Optimize audio transmission and playback in TWS headphones

As interest in high-definition audio continues to rise, huge demand for high-definition TWS earphones with advanced features is reaching a peak. This article explains the technology behind high-definition music transmission and how audio designers can meet the growing demand.

Industry experts note a steady increase in demand for high-definition audio wireless headphones. As people of all ages, including those with age-related hearing loss, seek a full HD sound experience, the research also points to an increasing demand for hearing personalization to bridge the gap between audio and hearing. These trends are driving the development of high-definition audio support at every stage of the music delivery chain.

The Evolution of High Definition Audio

The term high-definition audio (or high-resolution audio) has no strict technical definition. It is often used to describe audio systems that support higher data rates than were available in the early days of such equipment adoption. The term was originally used to describe digital audio systems that could support higher data rates than the compact disc (CD) standard. This included alternative disk formats, and later files containing digital audio recordings. It has also been applied to audio streaming and, more recently, to wireless headphones that are capable of providing better than typical audio quality.

Improvements in audio recording and distribution have increased the data rates available to mobile listeners. Wireless headphones have lagged behind due to Bluetooth limitations. However, newer Bluetooth codecs enable wireless headphones to deliver high-definition audio. This places new demands on audio hardware, including drivers. This article will explain technical terms related to high-definition audio, describe improvements in the audio chain from source to headphone, and how the use of balanced armature (BA) tweeters and dynamic woofers combined with ANC, occlusion reduction, and hearing personalization.

Technical Background

Digital audio formats are most simply described by their sampling rate and bit depth. Digital audio represents analog sound by sampling the amplitude of the signal rapidly in time. The amplitude of each sample is measured and stored as a binary number. The number of samples per second is the sampling rate. The size of the binary number used to describe the amplitude is the bit depth.

Sampling rate

Figure 1: Sampling rate and bit depth in digital audio.

Information theory, attributed to Nyquist (and Shannon and Whittaker), states that we must sample a sine wave at least twice per cycle to accurately capture its information. Otherwise, the reconstructed output will be an alias at the wrong frequency, as shown in Figures 2a, 3, and 2b. The alias is a frequency domain image centered at the "Nyquest frequency" (1/2 the sampling rate).

For music, the sampling rate must be at least 2 times the highest frequency to be reproduced. So if the maximum audio frequency to be represented is 20 kHz, a sampling rate of 40 kHz is required. This requires a very sharp 20 kHz low pass filter on the ADC. Any audio above 20 kHz that passes the filter will be recreated as audible aliases when converted back to analog, reducing the fidelity of the music. Because no filter is perfect, in real world practice some headroom is provided between the highest frequency to be represented and the sampling limit. In the compact disc format, the audio to be sampled is low pass filtered to 20 kHz. The sampling rate is set to 44.1 kHz, slightly above the theoretical limit of 40 kHz, to provide filter headroom.

Figure 2a and 2b: Aliases in the time and frequency domains

Bit Depth

The size, or bit depth, of the words used to describe the samples determines how accurately each sample is digitized. Imagine describing a person's height by rounding it off. You could then say that someone is two or three meters tall. A second number would be better, describing someone as 2.1 or 2.2 meters tall. This is still a bit rough, but you can keep adding digits until you have enough resolution.

The same is true for digital audio. Each time an additional bit is used in the binary word, the amplitude can be described by twice the number of values. This in turn reduces the error by a factor of two. While one would intuitively think that rounding would cause distortion, it turns out that the error (quantization error) can be converted from distortion to noise by applying a little bit of noise called dither. Thus, a good recording system has no distortion and the noise floor drops by 6 dB for each additional bit of depth. The ratio between the loudest undistorted sound and the noise floor is 6 * bit depth.

Since digital systems are built on multiples of 8 bits, digital audio uses multiples of 8 bits to represent its word length. Very early computer audio used only 8 bits. Having a noise floor just 48 dB below the loudest music is not very practical. Compact discs support 16-bit depth, which provides a signal-to-noise ratio of 96 dB. This covers a very useful range. If the playback volume is set to a reasonable level, the remaining hiss between songs is below the background sound in the listening environment. As the volume is increased, some audible hiss may appear between songs or in quiet passages. Therefore, an argument can be made for using higher bit depths to reduce this component of the total system noise.

