Electronics
How many watts do you really need?
In short: Watts are an expensive way to buy loudness. Doubling the power adds just 3 dB, which the ear hears as a small step. To make music sound twice as loud, you need ten times the watts. Speaker sensitivity matters far more: a 10 dB difference in sensitivity is worth ten times the amplifier power. At normal listening levels at home, your speakers get on average anywhere from a fraction of a watt to a few watts. You need power mainly for the peaks, which in music sit 10 to 20 dB above the average. Speakers are rarely killed by “too many watts”. They are killed by an amplifier that runs out of headroom and starts clipping the signal. When you compare numbers, trust only continuous power (RMS) at a stated distortion and load, never PMPO.
In this article:
- Watts and loudness: why 3 dB and 10 dB
- Speaker sensitivity and the 4 Ω trap
- Distance and the room
- Two speakers instead of one
- Average and peaks: headroom in music
- Calculation: 85 dB at three metres
- Table: sensitivity and the power you need
- What really kills tweeters
- Marketing watts: PMPO, peak and RMS power
- How much a speaker can take
- What to choose: practical advice
- Key takeaways
- Short glossary
Watts and loudness: why 3 dB and 10 dB
Your ear does not hear loudness on a linear scale. If it worked like a kitchen scale, twice the watts would sound twice as loud. In reality the ear works more like a ruler with a logarithmic scale: what counts is the ratio, not the difference. That is why loudness is given in decibels (dB).
Level change in dB = 10 · log10(P2 / P1)
2× power = +3 dB, 4× power = +6 dB, 10× power = +10 dB, 100× power = +20 dB.
What this means for your ears:
- 1 dB is roughly the smallest difference you can hear in a direct comparison, and only with a steady sound.
- 3 dB is clearly audible, but a modest step. Like nudging the volume knob a little.
- 10 dB sounds roughly twice as loud. This rule comes from psychoacoustic measurements (Stevens, the sone scale) and holds up well in practice for midrange frequencies and normal listening levels.
That leads to an uncomfortable fact: a 100 W amplifier instead of a 50 W one gives you only +3 dB. To make music sound twice as loud as with 50 W, you would need 500 W. And that is with the same speakers, which might not even survive such a level.
Speaker sensitivity and the 4 Ω trap
Sensitivity tells you how loud a speaker plays for a given input, measured 1 metre in front of it. Typical values:
- small bookshelf speakers: 82 to 86 dB,
- typical floorstanding speakers: 86 to 90 dB,
- high-efficiency designs, full-range drivers and horn speakers: 92 to 100 dB and more.
A 10 dB difference in sensitivity is the same as ten times the power. A 94 dB speaker with 10 W plays as loud as an 84 dB speaker with 100 W. That is why sensitivity matters more than the watts themselves when you choose an amplifier.
Two ways to state sensitivity
Manufacturers state it in two ways, and they look almost identical:
- dB / 1 W / 1 m: the level at a power of 1 W.
- dB / 2.83 V / 1 m: the level at a voltage of 2.83 V. Into 8 Ω this is exactly 1 W (2.83² / 8 = 1 W).
For an 8 Ω speaker both numbers are the same. Into 4 Ω, however, 2.83 V means 2.83² / 4 = 2 W. So the speaker got twice the power, and the number comes out 3 dB higher. A 4 Ω speaker rated at “90 dB / 2.83 V” really has only 87 dB at 1 W.
The 4 Ω trap: if a speaker is 4 Ω and its sensitivity is stated at 2.83 V, subtract 3 dB when you work out the power you need. It is not cheating, just two different ways of looking at it. Voltage is the more natural reference for today’s amplifiers, because they behave like voltage sources. But the power the amplifier has to deliver into 4 Ω is still doubled.
A good amplifier delivers almost twice as much power into 4 Ω as into 8 Ω, provided it has a strong enough power supply and output transistors. Weaker amplifiers are more likely to hit their current limit or overheat at 4 Ω. Check whether the manufacturer also states the power into 4 Ω.
If you design speakers yourself, you will find the driver’s sensitivity on its datasheet (see Where to get driver data). Watch out: the finished speaker is often less sensitive than its drivers. The crossover has to match the woofer and the tweeter, so the whole system ends up at the level of the less sensitive one. Loudsmith shows you the resulting response, and you can read the real sensitivity straight from it.
Distance and the room
Sensitivity is quoted at 1 metre. But you sit further away. In free space, say outdoors in a field, a simple rule applies: every doubling of distance costs 6 dB.
