Signal to noise ratio (SNR) is calculated by dividing the strength of your desired signal by the strength of the background noise, then typically converting that ratio to decibels. The core formula is straightforward: SNR = Signal ÷ Noise. But the specific way you measure “signal” and “noise” depends on whether you’re working with audio, wireless networks, digital images, or lab instruments. Here’s how the math works across the most common scenarios.
The Basic SNR Formula
At its simplest, SNR is a ratio. If your signal has a power of 100 watts and the noise has a power of 1 watt, the SNR is 100. That’s it. The challenge is that real-world signals span enormous ranges, so expressing SNR as a plain number gets unwieldy fast. That’s why most fields convert it to decibels.
To convert a power ratio to decibels:
SNR (dB) = 10 × log₁₀(Signal Power ÷ Noise Power)
If you’re working with amplitude (voltage, sound pressure, or any measurement that isn’t squared into power), the formula changes slightly because power scales with the square of amplitude:
SNR (dB) = 20 × log₁₀(Signal Amplitude ÷ Noise Amplitude)
Both formulas give the same result when applied to the same physical situation. The 10× version is for power quantities; the 20× version is for amplitude quantities. Mixing them up is one of the most common mistakes people make. A quick way to remember: if you measured it with something like a voltmeter or microphone (amplitude), use 20. If it’s already in power units (watts, milliwatts), use 10.
As a reference point, an SNR of 40 dB means the signal amplitude is 100 times larger than the noise amplitude. An SNR of 20 dB means a factor of 10. And 0 dB means the signal and noise are equal, which is generally useless.
Measuring Signal and Noise With RMS
For signals that fluctuate over time (audio, vibrations, electrical signals), you need a consistent way to measure their strength. The standard approach is the root-mean-square (RMS) value. RMS works by squaring the signal at every point in time, averaging those squared values, and taking the square root. This gives you a single number representing the signal’s effective amplitude.
To calculate SNR using RMS values, you measure the RMS amplitude of your signal during the period of interest, then measure the RMS amplitude of the noise alone (with no signal present, or in a region where you know there’s only noise). Divide the signal RMS by the noise RMS, and apply the 20 × log₁₀ formula to get decibels.
For example, if your signal has an RMS amplitude of 0.5 volts and the noise measures 0.005 volts RMS, the SNR is 20 × log₁₀(0.5 ÷ 0.005) = 20 × log₁₀(100) = 20 × 2 = 40 dB.
SNR for Wireless and Wi-Fi Signals
In wireless networking, signal strength and noise are already reported in decibels (specifically dBm, which references 1 milliwatt). Since both values are already in a logarithmic scale, you don’t need the log formula at all. You just subtract.
SNR (dB) = Signal (dBm) − Noise Floor (dBm)
If your Wi-Fi client receives a signal at −75 dBm and the noise floor sits at −90 dBm, the SNR is 15 dB. Most wireless tools and router dashboards report signal strength and noise floor separately, so this subtraction is all you need. For reliable Wi-Fi performance, you generally want an SNR above 20 dB; below 10 dB, connections become unreliable.
SNR in Digital Imaging
For cameras and image sensors, SNR is calculated per pixel. The signal is the number of electrons generated by incoming light (which is proportional to the number of photons hitting the sensor, multiplied by the sensor’s quantum efficiency). The noise comes from three sources combined: the natural randomness of photon arrival (called shot noise), the sensor’s readout electronics, and dark current from thermal activity in the chip.
The formula for a digital camera sensor is:
SNR = Signal Electrons ÷ √(Signal Electrons + Readout Noise² + Dark Current Noise²)
Notice the signal itself contributes to the noise (through the square root term). This means doubling the light doesn’t double the SNR. It only improves it by a factor of about 1.4, because the shot noise grows with the signal. In practice, if you’re trying to improve SNR in imaging, longer exposure times and cooling the sensor (to reduce dark current) help, but there are diminishing returns.
For a simpler approach that works with any image file: capture a uniformly lit frame, measure the mean pixel value in a region (that’s your signal), and measure the standard deviation of pixel values in that same region (that’s your noise). Divide mean by standard deviation, and you have a linear SNR you can convert to decibels with the 20× log₁₀ formula.
SNR in Chromatography and Lab Work
In analytical chemistry, SNR determines whether you can reliably detect a substance. The standard method uses peak height rather than area. You measure the signal from the middle of the baseline to the top of the chromatographic peak. For noise, you measure the total vertical distance between two lines that bracket the baseline fluctuations in a region near the peak where no compounds are eluting.
SNR = Peak Height ÷ Baseline Noise
Peak height matters more than peak area for this calculation because you’re comparing a height measurement (the peak) against another height measurement (the noise band). Regulatory guidelines typically require an SNR of at least 10 for quantitation and at least 3 for detection. If your SNR falls below those thresholds, the measurement isn’t considered reliable enough to report.
What Counts as a “Good” SNR
What qualifies as adequate depends entirely on the application. In medical imaging and visual detection tasks, the Rose criterion sets the bar: a signal needs to be about 5 times the standard deviation of the background noise for a human observer to reliably spot it. This holds even under ideal conditions, and in practice, you usually need more than 5 because real backgrounds aren’t perfectly uniform or predictable.
Some rough benchmarks across fields:
- Wi-Fi: 20 dB or higher for smooth streaming; below 10 dB causes dropouts
- Audio recording: 60 dB or higher is considered good; below 40 dB, noise becomes clearly audible
- Analytical chemistry: 10:1 minimum for quantitation, 3:1 minimum for detection
- Medical imaging: SNR of 5 or higher for reliable feature detection (Rose criterion)
Common Mistakes to Avoid
The most frequent error is confusing the 10× and 20× log formulas. If you apply 20 × log₁₀ to a power ratio, you’ll get a number twice as large as the correct answer. Always check whether your measurements are power or amplitude before choosing the formula.
Another common issue is measuring noise incorrectly. Noise should be measured in conditions as close as possible to your actual measurement, minus the signal. In audio, this means recording with the same gain settings but no input. In imaging, it means capturing a frame with the lens cap on (for dark noise) or analyzing a blank region of the image. If you measure noise under different conditions than the signal, your SNR won’t reflect reality.
Finally, keep in mind that averaging multiple measurements improves SNR by the square root of the number of measurements. Averaging 4 readings doubles your SNR. Averaging 100 readings improves it by a factor of 10. This principle applies across nearly every field, from astronomy (stacking exposures) to analytical chemistry (multiple injections).