Audio processing algorithms improve speaker sound quality

Modern smartphones are sleek and powerful, and while the size of the phone varies depending on the model, in general, a state-of-the-art device packs a lot of features into a package that measures approximately 110x60x15mm.

If you factor in the display and circuit boards, there is not much room left for speakers. Now, imagine the space occupied by a subwoofer speaker in a home theater. Most people would think that these are two completely different things and even incomparable. In a way, it is understandable. However, the fact is that even though they are indeed two completely different applications, the content they run is becoming more and more similar. High-speed mobile communication technology (3G, 3.5G, 4G) and its supporting networks have enabled the download and playback of audio and video on mobile phones. Mobile phone users want higher bandwidth and better audio and video quality.

The problem is that improving audio quality is not easy. Handheld device manufacturers face many constraints, two of which are the size of the phone and the degree of compression of the audio files. Let's discuss these two aspects.

Dimensions

A speaker converts electrical energy into sound waves by moving a diaphragm back and forth. The diaphragm pushes air, creating sound waves that are converted into sound by our ears. Given the size constraints mentioned above, there is not much room for movement in a phone, so only small speakers with very small diaphragms are used, which only allow for a small range of movement.

In static integrated circuits, speakers are a bit cumbersome because they need to move. Small speakers cannot produce good audio effects, and bass frequencies are the most affected. Getting high-quality audio effects from small portable consumer electronics is indeed a challenge, and to meet this challenge, it can only be achieved by a cross-functional team of designers from the fields of industrial, electromechanical, and electronics. Electronic engineers have come up with this initiative: use audio processing algorithms.

Compressed Audio

Audio is usually compressed into smaller files for users to download. File compression is achieved through encoding algorithms (such as MP3). The reduction of the file size may cause loss of information, which ultimately affects the audio quality. In this case, audio processing algorithms can also come in handy.

Audio processing algorithms

Currently, there are various algorithms for processing audio signals and improving the listening experience.

The basic processing algorithm is to overcome the speaker's imperfections by filtering the amplitude variations of different frequency bands through an equalizer. By looking at the frequency response of the speaker, we can determine what can be reproduced and what cannot, and then we can design the equalization curve accordingly. The goal is to obtain audio with constant amplitude, regardless of the frequency of the speaker.

The use of basic equalization is now commonplace, and most audio converters sold on the market use equalization. Unfortunately, sometimes this is not enough to improve audio quality. In fact, speakers have a frequency response that changes with the strength of the audio signal (see Figure 1).

Figure 1: Loudspeaker + speaker frequency response and signal level distortion. [page]

To compensate for this effect, dynamic filters must be used. The frequency response of the speaker changes with the signal amplitude, and the poles and zeros of these filters will also change with it. When implementing dynamic filters, DSP-like processing is essential. Most low-power audio converters are not powerful enough to achieve this.

Another interesting algorithm is Bass Boost, which improves the reproduction of bass frequencies by exploiting the acoustic principle of missing fundamental.

Looking at the frequency response of small speakers, we can see that their bass response is 3 dB, which extends to several hundred hertz. This means that such speakers cannot reproduce lower frequencies very well. Driving the speaker with these low frequencies is meaningless (the speaker cannot reproduce these low frequencies) and can even be harmful. The low frequencies will force the speaker to move beyond its capabilities, which will ultimately cause more distortion in the higher frequencies.

Bass boost (see Figure 2) takes the bass content that the speaker cannot reproduce and raises it an octave higher to where the speaker can work well. For example, if the speaker is 3 dB above 300 Hz and the content is only 200 Hz, the bass boost will raise it to 400 Hz so that it can be played. Considering that the audio content is an octave, the human ear and brain will be tricked into thinking that low frequency content is being heard (the missing fundamental principle). Now, we can use a filter to remove all of this low frequency content that cannot be reproduced so that it does not reach the speaker. The simultaneous use of bass boost and high pass filtering will greatly improve the bass reproduction capabilities of small speakers.

Figure 2: Bass enhancement principle.

Audio can also be improved through virtualization, also known as 3D. This enhances the audio played through speakers or headphones by creating an immersive listening experience. Virtualization algorithms enlarge the sound, effectively creating virtual surround sound even for small portable devices. They analyze the similarities and differences of the audio played through the two channels of a stereo system and enhance them to make the user believe that the sound is coming from all directions. The algorithm uses what is called a human brain-related transfer function (HRTF), which explains how sound interacts with and is interpreted by the human brain, ears, and brain system.

Other algorithms focus primarily on improving compressed audio. In this case, they try to recover information that was lost during the compression process. They tend to work particularly well with high-frequency content (around 1 kHz), improving clarity. This allows high-frequency sounds, such as the sound of rain in a movie or a guitar solo in a song, to be reproduced with lifelike clarity.

Many audio converters (ADC, CODEC and DAC) support advanced audio processing functions. At TI, these algorithms are run in the audio digital signal processor (DSP or miniDSP), which is integrated into the audio converter. This mini digital signal processor is programmed in the PurePath Studio graphical development environment. The TLV320AIC36 is one of the devices that can be used in many mobile phone products due to its analog input and output characteristics.