Receiver Sensitivity: MDS, Noise Floor and Noise Figure

Every transceiver data sheet lists a sensitivity figure, and almost every operator has wondered what it really means. Is a lower number always better? Does a preamplifier help? Why does a radio with a “worse” figure sometimes hear better on the air? The answers lie in a small amount of arithmetic and one idea: sensitivity is not about how much a receiver amplifies, but about how much noise it adds. This guide explains receiver sensitivity in practical terms, covering the thermal noise floor, bandwidth, noise figure, the minimum discernible signal (MDS), how those numbers relate to the S-meter, and when sensitivity matters on HF and on VHF.

What Receiver Sensitivity Really Measures

A receiver can only detect a signal that is stronger than its own noise. Amplification does not help, because noise is amplified along with the signal. Sensitivity is therefore the weakest signal level at which a receiver can still produce a usable result. Manufacturers define “usable” in different ways. For SSB and CW, the figure is often the minimum discernible signal (MDS), which is the signal power equal to the receiver’s noise power, so that the output rises by 3 dB when the signal is applied. For FM, the specification is usually written in microvolts for a given signal-to-noise-and-distortion ratio, such as 12 dB SINAD. These are different tests with different units, so comparing numbers from different radios requires care.

The related minimum detectable signal concept is used in many fields. For ham radio, the practical definition is simple: MDS is the signal level that gives the same audio output as the receiver noise alone. Anything below the MDS is hidden in the noise, and anything well above it is easy to hear. Levels are normally given in dBm, which is decibels relative to one milliwatt, so that very small numbers become convenient negative values.

The Thermal Noise Floor: -174 dBm per Hertz

Every resistor, every antenna and every piece of matter at room temperature produces thermal noise, described in the Johnson-Nyquist noise article. At a standard temperature of 290 kelvin, the available noise power is about -174 dBm in every hertz of bandwidth. To find the noise in the bandwidth of your receiver, add 10 times the logarithm of the bandwidth in hertz. A bandwidth of 1000 Hz adds 30 dB, giving -144 dBm; a bandwidth of 100 Hz adds 20 dB, giving -154 dBm. This is the best possible noise floor for that bandwidth: a perfect, noiseless receiver cannot do better.

Two consequences follow immediately. First, narrowing the filter lowers the noise floor and improves the MDS: halving the bandwidth improves it by 3 dB, and reducing from 2400 Hz to 240 Hz improves it by 10 dB. This is why a CW operator with a narrow filter hears signals that an SSB operator cannot. Second, the noise floor is set by the bandwidth you choose, not by the radio’s maximum capability. A CW signal in a 500 Hz filter is easier to hear than the same signal in a 2400 Hz filter even if the radio and antenna are identical. Our guide to receive filter bandwidth explains how to choose the width.

Noise Figure: How Much Noise the Receiver Adds

A real receiver adds noise of its own. The noise figure is the number of decibels by which the receiver degrades the signal-to-noise ratio compared with the ideal case. A noise figure of 10 dB means the receiver’s own noise is 10 dB above the thermal floor. Adding it to the thermal noise gives the practical formula: noise floor in dBm equals -174, plus 10 times the logarithm of the bandwidth, plus the noise figure. The noise figure article gives the formal definition.

The values come directly from the formula. They show that bandwidth matters as much as noise figure: going from 2400 Hz to 250 Hz improves the noise floor by about 10 dB, the same as improving the noise figure from 15 dB to 5 dB. Many modern HF transceivers have noise figures in the range of roughly 5 to 15 dB depending on band and preamp settings, but check the data for your own radio, because this varies and manufacturers do not always publish it. Independent lab reports often list the MDS at a stated bandwidth, and those numbers are comparable between radios if the bandwidth is the same.

Relating dBm to the S-Meter and to Microvolts

The usual convention for HF is that S9 corresponds to a signal of 50 microvolts across 50 ohms, which is -73 dBm. Each S-unit below S9 is 6 dB, so S8 is -79 dBm, S7 is -85 dBm, and S1 is -121 dBm. Above S9 the meter reads in decibels over S9. A noise floor of -130 dBm in a 2400 Hz bandwidth is therefore about 9 dB below S1, which is why a good receiver with the antenna disconnected shows no S-meter movement at all, and why a signal that barely lifts the meter can still be perfectly readable on CW. Many real radios do not follow the convention exactly, particularly at VHF, where the convention is different, so treat the S-meter as relative.

The same logic gives a link between dBm and microvolts: 1 microvolt across 50 ohms is -107 dBm, 0.5 microvolt is -113 dBm and 0.2 microvolt is -121 dBm. So the common FM specification of about 0.2 microvolt for 12 dB SINAD corresponds to a very weak signal, similar in level to the S1 reference point, although the two numbers measure different things.

When Sensitivity Matters and When It Does Not

This is the most important practical point. On the lower HF bands, external noise from the atmosphere, the Sun and human-made sources is far above the thermal floor. On 80, 40 and 20 meters in a typical residential area, noise from power lines, switching supplies and electronics often sets the limit, by tens of decibels. If the noise that arrives at your antenna is 30 dB above the thermal floor, a receiver with a noise figure of 10 dB adds hardly anything to it, and making the receiver quieter has no effect on the signals you can hear. Adding a preamplifier on those bands only amplifies noise and adds the risk of overload, a point we cover in our guide to the HF preamp and attenuator.

