you can hear. It comes from unavoidable thermal noise, man-made noise picked up by the antenna, and the receiver’s own electronics. A signal has to rise above this floor by enough margin (its SNR) to decode.
Does more gain improve SNR?
Only up to a point. Gain amplifies the signal and the noise together, so once a weak signal is already clear of the receiver’s internal noise, adding gain doesn’t change the ratio between them — it just moves both up toward the ADC ceiling. Past that point extra gain only risks overload. See the gain lesson for the full trade-off.
What is receiver sensitivity, or MDS?
Sensitivity describes the weakest signal a receiver can usefully recover, often quoted as the minimum discernible signal (MDS) — the input power, in dBm, that produces a signal just detectable above the noise floor. It’s set mainly by the receiver’s bandwidth and its noise figure. A lower (more negative) MDS means a more sensitive receiver.