The short version
- Cancellation is a filter with a shape, and one number describes a single point on it.
- Low frequencies are the hard part and the part that matters on transport.
- Passive isolation does most of the work above about 2 kHz, so the seal matters more than the electronics.
- A system that changes the tuning when you switch it on is a system you cannot monitor with.
- Wind noise is a separate problem that no current system solves well.
A pair of headphones advertising “up to 40 dB of noise cancellation” is telling you one true thing about one frequency and letting you assume it applies to all of them. It does not, and the difference between the assumption and the reality is the difference between a quiet train and a slightly muffled one.
Cancellation is a filter with a shape
An active system works by measuring incoming sound with a microphone and playing the inverse of it. That process has a shape: it works well over a certain range of frequencies and poorly outside it, because the maths gets harder as the wavelength gets shorter.
Plot attenuation against frequency for any pair of cancelling headphones and you get a curve, not a line. The manufacturer’s headline number is the highest point on that curve.
- 31.4dB
- Atlas II at 100 Hz
- 19.2dB
- Atlas II at 4 kHz
- 12.2dB
- The gap between them
Where rail rumble sits
Where voices sit
Twelve decibels is not a rounding error. It is roughly a four-fold difference in perceived loudness, from the same pair of headphones, depending on what noise you are sitting in.
The low end is the hard part, and the important part
Engine and rail rumble is concentrated below about 200 Hz. So is road noise, air conditioning, and most of what makes an aircraft cabin exhausting.
It is also the region where a cheap cancellation system falls apart, because low frequencies have long wavelengths and the microphone-to-driver timing has to be right over a bigger physical distance. This is why we quote every figure at 100 Hz first, and why a headphone that is excellent at 1 kHz and mediocre at 100 Hz is mediocre on a commute.
Above 2 kHz, the electronics stop mattering
High frequencies are attenuated mostly by physics — the mass of the earcup and the quality of the seal — rather than by processing.
That has a practical consequence. If the noise you want to escape is an open-plan office, where the problem is speech at 500 Hz to 4 kHz, an expensive cancelling headphone will not help much more than a well-sealed passive one. If the noise is a train, it will help enormously. Buy for your noise, not for the number.
The test nobody runs: does it change the sound?
A cancellation system is inside the signal path, and a badly implemented one alters the frequency response when you switch it on — usually by adding bass, because that flatters the demo.
We measure the response with cancellation on and off and publish the difference. The Atlas II shifts by 0.6 dB, which is inaudible. The class average is 2.4 dB, which is not. A headphone that changes its tuning depending on a button is a headphone you cannot use for anything where accuracy matters.
Wind
No current system handles it. The microphones that measure the incoming noise are also measuring turbulence across their own ports, and the result is a roar that the system then tries to cancel and cannot.
Every pair we have tested has this problem above about 15 km/h of apparent wind. Some have a wind-reduction mode, which works by turning the cancellation down. That is not a criticism of any particular product; it is the state of the art.
What to ask instead
- At what frequency is that decibel figure measured?
- What is the attenuation at 100 Hz?
- Does the frequency response change when cancellation is switched on?
- Are the earpads replaceable, given the seal does half the work?
Any manufacturer that can answer the first two has done the measurement. Most cannot.