Free-field loss = 20 · log10(r / 1 m)
2 m: −6 dB, 3 m: −9.5 dB, 4 m: −12 dB, 8 m: −18 dB.
In a room the drop is not as steep. Sound reflects off the walls, ceiling and floor, and those reflections add to the direct sound. In a typical living room the level therefore falls more slowly with distance, often only by 3 to 4 dB per doubling. In the bass the room also adds level (so-called room gain), but unevenly, with pronounced peaks and dips. More on this in The room plays along and in the room planner.
In the calculations below I assume a free field. It is a cautious estimate: in a real room you will get a few dB more, which is extra headroom.
Two speakers instead of one
In stereo, two speakers play at once. Their sound adds up, but how depends on the frequency:
- At midrange and high frequencies, the sound from two speakers combines more or less randomly at the listening position, sometimes in phase, sometimes against each other. On average this gives +3 dB, the same as doubling the power.
- At low frequencies, where the wavelength is much longer than the difference in distances, the signal from both speakers adds in phase. Twice the sound pressure means +6 dB. This applies to content that is the same in both channels, such as bass, vocals and a solo in the centre.
For the calculation I take the more cautious +3 dB. So each speaker has to play 3 dB quieter on its own than the level you want to hear from both together.
Average and peaks: headroom in music
Music is not a steady tone. A drum hit, a piano strike or a trumpet jumps above the average level for a fraction of a second. The ratio between the peak and the average (more precisely, the effective value, RMS) is called the crest factor.
| Signal | Crest factor, approx. |
|---|---|
| Pure sine tone | 3 dB |
| Heavily compressed modern pop, rock | 8 to 12 dB |
| Jazz, acoustic music, older recordings | 12 to 18 dB |
| Classical, film sound, live drums | 18 to 25 dB |
The amplifier has to handle the peak, not the average. If you want an average of 85 dB with peaks 20 dB higher, the amplifier must briefly deliver enough power for 105 dB. A 20 dB difference is a hundred times the power: when 1 W flows on average, the peak needs 100 W.
These numbers are not made up. Cinemas are calibrated to a reference level of 85 dB per channel, and the sound system has 20 dB of headroom above it, so peaks reach 105 dB. Studios work with similar headroom (for example Bob Katz’s K-20 system). For home listening, 85 dB is already quite loud. Most people listen at an average of around 70 to 80 dB.
Calculation: 85 dB at three metres
Now let’s put it all together. The brief:
- average level at the listening position 85 dB,
- headroom for peaks 20 dB, so peaks of 105 dB,
- distance 3 m, free field (−9.5 dB),
- two speakers (+3 dB),
- three speakers with a sensitivity of 84, 88 and 95 dB / 1 W / 1 m.
Step by step
- Each speaker has to produce 105 − 3 = 102 dB at 3 m.
- At 1 m that means 102 + 9.5 = 111.5 dB.
- How many dB you need above the sensitivity: 111.5 − sensitivity.
- Power = 10(difference / 10) W.
| Sensitivity | Extra needed | Peak power | Average power (20 dB less) |
|---|---|---|---|
| 84 dB | 27.5 dB | about 570 W | about 5.7 W |
| 88 dB | 23.5 dB | about 225 W | about 2.3 W |
| 95 dB | 16.5 dB | about 45 W | about 0.45 W |
Notice two things. First, the average power is ridiculously small: even when you listen loud, the speaker gets just a few watts. Second, the peak power for low-sensitivity speakers is huge. 570 W per channel for small 84 dB bookshelf speakers is more than most such speakers can take, and that is not the amplifier’s fault. Speakers like these are simply not built for 105 dB at three metres.
If you settle for 100 dB peaks (an 80 dB average, still loud), the power drops to a third: 180, 71 and 14 W. And in a real room, where the level falls more slowly with distance, roughly half of that is often enough.
Table: sensitivity and the power you need
For a quick estimate: peak power per channel at a distance of 3 m, with two speakers in a free field, for three loudness levels. The average is always 20 dB below the peak. Sensitivity is at 1 W (for 4 Ω with a 2.83 V rating, subtract 3 dB first).
| Sensitivity | Peaks 95 dB (average 75 dB) | Peaks 100 dB (average 80 dB) | Peaks 105 dB (average 85 dB) |
|---|---|---|---|
| 82 dB | 90 W | 285 W | 900 W |
| 85 dB | 45 W | 143 W | 450 W |
| 88 dB | 23 W | 71 W | 225 W |
| 91 dB | 11 W | 36 W | 113 W |
| 94 dB | 6 W | 18 W | 57 W |
| 97 dB | 3 W | 9 W | 28 W |
| 100 dB | 1.4 W | 5 W | 14 W |
The table shows why fans of 8 W per channel valve amplifiers reach for sensitive speakers of 95 dB and up. With speakers like that, 8 W really is enough. More in Valves or transistors.