The opposite is true at VHF and above. External noise is lower, so the receiver and the feedline can dominate. A long run of coaxial cable ahead of the receiver adds its loss directly to the system noise figure: 3 dB of loss in the cable means 3 dB more noise. That is the reason for mast-mounted preamplifiers on weak-signal VHF and UHF stations, covered in our mast-mounted preamplifier guide. For EME and other weak-signal work, every decibel matters, and the noise figure of the first amplifier is one of the main design choices.

• Low HF bands (160 to 40 meters): external noise dominates; bandwidth and antenna quality matter more than noise figure.
• Upper HF bands (20 to 10 meters): external noise is lower in quiet locations, and a quiet receiver can help, but only if your local noise is also low.
• VHF and UHF: receiver noise figure and feedline loss are important; a good preamplifier at the antenna is a common upgrade.
• Weak-signal modes: the same arithmetic applies, with long integration acting like a very narrow bandwidth.

If you suspect that local noise is the real limit, the right step is to find the source. Our guide to finding and killing RF noise explains the process, and removing a source of 20 dB of noise has a larger effect than any change to the radio.

Worked Example: Comparing Two Radios

Suppose a data sheet quotes an MDS of -138 dBm for radio A in a 500 Hz bandwidth and -132 dBm for radio B, in a 250 Hz bandwidth. At first glance radio A looks better by 6 dB, but the bandwidths differ. Narrowing radio A from 500 Hz to 250 Hz would lower its floor by 3 dB, giving about -141 dBm, so radio A is really 9 dB better. Now suppose both specifications use 500 Hz, and radio A quotes -138 dBm while radio B quotes -137 dBm. The 1 dB difference is far smaller than the variation you will see from the antenna, the band and the time of day, and you should ignore it when making a purchase.

The same reasoning applies to a preamplifier. If it improves the system noise figure by 3 dB, the noise floor falls by 3 dB, which matches the effect of halving the filter bandwidth. Doing both gives 6 dB. Whether this is worth the cost depends on whether noise from the radio or noise from outside is the limit on your band. If it is outside noise, neither change matters; if it is the radio, both will help. Always decide using the dummy load test above, not the data sheet alone.

The same arithmetic explains why digital modes such as FT8 work with such weak signals. A decoder that integrates over several seconds behaves like an extremely narrow filter, so the effective noise floor falls far below what a normal SSB filter allows. The standard reference bandwidth for WSJT-X signal reports is 2500 Hz, and signals decode at levels well below -20 dB in that reference bandwidth. The thermal floor is the same, but the processing gain lets the software reach much deeper into it.

A Simple Test You Can Do at Home

You can check your own receiver without special equipment. Set the radio to a quiet band, set the mode and filter width you normally use and turn off the preamplifier and the noise reduction. Connect a dummy load in place of the antenna and note the audio level or the S-meter reading, then reconnect the antenna. If the noise rises noticeably with the antenna connected, external noise is above the receiver’s own noise and sensitivity is not the limit on that band. If the noise stays the same, your station is limited by the radio, and a preamplifier or a quieter receiver might help. This simple test is the rule of thumb behind much of the advice in the amateur literature, and it is quick enough to try on each band.

For a more careful measurement you need a calibrated signal generator and a way to read the audio output level. With the generator connected to the antenna input at a known level, you adjust the level until the output rises by 3 dB compared with the noise alone, and read off the MDS. This is how laboratories do it, and it requires a generator accurate at very low levels. Most hams will not need to do this, but it helps to know how published numbers are produced.

Common Misunderstandings

• “More gain is better sensitivity.” Gain raises noise and signal alike. Only a lower noise figure or a narrower filter improves the signal-to-noise ratio.
• “A lower MDS number is always better, whatever the bandwidth.” MDS must be compared at the same bandwidth; a figure for 500 Hz cannot be compared with one for 2400 Hz.
• “Sensitivity decides how far I hear.” On HF, propagation and external noise decide most of it.
• “Dynamic range is the same thing.” Dynamic range describes the span between the noise floor and the strongest signal the receiver can handle; see our article on receiver dynamic range.
• “A preamp always helps.” It helps only when the receiver noise, not the external noise, is the limit.

The Bottom Line

Receiver sensitivity is set by three things: the thermal floor of -174 dBm per hertz, the bandwidth you select, and the noise figure of the receiver and feedline. Narrowing the filter and cutting feedline loss are the cheapest ways to improve it. On the lower HF bands, external noise swamps the receiver’s own noise, so a quieter radio or a preamplifier will not help; on VHF and above, the noise figure is a real design choice. Use the formula, check bandwidths when comparing radios, and look for noise sources at home before spending money on gear.

Frequently Asked Questions

What does MDS mean?

Minimum discernible signal: the signal level at which the output equals the receiver’s own noise, so the audio rises by 3 dB when the signal is applied. It is usually quoted in dBm at a stated bandwidth.

What is the thermal noise floor?

About -174 dBm per hertz of bandwidth at 290 kelvin. Add 10 times the logarithm of the bandwidth in hertz to get the noise in your filter.

Does a narrower filter improve sensitivity?

Yes. Halving the bandwidth lowers the noise floor by 3 dB, so a 250 Hz CW filter has a noise floor about 10 dB lower than a 2400 Hz SSB filter.

What is the noise figure?

The number of decibels by which the receiver degrades the signal-to-noise ratio compared with a perfect receiver. A lower number means a quieter receiver.

Do I need a preamplifier on HF?

Usually not on the lower bands, where external noise is much higher than receiver noise. It may help on the highest HF bands in a quiet location, and more often at VHF and UHF.

How many dBm is S9?

By the usual HF convention, S9 is 50 microvolts across 50 ohms, or -73 dBm, with 6 dB per S-unit below it.