How to lie about loud
Loudness is the new battle line for portable speakers, everybody is claiming to be the loudest, but how do you find out the truth?
The proper name for a speaker is “Loudspeaker”, and speakers should really be able to play music loud, but they vary hugely in their abilities. The portable speaker marketeers have gone from describing things as” loud”, “super loud”, “hyper loud” etc to using a loudness measurement in dB as a figure of merit. The more dB you have the louder the speaker is.
Sounds good, but as ever, the devil is in the detail, and you can easily produce dB figures that look good, but in reality are just hype.
When measuring loudness there are 5 major variables:
- Distance: How close the measurement microphone is to the speaker
- Signal: whether you use pink noise, sinewave, squarewaves or music
- Weighting: how the measuring meter responds to the signal
- Distortion: is the speaker clipping when it’s loud
- Space: how big or reverberant is the room you’re measuring in
Unless you specify all of the above conditions, you can’t compare speaker A with speaker B.
How to make an average speaker look great
So let’s measure a hypothetical speaker using a engineering standard:
- Distance: 1M on axis of tweeter
- Signal: pink noise with 6dB crest factor
- Weighing: dBA, – A response, slow response meter
- Distortion: no clipping
- Space: not specified, but assumed non reflective or open air
We do this, and get a reading of 75dBA, which is pretty average. So lets manipulate it:
- move the mic closer, so it’s 0.5M away – add 6dB (81dB)
- use a sinewave instead of noise – add 3dB (84dB)
- use CCIR peak weighting instead of Z weighting – add 6dB (90dB)
- clip the hell out of the signal – add 10dB (100dB)
- measure in a small reflective room – add 6dB ( 106dB)
So we’ve taken a poor figure – 75dBA, and converted it into a stunning figure – 106dB, good hey? To get this figure legitimately, we’d have to increase the speaker amplifier power enormously, to about 1000W at which point the speaker would melt! Both loudness figures are true, but if you drove music through the speaker to get 106dB, it would sound awful. Lets look at each of these manipulations to see what they do.
Sound is radiated in a pulsing sphere from the surface of the speaker driver, with the audio energy distributed across the surface area of that sphere. The surface area of a big sphere, is obviously bigger than a smaller one. If you do the math, it turns out that if you half the distance, the measured level goes up 6dB. In a normal room, this can be between 3-6dB because of reflections, but it’s 6dB in the open air, or anechoic chamber.
Moral: you must know what the measurement distance was.
Different types of signal have different amounts of energy in them, and loudness is a measure of sound energy. If we can put more energy into a speaker, we will get a louder measurement.
Now music has quiet bits and loud bits, so pink noise, which also has quiet bits and loud bits is often used to simulate the loudness of music. Because pink noise can be accurately specified, but music not (unless everybody agrees to check the loudness of a speaker with a defined music track), it’s often used as a signal.
Pink noise has a crest factor of 6dB. This means that its peak level is 6dB louder than it’s average level. Loudness however, measures average signal level, so a sinewave (which has a crest factor of 3dB) will automatically give you a result 3dB louder. There’s also a problem that because you’re using a fixed frequency sinewave, it’s actual level will be modified by the room you measure it in. Unless it’s done in a anechoic chamber, reflections and resonances can easily add another 6dB to the measurement. The proper way to test it is to take multiple measurements of different sinewave frequencies and average these, or use a fast swept sinewave.
Moral: you must specify the signal used to measure the loudness
The measurement meter needs to work out what the average level is, and it applies a weighting filter to do this. Different filters give different results, and loudness is usually measured with an RMS long time average A weighting filter – that’s the “A” in “dBA”
The A filter is actually misleading for speaker loudness: If I designed a speaker with no bass or treble, I could still get a good dBA figure. For speakers, I recommend a C weighing filter (dBC), which checks the levels of bass and treble loudness as well.
A CCIR filter is used to measure background noise, and it’s boosted at midrange frequencies, and also measures peak, not average levels. Consequently if you use this filter, the measurement you’ll get will be ~6dB louder.
Moral: you must specify the weighting filter used.
Now this is key. If you start to clip the amplifier, the crest factor gets smaller, and the average level increases – rapidly. It sounds really hard and nasty with music, but it’s certainly louder. You can easily get a 10dB increase in measured level.
This is sort of related to how modern music has got louder; it’s been made with compressors, which reduce the crest factor, and increase the average level of the music.
Most portable speakers distort like this on full volume, and consequently give good measured loudness figures, but sound dirty. It’s because the batteries can only produce so much voltage and current.
Moral: you should know if the speaker is clipping or not.
To be consistent, all speakers should be measured anechoically – this means without any reflected sound. You can sort of do this the open air, or with special measurement techniques. The reason for this is that sound bounces off walls and a room becomes an acoustic amplifier, boosting the loudness. A smaller room with hard walls and floor will sound louder than a large room with lots of curtains and soft furnishings.
Another factor is how widely the speaker radiates it’s sound. Some speakers are like torches, the sound is really focused in a beam. Once you get out of the beam, the sound level drops off. In the beam however the volume level is more intense. These sort of speakers will always measure really well anechoically, but poorly in a real life unless you’re right in front of them: it’s a minefield!
Moral: you have to know the environment the speaker was measured in.
Do it right
We measure our figures at 1M and 0.5M, with pink noise with C weighing filters, before clipping, in a medium sized normal room. Because the SoundBucket radiates in 360 degrees, the actual perceived loudness is actually higher than the already good figures suggest.
Ideally we want you to listen to the dynamics of music, not clipping distortion just to get better loudness figures. But we appreciate that some folk just want it loud! So:
- On batteries, we allow you to clip the amplifier slightly so you can choose whether you want to hear the sound clean or dirty. If it sounds dirty, just turn down the level slightly!
- If you use our optional external DC power supply, it provides a really big voltage to the SoundBucket internal amplifiers so they never clip. The result is loud but clean music.
It’s the best of both worlds!