What really kills tweeters
A common piece of advice goes: “Don’t put a more powerful amplifier on your speakers, you’ll blow them.” In reality, speakers are more often killed by a weak amplifier pushed beyond its limits.
How clipping happens
Every amplifier has a ceiling: its output voltage cannot exceed its supply voltage. When you ask for more gain than that, the tops of the waveform get clipped flat. The sine wave starts to look like a square wave. This has two consequences:
- Higher harmonics appear. The sharp edges of a square wave carry a lot of high-frequency energy. That energy goes straight through the crossover into the tweeter, which is built for only a fraction of the woofer’s power. A typical tweeter can handle only a few watts continuously.
- The average power rises. A clipped signal has a lower crest factor, so the average moves closer to the peak. A fully clipped square wave carries twice the power of a sine with the same peak. A “50 W” amplifier can then push around 100 W continuously into the speaker.
The tweeter’s voice coil overheats and burns out. From the outside it does not look dramatic: one day the treble is simply gone.
But a big amplifier can do damage too
To be fair, there is another side. A powerful amplifier that does not clip can also destroy a speaker if you turn it all the way up. Power beyond a driver’s thermal or mechanical limit does harm even when it is clean. The difference is that you can usually hear clean overload: the bass starts to bang and the sound turns hard. Clipping is harder to hear with dynamic music, until it is too late.
Rule of thumb: an amplifier with headroom that never clips is safer for your speakers than a weak amplifier at full tilt. If you hear the sound turning hard and distorted, turn it down. It is not “too few watts”, it is a sign that something is nearing its limit.
How an amplifier clips also depends on its design. Valve amplifiers clip more softly, class D often very sharply, and some have a limiter. More in Amplifier classes.
Marketing watts: PMPO, peak and RMS power
On the box of a portable speaker you will find “1,000 W”, on a quality amplifier “2 × 50 W”. The first number is often less honest than the second.
| Label | What it means | Can you trust it? |
|---|---|---|
| RMS, continuous power | the power the amplifier delivers continuously into the stated load at the stated distortion and bandwidth | yes, if the conditions are given |
| Peak | a short-term maximum, often just double the continuous figure | as a rough guide |
| Music power, “max” | short-term power on a tone, before the power supply sags | not much |
| PMPO | no binding standard, the manufacturer picks the number | no |
A good spec looks like this, for example: 2 × 50 W into 8 Ω, 20 Hz to 20 kHz, THD 0.1 %, both channels driven. Every condition matters. “Both channels” means the power supply can feed both at once. “20 Hz to 20 kHz” means the power holds in the bass too, where it is usually hardest. “THD 1 %” or “10 %” already means the onset of clipping; at 10 % THD the number comes out tens of percent higher.
By the way, “RMS watt” is a physically inaccurate term. Strictly speaking, it is average power calculated from the effective (RMS) voltage. In practice, though, everyone understands it as continuous power, and that is good enough.
How much a speaker can take
A speaker’s power handling rating also needs to be read with care. A speaker has two different limits:
The thermal limit
Most of the power you send into a speaker turns into heat in the voice coil. Only a fraction of a percent to a few percent becomes sound. The coil heats up, and during long, loud sessions the glue or insulation may give out. The stated continuous power (RMS, often to IEC 60268-5 or AES2) is measured with noise that has a music-like spectrum, over several hours. So it describes the long-term average, not the short peaks, which the speaker can take at higher levels.
Heat also brings thermal compression: a hot coil has higher resistance, draws less current, and the speaker plays quieter than it should. When you play loud, this can be a few dB.
The mechanical limit: excursion
In the bass, the cone moves a lot. Each octave lower at the same loudness needs four times the excursion. Once the cone goes beyond its linear stroke (Xmax), distortion rises sharply. With even more excursion, the coil hits the bottom of the magnet or the suspension gets damaged. This limit can arrive long before the thermal one, for example with a small woofer and deep bass.