Figure 3: Effects of increasing sampling rate and bit depth

compression

The combined bit rate of compact discs is 16 bits/sample ∙ 44,100 samples/second ∙ 2 channels = 1,411 kilobits/second (kbps). This data rate was too large to be used in early digital audio players. It took too long to download songs and available memory was limited. To solve this problem, compression methods were developed. Pattern recognition can be used to provide a more concise description of the pattern of 1s and 0s in a data stream. This is the idea behind compressing data files in computers. Some files can be compressed to less than 1/10 of their original size and then restored later without loss.

Music files cannot be compressed significantly using these methods. Some improvement can be obtained by using a linear prediction algorithm. However, compression ratios still rarely exceed 2:1. Since no data is lost, these are called lossless encoders. Two popular examples are FLAC and Apple Lossless.

To further reduce the data rate, psychoacoustic models of human hearing can be used to minimize the audibility of the discarded data. It is common for one sound to hide another to prevent human perception. (See Figure 4.)4 By encoding the hidden sounds at a lower rate or discarding them entirely, the overall data rate can be reduced. While psychoacoustic compression methods have improved, even the best-designed encoders still result in some audible artifacts5. Lower compression rates provide fewer and gentler audible artifacts at the expense of larger files and higher data rates. This is the science behind the movement toward high-definition audio streaming, made possible by the ability to transmit music at higher speeds.

Expanded bandwidth

The terms "High Definition Audio" and "High Resolution Audio" are often used synonymously, but this can lead to confusion. "High Definition Audio" is used to describe any audio that has a higher data rate than is traditionally used. The Japan Audio Society (JAS) authorizes the use of the "Hi-Res Audio" logo for hardware that meets certain requirements, including the ability to reproduce audio up to 40 kHz. (The exact requirements can be licensed from JAS).

Figure 4: Figure 4: Loud sounds mask the perception of quieter sounds with similar frequencies

High-definition audio is the result of any combination of increasing the sampling rate, increasing the bit depth, and reducing the compression rate. If the sampling rate is increased beyond 44.1 kHz, the digitized audio will be closer to the original signal, which in turn helps maintain fidelity throughout the transmission chain. Extending the bandwidth to above 20 kHz also becomes possible. Not all high-definition audio formats and devices have bandwidths beyond 20 kHz, but some do, especially those targeting JAS certification.

Music Communication Chain

In order for listeners to hear high-definition audio, every step in the pipeline must be of adequate quality. This includes the original song preparation process, download or streaming services, playback systems, headphone connections, and headphone drivers.

Recording and delivery

Music recording equipment is widely available, allowing storage at a variety of sampling rates and bit depths, many of which exceed compact disc specifications. Once music is recorded, it must be delivered to the user. File download services have supported high-definition audio for some time. More recently, most popular streaming services have announced or already offer lossless CD quality as a baseline, and increasingly have subsets of their libraries with high-bitrate high-definition, usually at no additional charge.

Music playback hardware

Today, people can listen to downloaded or streamed audio using a variety of devices, including dedicated music players, mobile phones, PCs, and more. High-definition audio playback is widely available in these devices. The additional cost of supporting high-definition audio is relatively low, and sample rates up to 192 kHz and word lengths up to 32 bits are generally supported. In some cases, the analog headphone outputs of these devices (if available) are limited to 20 kHz, and typically do not have noise performance better than 16 bits. However, in these cases, an external DAC can be plugged into the digital port to obtain higher performance.

Headphone Codec

The final step in the chain is the earphones. In wireless earphones, a key limiting factor is the Bluetooth radio link. The Bluetooth codec (COder/DECoder) reduces the data rate before transmission to the earphones. To meet the growing demand for high-quality audio in wireless earphones and TWS, codecs with higher data rates have been introduced. (Table 1.) In properly designed earphones, these newer codecs can provide a higher-fidelity listening experience, even if not completely lossless. Going a step further, Qualcomm recently announced their intention to support lossless transmission of CD-quality audio over Bluetooth. 6