Electronic Design

Challenges Await In Noise-Cancelling Headset Design

Up to 30% of adult hearing loss may be due to noise.1 One of the effects of hearing loss is the inability to selectively hear in the presence of background noise. Headsets that cancel noise will help save our hearing and enable us to have better conversations, more peaceful airplane flights, and overall higher quality of life.

“Passive” noise cancelling mechanically insulates the ear from the environment with earplugs or earmuffs made of plastic, foam, rubber silicone, or some other material. Depending on materials used, passive cancellation starts affecting noise from 500 Hz. Active noise-cancelling (ANC) techniques are effective from frequencies as low as 20 Hz up to a few kilohertz. The “active” solution fills the gap that typical passive solutions don’t fulfill.

Active noise cancellation uses sound generated from a speaker to cancel other sounds arriving at the ear. Two alternatives exist for detecting unwanted sounds: the design exposes the microphone to the noise and isolates it from the speaker, or it places the microphone as close as possible to the speaker (Fig. 1).

1. In a feed-forward noise-cancelling headset design (a), the microphone is isolated from the speaker. In a feedback design (b), the microphone is placed near the speaker. Feed-forward and feedback can be combined in the same headset (c).

With the feed-forward approach, if the acoustic isolation between the speaker and the feed-forward microphone is good enough, there will be no influence on the playback path. This is a great advantage when the noise cancellation is switched off or the battery runs flat and the headset works in passive mode. The music playback still sounds good. The feed-forward topology also makes it very simple to create an “assisted” hearing mode where the immediate acoustic environment can be amplified, removing the need to take the headset off to listen to someone.

Designing For Feed-Forward

With a feed-forward topology, the designer analyzes the acoustics of the headset to determine how the noise is affected by the time it reaches the ear, in terms of frequency, phase, and amplitude. This transfer function, G(w), is then modeled electrically and inserted between the microphone and speaker.

Feed-forward designs can be subject to directional issues, so the microphone must be omnidirectional. Also, the noise channel can’t be mechanically concentrated. Because the microphone must acquire the noise before it gets to the ear, parallel acoustic paths must be minimized.

Designing For Feedback

A feedback topology acts on the actual noise at the ear, which, by definition, is a more accurate representation of the noise compared to the simulated effects used by feed-forward. However, the speed of the system’s response constrains performance. Special attention is required when defining the transfer function to avoid introducing instabilities, such as positive feedback.

Feedback designs subtract the playback from the microphone. The resulting signal represents the noise around the ear. Using a transfer function, a phase-shifted version of the signal then is used to cancel the unwanted sound.

An ideal headset for feedback ANC would have zero delay between the speaker and the microphone that picks up the speaker response. That cannot be achieved because then the speaker and the microphone would have to share the same acoustic center. From a physical point of view this is impossible to achieve. The best that can be done is to place the microphone as close as possible to the speaker membrane to minimize this propagation delay.

Because the microphone in the feedback headset cannot differentiate between noise and playback, headsets using feedback typically experience some signal distortion. To overcome this, it’s common to add an equalization circuit in the playback path to ensure the sound is the same with and without ANC activated.

It’s possible to combine feed-forward and feedback in the same headset (Fig. 1c). This produces the advantage of both approaches, but introduces additional design complexity and cost. The table assesses the tradeoffs.

Practical Design Considerations

The biggest challenge with feed-forward solutions is ensuring the controlled environment around the user’s ears. Different users will have different ear shapes and sizes, and their headsets will fit differently. These different physical fits require different transfer functions if they are to achieve optimal noise cancelling.

This implies that when designing a headset for feed-forward, the headset has to fit everyone well, or it has to be designed so the effects of variations from head to head have minimum effect on the transfer function. A controlled leakage path accounts for this in feed-forward headsets.

When considering feedback, the main issue is where to place the microphone and how to ensure a controlled acoustic behavior across the frequency range. This is subtle, but it is not uncharted territory. Numerous patents exist concerning the position of the microphone and the dimensions/ratios of the front and back cavities.

In making design tradeoffs, it is common to prioritize the audio response, in which case the ANC performance fits as best it can. Under these conditions, the ANC behavior is typically limited in frequency band, but can achieve adequate or even significant noise-cancelling levels. This approach ensures that the headset sounds good even when the battery has run out.

The alternative is to prioritize ANC performance, in which case the audio response has to be corrected to sound good. In this case the headset should always remain powered, or alternatively some passive in-line equalization will be needed to recreate the “pleasant” sound.

Analog Vs. Digital

The designer can choose either a digital or an analog signal processing approach. The traditional digital solution relies primarily on prediction and enables noise cancelling only for “steady state” noise, such as constant engine noise or other constant-frequency sounds.