The type of enclosure makes a big difference here. In a bass reflex box, below its tuning the port stops controlling the cone, and the cone starts flapping. That is why a filter that rolls off the deepest bass is a good idea below the tuning (see Bass reflex and Sealed box).
When you design an enclosure in Loudsmith, you can see where your driver and box are still safe: where, according to the driver data, it hits its excursion limit and the highest level you can expect from it in the bass. That is a better guide than the wattage on the label, because it takes frequency into account.
How to match an amplifier to a speaker
- A common recommendation, especially in pro audio: an amplifier with a continuous power roughly 1 to 2× higher than the speaker’s continuous power handling. The headroom ensures the amplifier does not clip, and it is up to you not to turn it all the way up.
- A weaker amplifier is not a mistake, as long as you do not overload it. With sensitive speakers and quiet listening, 20 to 30 W is plenty.
- In your own speaker with a passive crossover, the tweeter is protected above all by the capacitor in the crossover and a sensible crossover point. A low crossover frequency close to the tweeter’s resonance increases its load. More in Speaker box calculations and in the manual.
What to choose: practical advice
| Situation | Sensible continuous power per channel |
|---|---|
| Small room, up to 2.5 m, sensitivity 88 to 92 dB, normal volume | 20 to 50 W |
| Living room, 3 m, sensitivity 85 to 88 dB, occasionally loud | 80 to 200 W |
| Small bookshelf speakers at 82 to 85 dB, you want it loud | 150 W and more, but you will hit the speakers’ limits first |
| Sensitive speakers of 95 dB and more | 5 to 30 W, valves work too |
| Large room, 4 to 5 m, home cinema, live drums | 200 W and more, or more sensitive speakers |
- Want it louder? Raise the sensitivity first (different speakers) or sit closer. 3 dB from sensitivity is free, 3 dB from the amplifier costs twice the watts.
- Comparing amplifiers? The difference between 50 and 80 W is only 2 dB. Decide on quality, inputs and how well it copes with 4 Ω instead.
- Want to compare two amplifiers by ear? Match the levels to within 0.5 dB first, or the louder one will win. A blind ABX test helps.
- Designing your own speaker? The driver’s sensitivity and the enclosure type affect the power you need more than anything else. In an active design with DSP, each driver has its own amplifier and you can set limiters (see Passive or active).
Key takeaways
- Twice the watts = +3 dB. That is only a small step in loudness. Twice as loud takes ten times the power.
- Sensitivity is cheaper than watts. A speaker 6 dB more sensitive saves four fifths of the amplifier power.
- At home, a speaker gets just a few watts on average. You need big power only for short peaks.
- Clipping kills speakers, headroom does not. A more powerful amplifier you never max out is safer than a weak one at full tilt.
- Trust only continuous power with stated conditions. PMPO and “max” figures tell you nothing.
If you build your own speakers, the biggest lever is in your hands: your choice of driver and enclosure decides both the sensitivity and how much bass the speaker can take. In Loudsmith you can try it all for free before you buy a single part: change the driver or the box and instantly see the sensitivity, the excursion and the maximum loudness.
Short glossary
- Decibel (dB): a logarithmic unit of ratio. +3 dB is twice the power, +10 dB ten times.
- Sensitivity: a speaker’s level 1 m in front of it at 1 W or 2.83 V.
- Impedance: a speaker’s “resistance” to alternating current, usually 4 or 8 Ω, varies with frequency.
- Crest factor: the ratio between a signal’s peak and its effective value, in dB.
- Headroom: how many dB above the average level the system can still play without distortion.
- Clipping: the amplifier reaches its voltage ceiling and the tops of the waveform are cut off flat.
- RMS power: continuous power under stated conditions.
- PMPO: a marketing number with no binding standard.
- Xmax: the linear stroke of the cone, within which distortion stays low.
- Thermal compression: a drop in loudness when the voice coil heats up.
Sources: S. S. Stevens, On the Psychophysical Law (Psychological Review 1957); F. Toole, Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms (3rd ed., Routledge 2017); B. Katz, An Integrated Approach to Metering, Monitoring, and Leveling Practices (JAES 2000); SMPTE RP 200 (cinema reference level); IEC 60268-5 and AES2 standards (power handling and loudspeaker specifications); FTC, Amplifier Rule, 16 CFR Part 432 (amplifier power disclosure in the USA). The power calculations are my own.
Free tools: Room planner · Speaker measurement · Blind ABX test