An ANC headset must minimize latency to work with non-uniform noise. The typical distance between the microphone and the speaker is 0.7 cm, equating to a 20-μs sound-propagation delay. For a digital implementation, there is only a 20-μs interval for analog-to-digital conversion, signal processing, and digital-to-analog conversion (Fig. 2).

2. Given a microphone-to-speaker distance of 0.7 cm, the signal chain propagation delay cannot exceed 20 μs.

The typical power consumption for a standard 150-MIPS, 24-bit DSP and an analog-to-digital converter (ADC) and digital-to-analog converter (DAC) that meet the timing requirements is approximately 113 mW. In contrast, a fully analog-based implementation with similar if not better performance, such as the AS3400 from ams, consumes less than 10 mW. Battery-life considerations tend to favor an analog approach.

On the other hand, the development of the solution in digital-based approaches becomes more mathematical and software-oriented. Also, the designer can more easily introduce features such as equalization, bass boost, and surroundsound—at the cost of additional computing and power consumption, of course.

Although the analog approach is more empirical and relies on a good deal of pre-existing electro-acoustic engineering expertise, it rewards designers who can apply that expertise with one other advantage beyond battery-life, size, and cost: they react instantaneously.

Levels Of Noise Cancelling

Above 1 or 2 kHz, every headset exhibits some form of passive noise cancelling because the insulating materials block ambient noise filtering through to the ear. Also, 20 to 30 dB of noise isolation is common above a few kilohertz. Good noise-isolating earphones boast passive noise filtering down to very low frequencies, exhibiting 5- to 30-dB noise isolation. Even without being powered up, then, the headset provides peace and quiet.

However, there’s a drawback to good passive noise cancelling: a phenomenon called occlusion. It’s best experienced by putting your fingers in your ears and talking. You hear your own voice dampened and deformed due to the sound transiting through the bone structure of your jaw and nose.

This effect typically leads you to raise your voice when talking, to the point of shouting, which is neither discreet nor comfortable. Other than removing the headset or ear buds when you’re speaking, workarounds include a technique called Side-Tone, as your voice is picked up by an independent microphone and played back slightly attenuated.

One unexpected advantage of ANC is that it can replace the need for high passive filtering at lower (i.e., speech band) frequencies. Since a user can turn an active solution on or off by flicking a switch, this can greatly simplify situations such as talking to your airline-seat neighbor when he asks you to stop the flight attendant while the cart is going by.

Feed-forward solutions have a sweet spot, that is, a narrow band of frequencies where the noise-cancelling performance can be maximized (Fig 3a). In addition, feed-forward makes it possible to cancel noise over a broad frequency spectrum, typically up to 4 or 5 kHz.

3. Battery-life considerations tend to favor an analog design, but power consumption is only one factor in design tradeoffs.

Feedback ANC tends to be more homogeneous across its frequency band (Fig 3b). However, the frequency band in question is typically limited to around 1 kHz. This limitation is driven by the need for stability to ensure that different people or different headset pressures don’t cause sufficient transfer function changes to initiate positive feedback—hence, oscillation. Noise cancelling up to 20 dB and 1-kHz bandwidths are common for feedback designs.

Hybrid designs benefit from the positive sides of both topologies while minimizing their drawbacks (Fig. 3c). For more information, visit www.ams.com/anc.

Design Example

During the product concept phase, an OEM engineer must look at all the tradeoffs and determine whether the product will use feed-forward, feedback, or hybrid active noise cancelling as well as the materials for the mechanical design.

The next step is to define the shape, size, and fit of the headset. ANC component suppliers can provide advice on the position of microphone, venting and leakage holes, and other features related to the acoustic architecture of the headset. ANC chips are available that integrate all the key functions and offer the necessary performance. The AS34x0 from ams features everything needed for a feed-forward design (Fig. 4).

4. Feed-forward solutions (a) generally have a narrow band of frequencies where the noise-cancelling performance can be maximized. Sweet spots in excess of 25 dB and bandwidths up to 4 kHz are common. The behavior of feedback ANC (b) tends to be flatter. Hybrid designs (c) provide the best of both worlds.

Next, the headset developer transforms 3D models into electroacoustic solutions during the prototyping phase. At this point, the OEM engineer decides on possible modifications to the sound of the headset while the IC vendor can participate in the development of the first ANC filter using a selection of simulation tools and development kits.

The output of this stage is a fully functional headset presenting the best ANC performance possible from this first prototype, accompanied by the respective frequency response graphics. Most vendors supply an extensive reference schematic for the circuit diagram together with suggestions for further improvement of the mechanical design of the headset to increase the ANC performance. This prototyping stage is typically completed in one or two iterations.

The final step is to integrate the approved modifications and design and layout the printed-circuit board (PCB). Once the production conformation sample is complete, the ANC transfer function can be fine-tuned.


9th International Congress on Noise as a Public Health Problem (ICBEN) 2008

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.